===============================================
                                   TM
               Help file for ProBID

 (c) Copyright 1993, 1994 by PROFIT Consultants
               All rights reserved
===============================================
           Developed by: Tarek Hegazy


Overview of ProBID
*Overview of ProBID                                
 
* NOTE: Use  <F1>  at any  MENU item to see the * 
*       HELP and have a quick ProBID tour.      *
*  *

ProBID is a software for predicting construction
outcomes and for supporting your bid decisions.

ProBID uses advanced Neural Network research 
of computer science to "intelligently" recognize 
the  risk  pattern  of your  upcoming  project. 
ProBID then matches your project environment 
with a number of stored cases of successful and
unsuccessful projects. Accordingly, ProBID 
predicts some indicators of the project's 
potential success or failure:
 
* possible construction delay;
* level of claims and disputes; 
* level of change orders; and
* level of actual profitability.

In addition, ProBID suggests a markup strategy 
for your bid and evaluates your chances of 
winning. 

ProBID predictions direct your attention to 
potential problem areas that you may consider
to adjust your estimate, think of alternative 
decisions, and take early counter measures to
help assure a successful bid.

ProBID is NOT a pure theoretical model. Rather, 
it is developed based on the experience of a 
large number of REAL-LIFE projects that were 
collected from general contractors in the United 
States and Canada. Although ProBID was initially 
intended for building projects, it is designed 
with a powerful ADAPTATION option that builds on 
your own PAST-projects' experience and enables 
you to develop custom neural networks that suit 
your particular work environment, locality, and 
types of projects.

ProBID organizes your past-projects' data in a 
complete and simple manner, incorporating 
subjective factors such as uncertainty and market 
conditions. ProBID stores these past experiences 
and utilizes it to enhance future predictions on 
new jobs.

ProBID can be used by general contractors,
construction managers, and owners alike. ProBID
is meant to supplement, rather than replace, 
your own gut feeling. ProBID helps you make 
informed bid decisions, identify possible 
conflicts and problems, as indicators of the 
project's potential success or failure.
                                <END>
END


BID PROJECT DATA
*Definition: Bid PROJECT
This is a project in its initial stages for which
you are using  ProBID  to predict its bidding and
construction outcomes.  Each Bid Project is saved 
in a separate DOS file ".BID".

*Load or Edit project...
 You can EDIT the data of an existing BID project
 through  5  input screens incorporating a number
 of factors describing the project environment:
                                                      
        (1) level of uncertainty;               
        (2) level of complexity;                
        (3) Competition and other Market
            factors;
        (4) your company's capabilities and how        
            desperate you need this job.            
                                                       
*Notes:                                             
   - Select PRODUCE PREDICTIONS from CONSULTATION
     menu to get predictions for the project.     
                                                       
   - After the bidding stage, you can add a BID
     project to the DATABASE of PAST-projects.
     This database can be used to adapt the
     prediction model and thus enhance future
     predictions.
                                       <END>
END

NEW BID PROJECT
*Definition: Bid PROJECT
This is a project in its initial stages for which
you are using  ProBID  to predict its bidding and
construction outcomes.  Each Bid Project is saved 
in a separate DOS file ".BID".

*New Project...
  You can start a NEW BID project and EDIT its
  DATA through 5 input screens that incorporate
  a number of factors describing the project
  environment:
                                                      
        (1) level of uncertainty;               
        (2) level of complexity;                
        (3) Competition and other Market
            factors;
        (4) your company's capabilities and how        
            desperate you need this job.            
                                                       
*Notes:                                             
   - Select PRODUCE PREDICTIONS from CONSULTATION
     menu to get predictions for the project.
                                                       
   - After the bidding stage, you can add a BID
     project to the DATABASE of PAST-projects.
     This database can be used to adapt the
     prediction model and thus enhance future
     predictions.                      
                                        <END>
END

DELETE A BID PROJECT
*Delete project...
                                      

  This option allows you to DELETE an existing
  BID project. A list of the BID projects is
  provided and you can select and confirm which
  project to delete.                  
                                         <END>
END

RENAME A BID PROJECT
*Save As...
                                                      

  This option allows you to SAVE a BID project
  to a different name, then you can edit its 
  data.                                
                                        <END>
END

SAVING AS A PAST PROJECT
*SAVING AS A PAST PROJECT
 
 This option allows you to ADD an existing BID
 project to the DATABASE of PAST-projects. You
 may do this after the end of construction. Once 
 a project is added to the database, you need to 
 edit it (under ADAPTATION menu) to input the 
 actual bidding and execution outcomes of the 
 project. You may also choose to select this 
 project to be used for training and developing
 CUSTOM neural networks (see ADAPTATION menu for 
 details).                            
                                        <END>
END

QUIT ProBID


  This will quit ProBID and return to DOS.         
                                    
                                    <END>
END


BUSINESS DATA
*Company Business Data
 
 These are company-dependent data, depicting
 your company's relative size in the market and
 your practice of allocating a percent markup. 
 This data is considered during consultation to 
 ensure that ProBID's predictions are compatible 
 with your own practice.      
                                        <END>
END
                
Project Data
*Project Data
  
 These are part of the data that describe the 
 Bid project at hand. The outcomes of this 
 project are predicted based on such data and 
 the neural network selected.
           

*Editing Keys
  Use the keyboard or click the
  left mouse button on:

         <F8>  --> REPORT to a printer or a file.
 [PgDn]/[PgUp] --> Next / Previous screen.  
                                                      
 Using a mouse, click on any input field using 
 the left button to edit that field. Right mouse 
 button brings back the bottom menu.

*Note: 
  After the bidding stage, you can add a project
  to the DATABASE of PAST-projects.            
                                       <END>
END


General Project Data
*General Project Data
 These are data that describe the type of project
 being considered, its contract type and contract
 duration. These data, in addition to those 
 included in the next screens, provide a detailed 
 description of the project environment including 
 its uncertainty level, complexity level, market 
 conditions, and your company's ability and need 
 for the job. All these data represent the input 
 to the prediction model.
------------------------------------------------  
*Note: more info. is in page 34 of the manual.
------------------------------------------------

*Editing Keys:
  Use the keyboard or click the
  left mouse button on:
        <F5>  --> EDIT Fields.                     
        <F8>  --> REPORT to a printer or to a 
                  file.
       [PgDn] --> MOVE to next screen.
 
 Using a mouse, click on any input field using 
 the left button to edit that field. Right mouse
 button brings back the bottom menu.

*Notes:                                           
   - Refer to page 41 of the manual for a note 
     about subjectivity in assessing the factors.
     
   - Select PRODUCE PREDICTIONS from CONSULTATION
     menu to get predictions based on the project 
     data.
                                                    
   - After the bidding stage, you can add a 
     project to the DATABASE of PAST-projects.            

                                       <END>
END

Project Uncertainty Data
*Job Uncertainty
------------------------------------------------  
*Note: more info. is in page 36 of the manual.
------------------------------------------------
 The project's level of UNCERTAINTY is determined  
 by a number of factors:                           
                                                     
*- SITE CONDITIONS: 
   clarity of soil reports; availability of 
   utilities; etc. 
*- OWNER ATTITUDE: 
   frequent changes; aggressiveness; and
   regularity of payments.                        
*- PROJECT LOCATION: 
   storage space; close/far; local by-laws; and 
   availability of local labour.
*- SAFETY HAZARD: 
   toxic materials; pollution; high elevations. 
*- INACCURACY IN ESTIMATE: 
   detailed vs rough estimating, bidding time 
   allowed.     
*- WEATHER SENSITIVITY: 
   site climatic conditions; nature of work 
   tasks.
------------------------------------------------

*Editing Keys:
  Use the keyboard or click the
  left mouse button on:
         <F5>  --> EDIT Fields.                     
         <F8>  --> REPORT to a printer or to a 
                   file.
 [PgDn]/[PgUp] --> Next / Previous screen.
 
 Using a mouse, click on any input field using 
 the left button to edit that field. Right mouse
 button brings back the bottom menu.
 
*Notes:                                             
   - Refer to page 41 of the manual for a note 
     about subjectivity in assessing the factors.
     
   - Select PRODUCE PREDICTIONS from CONSULTATION 
     menu to get predictions based on input data. 
                                                       
   - After the bidding stage, you can add a 
     project to the DATABASE of PAST-projects.            
                                       <END>
END

Project Complexity Data
*job complexity
------------------------------------------------  
*Note: more info. is in page 37 of the manual.
------------------------------------------------
 The project's level of COMPLEXITY is determined
 by a number of factors:                           
                                                     
*- TECHNOLOGY NEEDED: 
   available vs. required technology.
*- RESOURCES NEEDED: 
   small vs large amount of resources required.
*- JOB SIZE: 
   small vs large ($) and duration.     
*- QUALITY OF DESIGN: 
   clarity of drawings; accuracy of measurements.
*- STACKING OF TRADES: 
   number of people working in the same area;
   site organization.                   
*- % SUBCONTRACTED: 
   the number of trades involved and the need 
   for coordination and management.
*- RIGIDITY IN SPECIFICATIONS: 
   strictness in following the specifications.
------------------------------------------------                                                    

*Editing Keys:
  Use keyboard or click the
  left mouse button on:
         <F5>  --> EDIT Fields.                     
         <F8>  --> REPORT to a printer or to a 
                   file.
 [PgDn]/[PgUp] --> Next / Previous screen.
 
 Using a mouse, click on any input field using 
 the left button to edit that field. Right mouse
 button brings back the bottom menu.
 
*Notes:                                           
   - Refer to page 41 of the manual for a note 
     about subjectivity in assessing the factors.
     
   - Select PRODUCE PREDICTIONS from CONSULTATION
     menu to get predictions based on the project 
     data.
                                                    
   - After the bidding stage, you can add a 
     project to the DATABASE of PAST-projects.            
                                       <END>
END

Market Conditions
*Market Conditions
------------------------------------------------  
*Note: more info. is in page 39 of the manual.
------------------------------------------------
 The  MARKET  CONDITIONS at the time of project    
 execution is determined by a number of factors:   
                                                     
*- INFLATION RATE: 
   high vs low rate of change.       
*- ESCALATION RATE: 
   high vs low rate of change.      
*- COMPETITION: 
   how severe is expected competition.
*- ECONOMIC GROWTH: 
   boom vs recession.               
*- RESOURCE AVAILABILITY: 
   availability of skilled labour; technical 
   staff; and general labours.  
------------------------------------------------

*Editing Keys:
  Use the keyboard or click the
  left mouse button on:
         <F5>  --> EDIT Fields.                     
         <F8>  --> REPORT to a printer or to a 
                   file.
 [PgDn]/[PgUp] --> Next / Previous screen.
 
 Using a mouse, click on any input field using 
 the left button to edit that field. Right mouse
 button brings back the bottom menu.
 
*Notes:                                           
   - Refer to page 41 of the manual for a note 
     about subjectivity in assessing the factors.
   
   - Select PRODUCE PREDICTIONS from CONSULTATION
     menu to get predictions based on the project 
     data.
                                                    
   - After the bidding stage, you can add a 
     project to the DATABASE of PAST-projects.            
                                          <END>
END


Firm Ability and Need
*Firm Ability & Need for Work               
------------------------------------------------  
*Note: more info. is in page 40 of the manual.
------------------------------------------------
 your company's ability to do the job and your 
 need for the job are determined by several 
 factors:
                                                      
*- SIMILAR EXPERIENCE: 
   how familiar is the type of project to your 
   company's experience.             
*- MGMT & SUPERVISION: 
   utilization of scientific management practice; 
   follow-ups; and control.            
*- CONFIDENCE IN WORK FORCE: 
   experience of working staff and the trade-
   subcontractors involved.             
*- FINANCIAL CAPABILITY: 
   company's assets; bonding; and 
   cash availability.
*- NEED FOR WORK: 
   level of current work-load.     
------------------------------------------------

*Editing Keys:
  Use keyboard or click the
  left mouse button on:
         <F5>  --> EDIT Fields.                     
         <F8>  --> REPORT to a printer or to a 
                   file.
 [PgDn]/[PgUp] --> Next / Previous screen.
 
 Using a mouse, click on any input field using 
 the left button to edit that field. Right mouse
 button brings back the bottom menu.
 
*Notes:                                           
   - Refer to page 41 of the manual for a note 
     about subjectivity in assessing the factors.
   
   - Select PRODUCE PREDICTIONS from CONSULTATION
     menu to get predictions based on the project 
     data.
                                                    
   - After the bidding stage, you can add a 
     project to the DATABASE of PAST-projects.            
                                         <END>
END


Project Outcomes
*Actual Project outcomes               
------------------------------------------------  
*Note: more info. is in page 53 of the manual.
------------------------------------------------
  These are the ACTUAL bidding and construction 
  outcomes of a PAST project. The outcomes 
  include:               
                                                      
  1- The (%) Markup you decided;                      
  2- Bid result: wether you did Win/Lose the bid;
  3- The Difference between the winner and the 
     second lowest bidder (Money left on the 
     table);
  4- Experienced level of change orders (H-M-L);  
  5- Experienced level of claims (H-M-L);         
  6- Actual project duration (months); and        
  7- Actual profitability achieved (H-M-L-loss).  
                                                      
*Editing Keys
  Use keyboard or click the
  left mouse button on:
          <F1>  --> Help.           
          <F2>  --> Go to project number...           
      <F3><F4>  --> Previous / Next project.           
          <F5>  --> EDIT Fields.                     
          <F6>  --> ADD a NEW project...             
          <F7>  --> DELETE project...                
          <F8>  --> REPORT to a printer or file.
  [PgDn]/[PgUp] --> Next / Previous input screen.    
 
 using a mouse, click on any input field using 
 the left button to edit that field. Right mouse
 button brings back the bottom menu.

*Note:
  To select a project for TRAINING, edit the   
  project <F5> and type "Y" for the last field.
                                        <END>
END
                       
Derived Predictions
*Derived Predictions
------------------------------------------------  
*Note: more info. is in page 43 of the manual.
------------------------------------------------
 Predictions are based on the neural network you   
 selected and your assessment of the factors that  
 describe the project. The predictions include:    
                                                     
* 1- Estimate of a (%) Markup;                    
* 2- Estimate of your chance to Win/Lose the bid
     at the estimated markup;
* 3- ($) Money left on the table: 
     this is an estimate of the difference 
     between the winner and the 2nd lowest 
     bidder. This will help maximize your profit 
     while being a winner;
* 4- Estimate of possible change orders (H-M-L);  
* 5- Estimate of possible claims (H-M-L);         
* 6- Expected actual duration (months): 
     this, compared with the contract duration
     gives you the expected project delay; and       
* 7- Estimate of your actual profitability 
     (H-M-L-loss).
------------------------------------------------            

*Editing Keys
  Use the keyboard or click the
  left mouse button on:
         <F8>  --> REPORT to a printer or a file.
 [PgDn]/[PgUp] --> Next / Previous input screen.    
 
 Using a mouse, click on any input field using 
 the left button to edit that field. Right mouse
 button brings back the bottom menu.
                                                       
*Notes:                                              
 - Select "Sensitivity Analysis" from 
   CONSULTATION menu to examine how the 
   predictions may vary with changes in your
   assessment of the project factors. This 
   option also will provide you with an estimate
   of your chances of winning at different 
   markup levels. This will help you decide a
   markup that balances between winning and 
   profit.
                                                       
 - After the bidding stage, you can add a
   project to the DATABASE of PAST-projects.             
                                     <END>
END                                                       


PRODUCING PREDICTIONS
*Producing Predictions
------------------------------------------------  
*Note: more info. is in page 43 of the manual.
------------------------------------------------

 ProBID predicts the outcomes of your project
 based on the prediction model (i.e., neural 
 network) you select and your assessment of 
 the factors describing the project environment.
 
 ProBID comes with a default neural network 
 "GENERAL.NNT" that is trained on actual projects 
 collected from general contractors in the 
 United States and Canada. The training cases 
 included various project situations and 
 their associated actual outcomes (either 
 successful or not). The TRAINED network can, 
 thus, produce predictions for a new job, based 
 on its closeness to the training cases. 
 
 The default network, as its name indicates, has 
 prediction ability that is general and not job 
 or contractor-specific. The predictions made by 
 the default network, therefore, may not readily 
 suit the type of projects you are mostly working 
 on. It is important, therefore, that you use the
 "ADAPTATION" menu to develop custom neural 
 networks that suit your specific type of jobs 
 (e.g., Office Buildings) or specific conditions 
 such as a particular owner, locality and/or 
 market. 
                                                     
 The predictions made by a neural network 
 concern seven project outcomes:                                  
 
* 1- Estimate of a (%) Markup;                    
* 2- Estimate of your chance to Win/Lose the bid
     at the estimated markup;
* 3- ($) Money left on the table: 
     this is an estimate of the difference 
     between the winner and the 2nd lowest 
     bidder. This will help you adjust your 
     markup to maximize profit while being 
     a winner;
* 4- Estimate of possible change orders (H-M-L);  
* 5- Estimate of possible claims (H-M-L);         
* 6- Expected actual duration (months): 
     this, compared with the contract duration
     gives you the expected project delay; and       
* 7- Estimate of your actual profitability 
     (H-M-L-loss).
                                         <END>
END

SENSITIVITY ANALYSIS RESULTS
*Sensitivity Analysis
------------------------------------------------  
*Note: more info. is in page 47 of the manual.
------------------------------------------------
 
 Sensitivity analysis examines how ProBID 
 predictions may vary with changes in your 
 assessment of the project factors. The analysis 
 is performed using the MONTE CARLO simulation 
 technique. The simulation generates a number of 
 scenarios (simulations) that are minor random 
 variations of the assessment you provided during
 the editing of the project data. All simulations 
 are then input to the prediction model (i.e., 
 neural network) you select, and predictions are 
 produced. As a result, the mean and standard 
 deviation in all scenarios will be reported as 
 the "MOST LIKELY" predictions for the project 
 outcomes. Refer to the manual for guidelines on 
 the number of simulations to use and how to 
 interpret the results.
                                                      
*Definition - Percent Markup
 ProBID considers markup as a percent of all the
 project estimated costs (direct + indirect), to 
 be added to your bid as an allowance for: 
 1) profit only; or 2) profit + contingency 
 (select "Input Business Data" from "Consultation 
 menu") to enter your preference. Since many 
 project variables may affect your markup 
 decision, it is not wise to set a fixed percent 
 to be used in every project. Generally, however, 
 the lower the percent markup, the higher your 
 chances of winning the job, however, the less 
 is your profit. 
 
*Markup vs Probability of Winning
------------------------------------------------  
*Note: more info. is in page 48 of the manual.
------------------------------------------------
 As a by-product of the sensitivity analysis, 
 ProBID helps you decide a percent markup that 
 makes your bid low enough to win the job, while
 being high enough to obtain the highest profit 
 possible. ProBID provides you with the
 relationship between your chances of winning  
 and the percent markup value you choose. This 
 relationship is provided in two ways: 

 1) considering that markup variation in all 
    simulations follows a DISCRETE distribution; 
 
 2) considering that markup variation in all
    simulations follows a NORMAL distribution.
 
 You may select either option depending on the 
 shape of the markup histogram provided as a 
 result of the MONTE CARLO simulation. Using the 
 graphics screens provided, use <right> and 
 <left> arrow keys to increment the markup value
 and view its corresponding probability of 
 winning. Based on that, you can decide a percent
 markup associated with your desired probability 
 of winning.              
                                       <END>
END                                                      


PAST PROJECTS DATABASE
*Definition: Past Project
 A Past project is one that has already been 
 executed by your company. Past-projects may 
 include:                        
  
  - Projects that experienced no execution 
    problems.
  - Projects that experienced execution problems   
    such as claims, change orders, delays, 
    and/or low actual profit or loss.         
 
 All Past-Projects are saved in one Database file 
 "Past.dat". The data of a Past Project is input 
 through 6 screens. Five of those screens 
 describe the project environment (similar to a 
 bid project), in addition to an input screen of
 the ACTUAL outcomes experienced in the project 
 (change orders, claims,...etc). Note that a bid 
 project that passes the bidding and construction 
 stages may be added to the database of past-
 projects using the "PROJECT" menu.

*Editing Keys
  Use the keyboard or click the
  left mouse button on:

         <F1>  --> HELP.                            
         <F2>  --> Go to project number...           
     <F3><F4>  --> Previous / Next project.           
         <F5>  --> EDIT Fields.                     
         <F6>  --> ADD a NEW project...             
         <F7>  --> DELETE project...                
         <F8>  --> REPORT to a printer or a file.
 [PgDn]/[PgUp] --> Next / Previous input screen.    
                                                       
 Using a mouse, click on any input field using 
 the left button to edit that field. Right mouse 
 button brings back the bottom menu.

*Notes: 
 - To build a proper model, your past projects 
   do not all have to be successful ones. cases
   that were unsuccessful help prevent similar 
   undesirable outcomes.

 - To select a project for TRAINING, edit the
   project and type "Y" for the last field.


DEVELOPING CUSTOM PREDICTORS                                                      
------------------------------------------------  
*Note: more info. is in page 51, 57 of the manual.
------------------------------------------------
 ProBID utilizes your database of past-projects
 to represent your company's work environment and 
 your own assessment of it.  
                                                    
 ProBID allows you to TRAIN the present model on 
 some of your past projects to create custom 
 prediction models (i.e., neural networks) that 
 suit your specific types of projects. For 
 example, you can develop a custom network suited 
 for predicting the outcomes of SCHOOL projects. 
 Alternatively, it can trained on OFFICE 
 BUILDINGS in the NEW YORK area, for instance. 
 
 Custom networks can be developed in two ways: 
 
  1- adding your past-project cases to a group of 
     default cases, thus, adapting the general 
     model to your own environment. This is 
     beneficial if you have small number of
     training cases (a relatively new firm), and
     if you are not confident of you own bidding
     practice; or
  
  2- solely using your own past-project cases in
     the development of a custom neural network.
     You may select this option if you are an 
     established firm that has a lot of past 
     projects experience. ProBID will help you
     organize your projects' data, preserving 
     that knowledge and utilizing it to support
     your bidding decisions and to predict the
     success or failure of new jobs.

*Procedure
 The process of developing the custom neural 
 network is as follows:

1- First, you input some of your PAST-projects 
   of a certain type (e.g., tunnels) into the 
   DATABASE of PAST-projects (use ADAPTATION    
   menu, Input Past-Projects ).
 
2- Check the sufficiency of training cases using
   the second option under ADAPTATION.
 
3- Start TRAINING either by adapting default
   knowledge or using only your own cases.
 
4- After training, use the new network for 
   predictions on new jobs of the same type.
 
5- Use ProBID to develop custom-trained networks 
   for other project types. Such networks will 
   have a well-defined and limited domain on 
   which they work best.


*Notes on TRAINING:                                
 - Refer to page 57 in the manual on how to 
   speed up training and avoid "over training".
                                               
 - The process of training usually takes some
   time (minutes to hours), depending on your  
   machine's speed as well as on the number of 
   training cases used.
    
 - The practicality of your CUSTOM network is   
   largely dependent on the number of PAST      
   projects you use for training the network,   
   the completeness and correctness of the      
   projects' data, and the suitability of the
   training projects to the domain for which    
   the network will be producing predictions.   
                                                    
 - After a training session, you have the option
   to view the errors of the TRAINED network,   
   accordingly, you can proceed with additional 
   training, quit, or delete the network if it  
   is not satisfactory. It is noted, however,   
   that over-training to very low errors might  
   not be suitable for generalized predictions. 
   The network in this case becomes very specific
   to the training examples and not others.
                                         <END>
END


Adapt Default Knowledge
 
 ProBID allows you to TRAIN the present model on 
 some of your own past-projects to create custom 
 prediction models (i.e., neural networks) that 
 suit the specific types of projects you mostly
 work on. For example, you can develop a custom 
 network suited for predicting the outcomes of 
 SCHOOL projects. Alternatively, it can trained 
 on OFFICE BUILDINGS in the NEW YORK area, for 
 instance.
 
 One way of developing a custom network is by
*Adapting Default Knowledge, which means:
 
     adding your past-project cases to a group of 
     default cases, thus, adapting the general 
     model to your own environment. This is 
     beneficial if you have small number of
     training cases (a relatively new firm), and
     if you are not confident of you own bidding
     practices.

*Training Procedure
 The process of developing the custom neural 
 network is as follows:

1- First, you input some of your PAST-projects 
   of a certain type (e.g., tunnels) into the 
   DATABASE of PAST-projects (use ADAPTATION    
   menu, Input Past-Projects ).
 
2- Check the sufficiency of training cases using
   the second option under ADAPTATION.
 
3- Start TRAINING either adapting default
   knowledge.
 
4- After training, use the new network for 
   predictions on new jobs of the same type.
 
5- Use ProBID to develop custom-trained networks 
   for other project types. Such networks will 
   have a well-defined and limited domain on 
   which they work best.

*Notes on TRAINING:                                
 - Refer to page 57 in the manual on how to 
   speed up training and avoid "over training".
  
 - The process of training usually takes some
   time (minutes to hours), depending on your  
   machine's speed as well as on the number of 
   training cases used.

 - The practicality of your CUSTOM network is   
   largely dependent on the number of PAST      
   projects you use for training the network,   
   the completeness and correctness of the      
   projects' data, and the suitability of the       
   training projects to the domain for which    
   the network will be producing predictions.   
                                                    
 - After a training session, you have the option
   to view the errors of the TRAINED network,   
   accordingly, you can proceed with additional 
   training, quit, or delete the network if it  
   is not satisfactory. It is noted, however,   
   that over-training to very low errors might  
   not be suitable for generalized predictions. 
   The network in this case becomes very specific
   to the training examples and not others.
                                        <END>
END


Build Your Own Model
 
 ProBID allows you to TRAIN the present model on 
 some of your own past-projects to create custom 
 prediction models (i.e., neural networks) that 
 suit the specific types of projects you mostly
 work on. For example, you can develop a custom 
 network suited for predicting the outcomes of 
 SCHOOL projects. Alternatively, it can trained 
 on OFFICE BUILDINGS in the NEW YORK area, for 
 instance.
 
 One way of developing a custom network is by
*Building Your Own Model, which means:
     
     solely using your own past-project cases in
     the development of a custom neural network.
     You may select this if you are an 
     established firm that has a lot of past 
     projects experience. ProBID will help you
     organize your projects' data, preserving 
     that knowledge and utilizing it to support
     your bidding decisions and to predict the
     success or failure of new jobs.

* Training Procedure
 The process of developing the custom neural 
 network is as follows:

1- First, you input some of your PAST-projects 
   of a certain type (e.g., tunnels) into the 
   DATABASE of PAST-projects (use ADAPTATION    
   menu, Input Past-Projects ).
 
2- Check the sufficiency of training cases using
   the second option under ADAPTATION.
 
3- Start TRAINING either by using only your 
   own cases.
 
4- After training, use the new network for 
   predictions on new jobs of the same type.
 
5- Use ProBID to develop custom-trained networks 
   for other project types. Such networks will 
   have a well-defined and limited domain on 
   which they work best.

*Notes on TRAINING:                                
 - Refer to page 57 in the manual on how to 
   speed up training and avoid "over training".
 
 - The process of training usually takes some
   time (minutes to hours), depending on your  
   machine's speed as well as on the number of 
   training cases used.
    
 - The practicality of your CUSTOM network is   
   largely dependent on the number of PAST      
   projects you use for training the network,   
   the completeness and correctness of the      
   projects' data, and the suitability of the       
   training projects to the domain for which    
   the network will be producing predictions.   
                                                    
 - After a training session, you have the option
   to view the errors of the TRAINED network,   
   accordingly, you can proceed with additional 
   training, quit, or delete the network if it  
   is not satisfactory. It is noted, however,   
   that over-training to very low errors might  
   not be suitable for generalized predictions. 
   The network in this case becomes very specific
   to the training examples and not others.
                                        <END>
END


More Training

 This option will continue with training the 
 custom neural network being developed. You can 
 decide on that if the network errors reached 
 after the previous training session are higher 
 than desired.
                                         <END>
END


PASSWORD
*Set or Change PassWord
  
  This option allows you to enter a PassWord to
  restrict access to the company's corporate DATA 
  (e.g., PAST-projects DATABASE).              
                                                      
  Only Password Holders can use ProBID to conduct     
  consultation and custom train the present 
  model. This may suit companies in which a Cost 
  Estimator only inputs the project data while 
  Managers make final bid decisions.                              
                                      <END>
END


TRAINING DATA VALIDATION

*Definition: Training cases
 these are selected projects of those in your 
 past-projects database. You can select a project 
 to be a training case by typing "Y" when you 
 are editing the last field of that past project. 
 Those training cases have to be of the same 
 category, for example "Hospital" projects. The 
 purpose is to use those projects to develop a 
 custom prediction model (i.e., neural network) 
 that suits the environment of hospital projects.
 
 You are advised to create different custom 
 networks for different types of projects or 
 location. Remember each time to edit all your 
 past-projects and select which projects to use 
 for training and which not.

*Training Data Validation               
------------------------------------------------  
*Note: more info. is in page 55 of the manual.
------------------------------------------------
 This option conducts a simple test on the 
 training examples selected from the PAST 
 projects DATABASE. The test is particularly 
 useful if you intend to develop custom 
 prediction models (neural networks) using
 solely your own past project, excluding the
 ProBID's default cases.
 
 This test examines the relationships depicted 
 in the projects' data against industry-known
 rules of thumb. A linear regression analysis 
 is used to identify the following data trends: 
      
  1. Markup vs need for work.                   
  2. Markup vs competition.                     
  3. Change orders vs quality of design drawings.
  4. Claims vs uncertainty of site location.    
  5. Profitability vs company expertise.           
   
 If the trends depicted in the past-projects
 selected for training are not "LOGICAL" and      
 do not comply with industry-known rules of 
 thumb, this indicates the insufficiency of the
 training data.
                                        <END>
END


TEST NEURAL NETWORK               
*Network Errors                                    'as is
  
 this option allows you to TEST the quality of
 predictions produced by a neural network against 
 the actual outcomes experienced in the company's 
 PAST-projects. You might need to use this option 
 right after a training session to test the 
 results of the custom network developed. This 
 test provides a table of the errors in the 
 prediction of all the 7 outcomes of the training 
 projects.
                                         <END>
END


SOFTWARE LICENSE AND WARRANTY
*Software License
 Purchase of the ProBID software system entitles
 unlimited use of the product on one computer. 
 You may physically transfer the software from 
 one computer to another, provided that the 
 software is used on only one computer at a time.
 You may not electronically transfer the software 
 from one computer to another over a network. You 
 may not distribute copies of the software or 
 accompanying written material without the prior
 written consent of PROFIT Consultants. The cost 
 of site license may be obtained from 
 PROFIT Consultants.

*Limited Warranty
 The media on which you receive ProBID
 software are warranted not to fail to execute
 programming instructions, due to defects in
 material and workmanship, for a period of 90
 days from date of shipment, as evidenced by
 receipts or other documentation. PROFIT
 Consultants will, at its option, repair or
 replace software media that do not execute
 programming instructions if PROFIT
 Consultants receives notice of such defects
 during the warranty period. PROFIT
 Consultants does not warrant that the
 operation of the software shall be
 uninterrupted or error free.
 
 ProBID software (including instructions for its
 use) is provided "as is" without warranty of any
 kind. Further, PROFIT Consultants and the
 author do not warrant, guarantee, or make any
 representations regarding the use, or the
 results of the use, of the software or written
 materials concerning the software in terms of
 correctness, accuracy, reliability, currentness,
 or otherwise. The entire risk as to the results
 and performance of the software is assumed by
 you.

 In no event shall PROFIT Consultants be liable
 for any damages arising out of or related to
 this software and manual or the information 
 contained in it. Neither the author nor PROFIT 
 Consultants nor anyone else who has been 
 involved in the creation, production, or 
 delivery of this software shall be liable for 
 any direct, indirect, consequential, or 
 incidental damages (including damages for loss 
 of business profits, business interruption, 
 loss of business information, and the like) 
 arising out of the use of or inability to use 
 such software even if PROFIT Consultants has 
 been advised of the possibility of such damages. 
 Customer's right to recover damages caused by 
 fault or negligence on the part of PROFIT 
 Consultants shall be limited to the amount 
 theretofore paid by the customer to purchase 
 the software.
                                      <END>
END

About
*                         TM
*                    ProBID
 Software for predicting construction outcome 
      and for supporting bid decisions
 --------------------------------------------
          Developed by: Tarek Hegazy 
 (c) Copyright 1993, 1994 by PROFIT Consultants
             All rights reserved
 --------------------------------------------
           P.O. Box 113, Station H,
        Montreal, Que., H3G 2K5, Canada.
 Tel: (514) 843-1356;     Fax: (514) 849-9995
                                            <END>
END

Why ProBID ?                    
* Why ProBID ? 
* ------------       
    If you are much involved in the construction 
  business, you must have experienced how 
  difficult it is, at early stages of a project, 
  to predict its potential success or failure 
  and to decide on a suitable bidding strategy.  
  Despite the importance of these decisions to a    
  costly commitment, you might have to decide on  
  them while a lot of information is lacking and 
  may be under pressures to speed up the 
  preparation of a bid before its submission
  deadline. Often, you, and many other 
  practitioners, are left to your own
  intuition and "Gut Feeling", with little or no  
  help from available tools. It is for these 
  reasons that our ProBID software was developed.
                   
*          ProBID is a software for
*      predicting construction outcome and
*       for supporting your bid decisions.
                                            <END>
END   

What Technology is Used ?
* What Technology is Used ?
* -------------------------
   
     Often, the life or death of a business
  organization is dependent on expert decision 
  makers who are very expensive and scarce.  
   
*     Can the superior performance of human   
*     experts be analyzed and structured ? ...
*     Can it be emulated by a machine that 
*     makes no "human" errors ?
   
 These questions are raised every time a human 
 expert moves from one company to another and 
 with him the many years of experience. It is 
 only in the 1940s when research in Artificial  
 Intelligence (AI) branch of computer science
 has provided various models of how humans store 
 and retrieve their problem-solving knowledge 
 and the manner they use this knowledge to 
 arrive at decisions or conclusions. As a result 
 of that research, Neural Networks have evolved 
 as a computer system that emulates the 
 structure of the human brain and its ability 
 to recognize patterns in the environment. This 
 technique is based on a simple "LEARN BY 
 EXAMPLE" approach. It can provide a reasonable
 DECISION-AID for problems that are experience-
 dependent and involve pattern recognition tasks. 
 This thechnique has, thus, been used in ProBID 
 to tackle the problem of recognizing the 
 potential success or failure of projects.
                                            <END>
END   
   
How does ProBID Work ?
* How does ProBID Work ?   
* ----------------------

      ProBID is intended to help you organize,  
  store, and build on your construction 
  experience to support your decisions on new  
  jobs. ProBID uses the Neural Network technique  
  of computer science to "intelligently" 
  recognize the risk pattern of your upcoming  
  project. ProBID then matches your project
  environment with a number of stored cases of  
  successful and unsuccessful projects. ProBID,     
  accordingly, predicts some indicators of the
  project's potential success or failure,
  including: 
      - possible construction  delay; 
      - level of claims and disputes; 
      - level of change orders; and 
      - level of actual profitability.  
  
  In addition, ProBID suggests a markup strategy  
  for your bid and evaluates your chances of 
  winning. ProBID predictions direct your 
  attention to potential problem areas that you  
  may consider to adjust your estimate, think of  
  alternate decisions, and take early counter 
  measures to avoid problems.
  
  The neural network included in ProBID was 
  trained to be "Intelligent"  as a general 
  predictor. It is not job specific, although 
  geared towards building projects. To fully 
  utilize the power of this predictor, ProBID  
  provides you with a powerful ADAPTATION option 
  that you can use to custom-train such network 
  on some of your past projects that are 
  representative of your particular job 
  environment (this is like employing an expert 
  who has to have a training period to fully 
  adapt himself to your business environment). 
  This issue is emphasized in the tutorial.
                                            <END>
END   


What is ProBID Output ?
* What is ProBID Output ?
* -----------------------   

    ProBID deals with a problem of predicting a
  set of future occurrences concerning the 
  project's bidding and execution outcomes. 
  ProBID predictions are based on your assessment     
  of the project's environment and the factors 
  that have a direct impact on its outcome.

  ProBID predictions for a project are:
  
  1. Estimate of a suitable percent Markup;
  
  2. Estimate of your chances  to Win/Lose 
     the bid at that markup;
  
  3. Estimate  of the  difference between  
     winner (lowest  bidder) and the 2nd 
     lowest bidder (i.e., how much you can  
     potentially increase your profit);
  
  4. Estimate of possible change orders 
     (L-M-H);
  
  5. Estimate of possible claims and disputes 
     (L-M-H);
  
  6. Expected actual duration (months); and
  
  7. Estimate of actual profitability level     
     (Loss-L-M-H).
  
  Generating these predictions is by nature a
  "fuzzy" and difficult task even for human 
  experts due to the multitude of variables
  involved. The neural network utilized in  
  ProBID is not designed to provide precise    
  solutions. Instead, if you develop custom
  prediction networks that adapt ProBID to a   
  specific type of projects (Tunnels, for 
  instance), it will produce the most reasonable 
  predictions based upon the variety of learned 
  cases. If you use more training cases, narrow
  the domain, and have good quality and diversity  
  in the learning cases, the network is expected 
  to make better predictions. However, it is not
  guaranteed to always give an absolutely 
  "correct" predictions, particularly if your 
  training cases are in some way incomplete or
  contradicting.
  
  ProBID requires the user to input subjective 
  assessment of a number of factors that evaluate 
  the project in terms of: job uncertainty; job
  complexity; market conditions; firm ability;   
  and need for work. The assessment of those 
  factors may differ from one person to another
  although they might have the same level of 
  expertise. If more than one person is involved 
  in the project, a brain storming session can 
  be conducted to arrive at an agreed-upon 
  assessment. To help you account for the 
  subjectivity involved, ProBID incorporates  
  a sensitivity analysis option to evaluate the 
  impact of changes in the assessment of project 
  factors on the predictions made.
                                            <END>
END   

   
Who are ProBID Users ?
* Who are ProBID Users ?
* ----------------------

     ProBID is meant to supplement, rather than  
  replace, the decision-making abilities of the 
  following users:
   
* - general contractors: 
    to support your bid/no-bid decision; evaluate 
    your chances of winning at different markup
    levels; identify possible conflicts and 
    problems; and organize your past experience;
   
* - owners, construction managers, and A/E firms:
    to help you select a proper contracting 
    strategy; resolve coordination problems; and
    examine project feasibility;
    
* - estimators:  
    to help you produce timely estimates, 
    accounting for soft factors and risk.
                                            <END>
END   

Where to Find More Info. ?
*           Where to Find More Info. ?
*           --------------------------
                    
* a) References on Bidding Strategy Models:
   
 - Ahmad, I. and Minkarah, I. "An Expert System  
   for Selecting Bid  Markups," Proceedings  of  
   the Fifth Conference on Computing in Civil  
   Engineering, ASCE, March, 1988.
   
 - Ahmad, I., and Minkarah, I. "An Additive 
   Utility Model for Selecting Optimum Bid 
   Price,"  Proceedings of the 18th Annual 
   Pittsburgh Conference on Modeling and 
   Simulation, 18, Part 1, April 1987, 367-373.
   
 - Carr, R., "General Bidding Model," Journal  
   of the Construction Division, ASCE, Vol. 108,  
   No. CO4, Dec. 1982,  639-650.
   
 - Carr, R., "Optimum Markup by Direct Solution,"  
   Journal of Construction Engineering and
   Management, ASCE, Vol. 113, No. 1, March 1987,  
   138-150.
   
 - Friedman, L. "A Competitive Bidding Strategy,"  
   Operations Research, Vol. 4, 1956, 104-112.
   
 - Gates, M., "Bidding Strategies and 
   Probabilities,"  Journal of the Construction    
   Division, ASCE, Vol. 93, No. CO1, March 1967, 
   75-107.
   
 - Hegazy, T. and Moselhi, O. "Analogy-Based  
   Solution to Markup Estimation Problem," 
   Journal of Computing in Civil Engineering, 
   ASCE, Vol. 8, No. 1, Jan. 1994.
   
 - Park, W. "Construction Bidding for Profit,"  
   John Wiley and Sons, Inc., 1979.
   
 - Seydel, J. and Olson, D. "Bids Considering    
   Multiple Criteria," Journal of Construction  
   Engineering and Management, ASCE, Vol. 116,  
   No. 4, Dec. 1990, 609-623.
   
 - Skitmore, M. "Contract Bidding in 
   Construction," Longman Group UK Limited, 1989.
   
 - Tavakoli, A., and Utomo, J. "Bid Markup 
   Assistant: An Expert System," Cost Engineering
   Magazine, Vol. 31, No. 6, June 1989, 28-33.

   
* b) References on Neural Networks:
   
       There are different neural network
   formulations available, each with distinct  
   characteristics (Moselhi et al. 1992). The  
   neural network paradigm used in ProBID 
   developments is "backpropagation (Rumelhart  
   et al. 1986)"  which has powerful pattern      
   recognition capabilities. Important references 
   include:
   
 - Moselhi, O., Hegazy, T. and Fazio, P. 
   "Potential  Applications of Neural Networks  
   in Construction," Canadian Journal of Civil  
   Engineering, Vol. 19, June 1992, 521-529.
   
 - Pao, Y. H. "Adaptive Pattern Recognition and  
   Neural Networks," Addison-Wesley Publishing    
   Company Inc., 1989.
   
 - Rumelhart, D.E., Hinton, G.E., and Williams,  
   R.J., "Learning Internal Representations by 
   Error Propagation," In "Parallel Distributed  
   Processing, Vol. 1, MIT press, Cambridge, MA, 
   1986.
   
 - Sietsma, J. and Dow, R. "Creating Artificial   
   Neural Networks that Generalize," Neural 
   Networks, Vol. 4, No. 1, Pergamon Press, 1991, 
   67-79.
   
 - Wassermann, P.D. "Neural Computing: Theory 
   and Practice," Van Nostrand Reinhold, NY, 
   1989.
   
 - White, H. "Neural-Network Learning and   
   Statistics," AI Expert, Dec. 1989, 48-52.
                                            <END>
END   


A Tutorial
*                 A Tutorial   
*                 ----------
  These are 6 tutorial sessions in the help menu
  of ProBID. Please follow them sequentially. 
  They will give you a quick and simple hands-on
  experience with ProBID. Also, use the <F1> key 
  at any Menu or sub-menu item for a description 
  of its function.
 
 Note: the 6 tutorials are included in the file 
       "Pbidhelp.txt". If desired, use the DOS 
        editor to edit the file and print the 
        tutorials.                          <END>
END    

Tutorial 1:
EXAMPLE BID PROJECT
*              Example Bid Project
               -------------------
* 1.1) Loading the Example Bid Project:
    
    An example of a bid project "Test.bid" is 
  included with ProBID. Select "Project" from  
  the main-menu. This menu allows you to:
  
  - CREATE, EDIT, LOAD, RENAME, and DELETE 
    bid projects;
  - ADD a bid project to the database of past 
    projects (usually after construction ends); &
  - EXIT ProBID to the DOS prompt.
 
      To load and edit the data of the TEST 
  project, move the menu-selection bar to "Load" 
  and strike the <Enter> key. Select "TEST.BID" 
  and do one of the following:  (1) strike 
  <Enter>; (2) move to the <OK> button using the
  <TAB> key and then press <Enter>; or (3) click  
  on the file name using the left mouse button. 
  Notice that the status bar at the bottom of 
  the screen shows "..Loading.." and the project  
  name "TEST".

* 1.2) Input of Project data:
      This is the data which describes the 
  complete environment of a Bid project: its 
  contract type, total cost, duration, its 
  uncertainty level, complexity level, market 
  conditions, your company's ability to do the  
  job, and your need for that job. These data
  are input through 5 screens.
  
* 1.2 a) Screen 1 - Project General Information:
      The first input screen of project data 
  (General information) will appear with a menu 
  at the bottom:
  
          <Esc>  = save and back to main-menu;
          <F1> = get Help;
          <F5> = edit data;
          <F8> = print a report to a 
                 printer or to a file; &
  <PgDn>/<PgUp>= go to next/prev. input screen.
  
  If you are using a mouse, you can click the 
  left mouse button on any input field to start 
  editing, notice the bottom menu disappears. To
  move across the different editing fields, use  
  the <Up> and <Down> arrows. Alternatively, you    
  can position the mouse cursor on any input 
  field and click the left button to start 
  editing.
  
  To exit from the EDIT mode, press the <Esc> 
  key or the right mouse button and, accordingly,  
  the bottom menu will re-appear. You can 
  activate any menu option by clicking the left
  mouse button on the desired option. You can 
  also move to next/previous screens by clicking 
  the left mouse button on the [PgDn] and [PgUp] 
  characters.
  
  Notice in the (General Information) screen that
  it includes information regarding the project
  name, location, type, contract type, duration,  
  and cost. Notice that the categorization of 
  project types is very broad and does not 
  include detailed types such as "Hospitals",   
  "Schools", "Tunnels", etc. This, however, is 
  intentionally programmed in ProBID and is not  
  a limitation. ProBID allows you to customize  
  the general nature of the default prediction 
  model using some of your past projects that 
  belong to a well-defined domain (e.g., 
  Tunnels), thus, adapting the default model to  
  your own work environment.
  
* 1.2 b) Screens 2 to 5:
      If you are done with the "General 
  Information" press, or click on, <PgDn> to go  
  to next 4 screens. These screens require your 
  assessment of 4 groups of factors pertaining  
  to:     
         Screen 2 - Job Uncertainty.
         Screen 3 - Job Complexity.
         Screen 4 - Market Conditions.
         Screen 5 - Company Ability and 
                    Need for Work.

  In these screens, there are some guidelines 
  for entering your assessment of the factors 
  involved based on a scale from 1 (for Low) to 
  5 (for High). As a default, a value of 3 
  (Medium) is used when you create a new project. 
  If you are not sure which value to assign, 
  the default value would suffice. 
  
  Also, you can press <F1> for additional help
  for every screen. 
  
  For editing the input fields, the rules 
  described for the "General Information" screen 
  should be followed.
  
     Once you finish editing the last screen of 
  project description, press the <Esc> key to 
  return back to the main-menu of ProBID.
                                            <END>
END

Tutorial 2:
PROBID PREDICTIONS
*              ProBID Predictions
               ------------------
* 2. Predicting the Project Outcome:
    
    To start a consultation session with ProBID, 
  select "Consultation" from the main-menu. Use 
  the down and up arrows to highlight any of the
  three options available:  

      (1) Input BUSINESS DATA;
      (2) Produce PREDICTIONS; and
      (3) Conduct SENSITIVITY analysis.
  
*  2.1) Business Data Input:
      These are data that depict your company's 
  relative size in the market and your definition 
  of a percentage markup. These data are used 
  during consultation to ensure that ProBID  
  outputs are compatible with your own practice.
       
  Activate "Input Business Data" to start 
  editing the default business. Use the mouse
  or arrow keys to move between the two editing  
  fields and enter a selection that suits your    
  own practice. After editing, strike <Esc> to
  exit from the editing mode and bring back the 
  bottom menu. Then, you can use <F8> to print  
  a report or <Esc> to return to ProBID's 
  main-menu.
  
* 2.2) Producing Predictions:
     To predict the outcome of the TEST project,  
  select "Consultation" from the main-menu and 
  use the  <Down> arrow key to highlight 
  "Produce Predictions" and hit <Enter>. You 
  will be prompted to confirm. Following that,  
  a window will appear asking you to select a  
  neural network  to be used for producing the
  predictions. ProBID has a default neural 
  network "GENERAL.NNT" that is trained on actual    
  projects collected from general contractors in  
  the United States and Canada. The default 
  network has prediction ability that is general   
  and not job or contractor-specific.
  
  For the TEST project, use the arrow keys or  
  mouse to select the "GENERAL.NNT" and hit 
  <ENTER>. A note about the general nature of   
  the network will appear. As mentioned in that
  note, ProBID provides you with the utility to  
  develop CUSTOM neural networks that suit our 
  specific work environment (e.g., Office 
  Building) or specific conditions such as a
  particular owner, locality and/or market. This  
  option is described under the "ADAPTATION 
  menu".
  
  Once you read the note, press any key to 
  continue consultation. ProBID then processes  
  the data of project (TEST) and inputs it to  
  the neural network. A few seconds later, 
  predictions will be produced and presented  
  on the screen. ProBID predictions can be read  
  as such:
 
*  "Out of the many projects stored in the 
*  selected neural network, you are likely to 
*  WIN project "TEST" using a markup of 5.1%. It 
*  is predicted that the difference between your 
*  bid and second lowest bidder is about $3,000.  
*  Upon execution, the project is predicted to 
*  exhibit a High level of change orders, a Low 
*  level of claims, an actual duration of 10 
*  months, and a Medium level of actual 
*  profitability".
  
     You can use [PgUp] to view previous screens   
  containing the project description data that
  represent the inputs to the prediction network. 
  You can also use <F8> to print a report of the
  predictions made. The report can be directed 
  to the printer (on first parallel port) or to  
  an ASCII file that you can later load from any 
  word-processing software. Once you finish, 
  press <Esc> to return to the main-menu of 
  ProBID.        
                                            
* 2.3) Sensitivity Analysis:
     The purpose of sensitivity analysis is to 
  examine how the project predictions may vary 
  if the project description is changed. This 
  issue is important since some of factors, for 
  instance owner attitude, may not be accurately
  known, particularly if you are dealing with 
  this owner for the first time. To start the 
  analysis, select "Consultation" from the main-
  menu and use the <down> arrow key to highlight
  "Conduct Sensitivity Analysis" and hit <Enter>. 
  You will be prompted to confirm. Following 
  that, you will be asked to enter the number of
  simulations, as multiples of 100, select 100, 
  for example. 
  
  The sensitivity analysis is then then performed  
  using MONTE CARLO simulation technique. The  
  simulation generates a number of project
  scenarios (simulations) that are minor random
  variations of your own assessment of the 
  project. All simulations are then input to the 
  neural model selected, and predictions are 
  made. As a result of this analysis, a menu 
  will appear on the screen, with five options:

   0)  Return to Main-menu;
   1)  "Most Likely" predictions;
   2)  markup histogram;
   3)  markup vs. chances of winning (discrete);
   4)  markup vs. chances of winning (normal).
  
  Move using the <Down> arrow to highlight the 
  second menu item "Most Likely Predictions" and    
  press <Enter>, or click the left mouse button   
  on that item. Accordingly, "Most  Likely" 
  predictions for the project are presented on  
  the screen in terms of the mean and standard
  deviation of the seven outcomes in all 
  scenarios. Notice the variation of the mean %
  markup from the original prediction produced   
  (5.1%) before the simulation. For a description  
  of the meaning of these results press <F1> for  
  help, or, if you need a report, press <F2> for  
  a printout of the results. When done, press  
  <Esc> to return to the sensitivity analysis
  window.
                                            <END>
END 
 
Tutorial 3:
A MARKUP STRATEGY 
*           Deriving a Markup Stratey
            -------------------------
  
    As a by-product of the sensitivity analysis,    
  ProBID predicts your chances of winning the 
  job at various levels of markup. Move in the
  sensitivity analysis window using the <Down> 
  arrow to highlight the third option "Markup 
  Histogram" and  press <Enter>, or click the 
  left mouse button on that item. Accordingly, a
  plot of the markup variation in all the  
  simulations (histogram) will be presented. 
  Press <Esc> to return to the sensitivity 
  analysis results window. The last two options
  in the results window give you two methods in  
  plotting the relationship between markup and  
  probability of winning:
  
 (1) Considering markup variation in all 
     simulations (shown in the markup histogram)
     to be a "DISCRETE distribution"; or
  
 (2) Making an approximation that markup 
     variation in all simulations follow a 
     standard normal distribution. You might
     decide that if the markup histogram shows   
     a bell-shape "NORMAL distribution".
  
  For this tutorial, move using the <Down> arrow  
  to highlight the fourth option "Markup vs. 
  chances of winning-Discrete" and press <Enter>,
  or click the left mouse button on that item. 
  The relationship is then plotted on the screen.  
  As shown in the bottom menu, you may use the  
  <right> and <left> arrows to increment the 
  percent markup and, accordingly, read its 
  corresponding probability of winning. 
 
  Based on the relationship shown, you may  
  select a percent markup associated with a 
  desirable probability of winning, and 
  use it to adjust your bid proposal project.  
  Once you are finished, press <Esc> to bring  
  the menu selection window and select the other
  distribution or exit to the main-menu.
                                            <END>
END

Tutorial 4:
ORGANIZING PAST EXPERIENCE
*           Organizing Past Experience
            --------------------------
          
      As  mentioned earlier, the default 
  prediction network of ProBID "GENERAL.NNT" is 
  designed as general and not job or contractor-
  specific. Although this network is geared more    
  towards building projects, ProBID allows you  
  to adapt it and develop other custom predictors
  that suit your own environment.

  To develop a custom prediction network, select 
  "Adaptation" from the main-menu. Use the 
  <Down> and <Up> arrow keys to highlight any of
  the four options available: 
  
        (1) Input PAST projects; 
        (2) Check training data; 
        (3) Start training; and 
        (4) TEST neural network.
  
* 4.1) Example Past Projects:
     Past projects are projects that you have   
  executed in the past. Ideally, they have to    
  include successful and unsuccessful projects:
  
  - Projects that experienced no execution 
    problems.

  - Projects that experienced execution problems   
    such as claims, change orders, duration
    extension, and/or low actual profits.
  
  Certainly, these projects are very specific to 
  your firm and its particular work environment. 
  ProBID organizes those past projects into a
  database that is utilized to enhance future 
  predictions on similar projects. To facilitate 
  the tutorial session, the past projects' 
  database of ProBID comes with 10 example past
  projects that you may use it for tutorial and 
  later delete it and input your own.

  An important feature of ProBID is that you can 
  add to this database any bid projects that 
  passed the bidding and execution stages. As 
  such, ProBID preserves your experience and
  continuously builds on it. To do that, while 
  project TEST is loaded, select "Project" from 
  the main-menu, then use the <Down> arrow to
  highlight "Add to past projects", and press 
  <Enter>. You will be prompted to confirm and 
  accordingly proceed.
  
  To load the database of past projects, select 
  "Adaptation" from the main-menu, then use the 
  <Down> arrow to highlight "Edit past projects" 
  and press <Enter>. In a few seconds, the first 
  screen of the database appears with a menu at 
  the bottom line. Notice that the past project 
  loaded on the screen is project number 11, 
  which is the last project in the database. 
  This is project TEST which you have just added 
  to the database. The bottom menu allows you to:

        <F1> = get Help.
        <F2> = Go to project number...
    <F3><F4> = Move to Previous/Next project.
        <F5> = EDIT this input fields.
        <F6> = ADD a NEW project...
        <F7> = DELETE this project.
        <F8> = REPORT to a printer
               or to a file.

  Note: to help you prepare past projects' data 
  for input to ProBID, a form for data input is  
  included in the text file "FORM.TXT" of ProBID.
  Print this form to the printer from DOS by 
  typing "Print form.txt <Enter>". Make copies 
  of this form, if you like, and fill the data
  related to each project. Later you can simply 
  use the filled forms to quickly fill the past 
  projects' database.
 
* b) Entering Actual Project Outcome:
      Use <PgDn> to move to the next screens of 
  past project number 11. Notice that the first 
  5 screens are exactly the same as those used 
  to enter the description of a bid project 
  earlier in the tutorial. There is, however, an 
  additional screen (screen 6) concerning the
  actual outcome that you have experienced for 
  that project.
  
  To edit the data, use <F5> and edit the seven 
  fields pertaining to:
  
     1- The percent Markup decided;
     2- Bid result: Win/Lose;
     3- The Difference in dollars between winner 
        and 2nd lowest bidder (i.e., Money
        left on the table);
     4- Experienced level of change orders;
     5- Experienced level of claims (H-M-L);
     6- Actual project duration (months); and
     7- Actual profitability attained.
  
  Notice that there is an additional field for 
  selecting the project as a training case, (Y) 
  or not (N). Press "Y" for yes in this field 
  and then <Esc> to bring the bottom menu. Note
  that you can use the mouse for editing and 
  selecting any menu option, following the rules 
  described earlier.
                                            <END>
END

Tutorial 5:
MAKING CUSTOM PREDICTORS
*         Developing Custom Predictors
          ----------------------------
* 5.1) Your Specific Work Environment:
      
      ProBID allows you to interactively TRAIN 
  the default model on some past projects in the 
  database and, accordingly, creates a custom 
  neural network that predicts the outcome of
  specific types of projects. For example, you 
  can develop a custom network that is trained 
  to predict the outcome of SCHOOL projects.
  Alternatively, it can train on OFFICE
  BUILDINGS executed in the NEW YORK area during 
  the past five years. The usefulness and 
  accuracy of your CUSTOM network are largely 
  dependent on the number of past projects 
  selected for training, the completeness and 
  correctness of the projects' data, and the 
  suitability of the training cases to represent
  the domain in which the network will be 
  producing predictions.
       
  As mentioned earlier, you can select some
  of your stored past projects to be used in 
  training. 
   
* 5.2) Are The Training Cases Sufficient ?
      ProBID allows you to run a simple test to 
  check the sufficiency of the training cases 
  you selected. This ensures that the custom 
  network you are developing is realistic and is
  likely to produce correct predictions. The 
  test is particularly useful if you later will 
  develop a custom prediction network using 
  solely your own past projects, and excluding 
  the default cases stored in ProBID. The test 
  depicts the general trends that are contained 
  in the training data, using linear regression 
  analysis, and then compares these trends with 
  well-known relationships in current bidding
  practice. ProBID plots five trends:
  
     1. % Markup vs. need for work.
     2. % Markup vs. competition.
     3. Change orders vs. quality of drawings.
     4. Claims vs. uncertainty of site location.
     5. Profitability vs. firm expertise.
  
  If the trends depicted in the past projects 
  selected for training are not "LOGICAL" and do 
  not comply with industry-known relationships, 
  this indicates the insufficiency of training 
  data.
  
  To start the validation, select "Adaptation" 
  from the main-menu, then use the <Down> arrow 
  to highlight "Check Training Cases", and 
  press <Enter>. In a few seconds, a menu 
  selection window will appear on the screen 
  with five trends to be plotted. Press <Down> 
  arrow key twice, or use the mouse, to highlight
  "% Markup vs. Competition", for example, and 
  hit <Enter>. This will plot the relationship 
  between percent markup and degree of 
  competition, depicted in the 11 training cases
  currently selected.
  
  Despite the low number of training cases used 
  in this tutorial (at least 20 cases are 
  recommended), shows that markup reduces with
  higher competition, a trend that is logical. 
  Press <Esc> to return to the validation menu 
  and continue examining the other trends.
  Accordingly, you will be able to decide wether 
  or not you need more training cases. To return 
  to the main-menu, select the first option and 
  hit <Enter>.
  
* 5.3) Training Options:
     Once training cases are validated, you can 
  start a training session to develop a custom 
  neural network that specializes on predicting 
  the outcome of the category of projects used 
  in training. ProBID gives you two options in 
  developing the custom network:
  
  1- Adapting Default Knowledge:
     this adds your selected training cases to 
     a group of default cases, thus, adapting 
     the general model to your own environment. 
     This is beneficial if you have small number 
     of training cases (a relatively new 
     company), and if you are not confident of 
     your own bidding practice; or
  
  2- Building Your Own Model:
     this option uses solely your own training 
     cases in the development of the custom 
     neural network. You may select this option 
     if your company has a lot of past projects 
     experience. ProBID will help you organize
     your projects' data, preserving that 
     knowledge and working as a support for your 
     bidding decisions.
  
  To select either option, first select 
  "Adaptation" from the main-menu, then use the 
  <Down> arrow to highlight the training option 
  you need, and press <Enter>. You will be 
  prompted to confirm and provide a file name 
  for the new network, and accordingly, proceed. 
  A percentage box will then appear showing the
  percentage of training done.
   
  A useful hint: if you would like to quit from 
  a long training session, press the letter "e" 
  or "E" several times to skip the internal 
  training cycles.
  
  After training, you will be prompted with 
  several options:
  
  1- accept the network and return to the main-
     menu (if training is satisfactory);
  2- view the errors of the trained network;
  3- print network errors to the printer;
  4- proceed with additional training 
     (if errors are not satisfactory); or
  5- delete the network and return to main-menu.
  
  Use the <Down> arrow key to highlight the 
  second option and press <Enter> to view the 
  training errors pertaining to the 11 training 
  cases. The error is calculated as the 
  difference between the actual project outcome 
  (stored in the database) and the network's own
  prediction of that outcome, as a percent of 
  the actual outcome. Based on these errors, you 
  might decide the to accept the network as is 
  or continue with more training.
  
  If you decide to keep a custom network, 
  highlight the first option in raining results 
  menu and press <Enter> to return to the main-
  menu. The network can then be utilized in 
  producing predictions for new projects that 
  are of the same category as those used in 
  developing the network.
                                            <END>
END

Tutorial 6:
USING CUSTOM PREDICTORS
*           Using Custom Predictors
            -----------------------
    If your company is engaged in more than one 
  type of project or has a business in different 
  localities, use ProBID to develop custom 
  networks for the different project types. Such 
  networks are expected to have good performance
  since they are more specific and trained on 
  narrow and well-defined domains. Note that 
  each time you develop a custom network, you 
  will have to edit all your past projects and 
  select which projects to use for training by 
  typing "Y" in the last field, or "N" to 
  exclude from selection.
    
  Custom prediction networks can be used in
  predicting the outcomes of new projects. Such
  new projects, certainly, are additional 
  experiences that you, at some point, would 
  like to add them to the custom network. To do
  this, add these projects to the database of 
  past projects using "Project-menu", then, edit 
  the past projects database and input the 
  actual outcomes experienced in such projects. 
  Following that, edit the past projects database 
  and select the projects to be used for 
  re-training. They have to include the ones
  used in previous training in addition to the
  new projects. This process, thus, requires you 
  to keep record of previous training cases.
  
  Note that using a single new experience to 
  re-train a custtom network is not expected to 
  noticeably affect its performance, particularly 
  if it was originally trained on a large number
  of cases. An updating frequency of twice a year 
  or three new projects is reasonable.
  
* More Experimentation:
      Try to do some experimentation on your 
  custom network, examine how practical the 
  predictions are, and how sensitive these 
  predictions are to various changes. You could
  for example, change the score assigned to one
  factor at a time (for example; your need for 
  work) and see how that only would affect the 
  predictions. Experiment with other factors to 
  see which factors do affect the predictions 
  stronger than others, for your particular 
  case. Also, test the predictions against 
  projects for which you know the actual outcome. 
  Experiment also with developing custom networks 
  with varying numbers of training cases and see
  see what number is good enough for your 
  particular environment. Have a feel of the 
  effect of subjectivity on the decisions 
  produced and establish for yourself some 
  useful hints that may help at last minute 
  negotiation of the bid.
                                           <END>
END                    
