Cost Variance Analysis

Explain every gap between estimate and actual: which assumption was wrong, and what does that teach the model?

Time30 min
FormatSolo
StageTrack

Cost Variance Analysis: what it is and why it works

Cost Variance Analysis compares each estimate with its actual once real costs are known, computes the difference as a percentage, and explains it. Each variance is classified by cause: a wrong assumption (for example, a cycle time or scrap rate), wrong input data (a stale price or wrong quantity), a scope change that the estimate never covered, or a learning effect that should be modeled rather than corrected. The observed ratios are then used to update estimating factors, and the lessons are shared with every estimator so the next estimate starts from better data.

The method works because estimating is a feedback discipline. Without systematic comparison against actuals, the same optimistic factor survives project after project, and each team rediscovers the same overrun. Classifying causes matters: scope changes should not be used to recalibrate unit-rate factors, and learning effects belong in a learning curve rather than in a flat correction. The logic is the estimate-versus-actual feedback loop of the cost control cycle. Variance Analysis upgrades the data behind Estimate Confidence Grading, explains the gaps the Cost Dashboard displays, and provides the historical pairs that Parametric Costing and the Learning Curve need.

What you need

Estimates with their line structure and assumptions
Actual costs booked in the same structure
Records of scope changes between estimate and execution
Current estimating factors and the database where they are stored

What you get

Variance per line in value and percent
Causes classified: assumption, data, scope change, learning
Updated estimating factors with the evidence behind each change
A short lessons note distributed to all estimators

When to use it

When actual costs arrive and the model is not updated.

How to do it, step by step

Pick the estimates with actuals available.
Compute the variance for each: estimate minus actual, in percent.
Classify causes: wrong assumption, wrong data, scope change, learning effect.
Update the estimating factors with the observed ratios.
Share the lessons with all estimators.

Worked example: Calibrating weld-hour factors in a pipe spool shop

Illustrative scenario — figures are realistic but not from a real company.

A fabrication shop producing pipe spools and skids for chemical and water treatment plants reviewed 12 completed jobs. Estimates used a factor per weld inch-diameter by material. The shop suspected that stainless jobs overran, but nobody had quantified it.

The estimator pulled estimated and actual hours for welding, fit-up, handling and testing on each job, expressing variance as (estimate - actual) / estimate.
Carbon steel jobs averaged -3%, within noise. 316L stainless jobs averaged -19% on welding hours.
Classification showed that about half the stainless overrun came from back-purging and interpass temperature control not included in the factor, a wrong assumption. Two jobs included late scope additions and were excluded from recalibration. Another portion came from one job with a new welder team, treated as a learning effect.
The stainless weld factor was raised by 11%, and a separate purge allowance per joint was added to the estimating sheet.

Result. The next five stainless jobs came in within 5% of estimate. The shop also began recording actual hours in the same breakdown as its estimates, which made the next review take half a day instead of a week. The lessons note went to all four estimators and the sales engineer.

Open it in the FieldKit Design to Cost app → Get the Design to Cost template pack — join the waitlist

Common pitfalls and how to avoid them

Recalibrating factors on jobs with scope changes.Separate scope-change variances and exclude them from factor updates.
Booking actuals in a structure that does not match the estimate.Align cost codes with the estimate breakdown before the job starts.
Correcting a factor on the basis of one job.Look for consistent patterns across several jobs and state the sample behind each change.
Keeping lessons with the analyst.Update the shared estimating database and brief every estimator.

Frequently asked questions

How do you calculate cost variance?

Subtract the actual cost from the estimated cost, then divide by the estimate to express it as a percentage. Some organizations divide by the actual instead; either works if applied consistently and the sign convention is stated. In earned value management, cost variance has a specific definition, earned value minus actual cost, which measures work performance rather than estimate accuracy.

What causes cost estimates to be wrong?

Typical causes are wrong assumptions such as cycle times, scrap or productivity; outdated or wrong input data such as prices or quantities; scope changes after the estimate; learning effects not modeled; and optimism bias. Classifying variances by cause tells you which part of the estimating process to fix.

How often should estimating factors be updated?

Update them when a consistent pattern appears across enough jobs to be credible, and review them on a regular cycle, such as after each project closeout or quarterly for high-volume estimating. Record the date, the evidence and the old value so changes can be traced and reversed if later data disagrees.

Origin

Estimate-versus-actual feedback — cost control cycle (engineering economics).

Used in these playbooks

Gate cost review pack 1 day

One day before each gate: dashboard refreshed, variances explained, attainment measured — and the go/no-go decision documented.

Related methods

More in “Track”

Keep the target visible: dashboards, gates and ownership.