# Defendable Estimating: Part I— How to Predict Cost Drivers based on Requirements using TrueFindings

Predicting cost from requirements is a fundamental goal of parametric estimating. Many say they do it.

But often, analogies are not based on rigor and do not represent organizational experience/productivity

not capture the complexity of project deliveries driven by requirements.

TruePlanning allows for calibration of the latter. TrueFindings allows for aggregation, filtering and analysis

of these calibrations. But then what? How can this knowledge base of calibrated metrics best be utilized?

**The answer is aligning knowledge of corresponding requirements that drove the costs that informed the calibrations.** In this way, we predict an appropriate set of model input-drivers based on the organization

and its project/program histories!

Here’s an example for the following notional Software Development data in Excel. In blue are the project names and SW components delivered. In green are the two key TruePlanning for Software (“True-S”) inputs, derived by calibration of the gray actuals. In yellow are component Key Performance Parameter scores.

*{Note: these KPPs could have been KSAs or any other measure of requirements for past/future deliveries.}*

Again, the goal is to perform Predictive Analytics to determine a method for best justifying cost driver-inputs

as a function of KPPs {or KSAs}. Certainly, we can estimate sufficiently using PRICE’s knowledge base and True-S estimating models. HOWEVER, we also have the above relevant program histories to leverage. The challenge (typical at early stage, e.g., NASA Phase-A) is the estimator ONLY knows mission requirements without yet an adequate description of product/functional requirements and detailed architecture.

Our job is to fine-tune the use of True-S models by best applying knowledge from KPPs to predict key drivers,

in this case above Functional Complexity. Assuming we know the future program’s scores (and again, the only quantified understanding of new requirements), the solution is to perform regression analysis of the calibrated values of functional complexity versus their associated KPP scores as candidate predictors. Certainly, a simple bivariate regression is feasible. But why stop there? In Part II, we will review step-by-step how to use the new TrueFindings application to produce a most appropriate CER using multiple regression. The result will look like this equation, with associated residual plot and statistical-significance metrics—

Functional Complexity = 0.449 * KPP#1 + 0.468 * KPP#2 - 81.39

For now, appreciate that we can indeed take advantage of relevant calibrated comparables (i.e., knowledge) to produce defendable estimates tied directly to parametric modelling drivers that are based (only!) on early-stage requirements. Today, the latter were expressed as KPP (or KSA) scores. For you, the pool of candidate predictors can be any metric. Essentially, you can not only develop a calibration knowledgebase; you can now also leverage the latter to best fine-tune TruePlanning reflecting your requirement metrics.

Success can be accelerated by the PRICE® **Predictive Cost Analytics (PCA) integration** of business, engineering, and program management objectives in order to produce competitive solutions.

TruePlanning immediately facilitates objective **quantitative assessment** of customer needs/

requirements, winning-business solutions, and project lifecycle cost management.

Contact me for a demonstration, either in person or via telecon.

We’ve had many success stories here. You can be next!