Home / Formulas / Predict defect density
§2 — Prediction Formula 2 of 7

Survey tab & predict defect density

The Survey tab has several categories of questions. Answer or edit all of the required inputs, then answer at least some questions in each remaining sub-tab — the machine learning model predicts defect density from your assessment.

How To Answer

Answering the survey

  • You must answer or edit all of the required inputs.
  • Beyond that, it’s recommended to answer at least some questions in each sub-tab.
  • If you don’t know the answer, don’t guess — answer “unknown” and the question won’t count toward the prediction.
  • The import file has the questions that are typically easiest to answer shown in bold.
  • The import file also explains what each question is used for.
  • Some questions are used for defect density prediction, and some are used for the FMEA. If you’re doing only a prediction, you don’t need to answer any FMEA-only questions.
Categories

The survey categories

Category
Description
Required inputs
All inputs in this sub-tab must be confirmed or edited in order to see prediction results on the Dashboard tab.
Product information
Check all that apply. Used only by the software FMEA.
Customer
How your customer sees your software.
Failure modes
Used only by the Software FMEA product.
Inherent risks
Risks that are often not easy to change — new system, new hardware, etc.
Software personnel
The software people — their experience, location, etc.
Planning
How much, if any, advance planning is done for this software release.
Execution
How well the team can execute to the plan.
Software requirements
Requirements-writing activities, how good those requirements are, and whether they cover fault management.
Design
Design activities, how good the design is, and whether it covers fault management.
Code
Coding activities, how good the code is, and whether it covers fault management.
Unit testing
Module-testing activities, how good that testing is, and whether it covers fault management.
Testing
Black-box testing activities, how good the testing is, and whether it covers fault management.
Corrective action
Fix effectiveness of the software.
Configuration management
How well software versions and documents are controlled.
Defect tracking
How well defects are tracked.
Reliability
How reliability is measured.
Software QA
Activities conducted for software quality assurance.
Process
The processes used for developing and testing software.
Defect Density

Why defect density is normalized

The survey inputs are used to predict defect density — a normalized measure of defects per size of code. Smaller projects have fewer defects than larger ones, so it’s difficult to predict defects without a normalized measure. Requs AI Predict predicts defect density via a machine learning model.

Defect density can be measured against any size measure, but in industry it’s typically measured as defects per EKSLOC (1,000 Effective Source Lines of Code). There are two major milestones defect density is measured against:

Defect density milestone
How it is used
Start of testing
Predicts the total number of defects that will be found during system-level testing. Does not include defects found in code reviews, unit testing, integration testing, or field usage.
Field
Predicts the total number of defects that will be found once the software is in operational use, through the end of growth for this version — typically 2 to 8 years, depending on installed sites and end users. Does not account for defects introduced by subsequent releases.

Continue to the next stage of the prediction.

Defect density feeds directly into the total defect count.

Predict Defect Counts → ← Back to Predict Size