Planned vs Actual: What the Variance Is Telling You
A variance is not a failure report. It is the earliest place your cost base tells you something changed.
Summary
- A cost variance is a question, not a verdict: 470 hours against an estimate of 340 has at least four causes, and each belongs to a different person.
- Set the plan number bottom-up at achievable performance and reset it once a year or when an input moves materially. A plan set at perfect, or at whatever you currently do, tells you nothing.
- Read variances in pairs. Rate against efficiency, price against usage, volume against cost per unit and revenue against delivery cost each tell a story that a single variance cannot.
- Start with three products, project types or procedures, run the comparison monthly for a quarter, and widen it only if it earns its keep.
A project was estimated at 340 hours. It delivered in 470.
Most firms record that as a margin miss, note it in the review, and move on. The project is finished; what is there to do?
A great deal, because 470 against 340 has at least four different causes and they point at four different people.
- Did we scope it wrong? The estimate was never achievable. That is a pre-sales and estimation problem, and the same error is probably sitting in three quotations currently out.
- Did the client change it? Scope grew after signature and nobody raised a change order. That is a contracting and account management problem.
- Did we staff it wrong? Three juniors where the estimate assumed one senior and two mid-level. Hours rose, cost per hour fell, and the net effect needs working out rather than assuming.
- Or did we just execute badly? Rework, poor handover, waiting on an internal dependency.
Four causes. Four owners. Four completely different fixes. And a single number, "we were 130 hours over", tells you none of it.
Standard costing works outside the factory too
The manufacturing version is called standard costing and is well documented. The concept is universal: what we assumed delivery would cost versus what it did.
What planned and actual mean in seven kinds of business, and what the gap between them usually reveals
| Business | Planned | Actual | What the variance usually reveals |
|---|---|---|---|
| IT / tech services | Estimated project hours and mix | Timesheet hours and seniority | Scoping error, scope creep, staffing mix, execution |
| SaaS | Assumed infra and support cost per account | Metered compute, tickets logged | A plan priced on usage that has since doubled |
| Hospitals | Package rate assumptions | Actual case cost | Length of stay, consumables, complications, disallowance |
| Diagnostics | Assumed reagent and run time per test | Actual consumption | Batch size, repeat rate, calibration frequency |
| D2C | Planned CM3 by channel | Actual CM3 | Return rate, ad cost, coupon stacking |
| Manufacturing | Standard material and labour | Actual issued and booked | Price, usage, rate, efficiency |
| Pharma | Standard batch cost | Actual batch | Yield, campaign length, QA rework |
Setting the plan number
Three decisions, and the second is where it goes wrong.
What goes in it. Build it bottom-up: inputs at current rates, effort at realistic levels, and the capacity consumed, with any shared cost apportioned on a basis you can defend. Not last year's number with an inflation adjustment. It is the same build as working out what one unit costs, done before the work rather than after it.
At what level of performance. You can set it at perfect, at achievable, or at current.
Set it at perfect and every variance is adverse forever. People stop reading a report that only ever says you failed.
Set it at current and you have learned nothing. It simply records whatever you happened to do.
Set it at achievable. What a competent team, under normal conditions, actually delivers. The variance then means something.
How often to reset. Annually for most, plus whenever an input moves materially. Reset too often and you erase the trend you were trying to see: if the plan follows the actual, the variance vanishes and so does the signal.
Read variances in pairs, never one at a time
The single most useful habit, and the one almost nobody has.
A variance on its own is ambiguous. Two together usually tell a story.
- Rate against efficiency. Hours up but cost per hour down almost always means you staffed junior. Whether that was good or bad depends on the net, and on what it did to quality.
- Price against usage. A favourable input price alongside adverse consumption is the signature of a cheaper input performing worse. Read them separately and you will congratulate purchasing for a decision that cost you money.
- Volume against cost per unit. Cost per unit up while volume is down is usually absorption, not inefficiency, and the fix is a capacity conversation, not a performance one. Why unit cost moves with volume works through the arithmetic.
- Revenue against delivery cost. In services, a project that came in on hours but under on revenue means the discount happened somewhere nobody logged.
The revenue side of the same discipline, a miss split into what volume cost and what price cost, is worked through in what FP&A is. This page stays with the cost side.
Putting each variance in front of its owner
A variance report nobody discusses is a filing exercise. The general mechanics of a review that decides something apply here unchanged: a monthly forum, a named owner, a threshold that triggers a conversation. What is specific to cost is the trend rule, the size of the threshold, and which owner each cause maps to.
- Trend, not month. One adverse month is noise. Three in the same direction is a change in the business.
- Size the threshold. Anything below 3% or 5% of the cost element is not worth a conversation. Otherwise the review drowns in small items and misses the large one.
- Map each cause to an owner, not the report to finance. Estimation owns scoping. Account management owns scope creep. Resourcing owns mix. Delivery owns execution. A report owned by "finance" produces no action, because finance cannot change any of the causes.
What it catches that nothing else does
- Drift. Costs creeping up 1% a month show up as an adverse variance in month two. They show up in your annual accounts in month fourteen.
- A silent specification change. Same price from a supplier, slightly different grade, worse yield. Same plan from a cloud provider, different instance behaviour. The usage variance catches it; the invoice does not.
- Estimation bias. If your projects are consistently 20% over, the issue is not delivery. It is that your estimates are systematically optimistic, which means every quotation currently outstanding is wrong by about 20%.
- A price that has quietly become unprofitable. If your plan says ₹2,400 and your actual has been ₹2,700 for four months, every quotation issued at a margin on ₹2,400 has been wrong, by ₹300 a unit (₹2,700 − ₹2,400), or 12.5% of the planned cost.
That last one is why this sits in the pricing practice, beside the contribution floor every quotation is meant to clear.
Your floor is only as current as the cost underneath it.
Start with three items, not the whole business
You do not need to do this everywhere.
- Pick three. Your three highest-volume products, or three representative project types, or three most common procedures.
- Write down what each should cost. Inputs, effort, capacity. An hour's work per item.
- Capture the actual for the same three. Most of it already exists: a timesheet, a consumption log, a claims record. It has simply never been compared.
- Run it monthly for a quarter. By month three you have a trend, and the trend is the useful part.
- Then widen it if it is earning its keep, and not if it is not.
The question that makes this worth doing: when were your cost assumptions last updated, and what has happened to your inputs since? If the answer is "last year" and "they went up", every quotation you have issued since is built on a number that is no longer true.
Our Pricing Maturity Assessment gives you a short read across your pricing and the cost floor underneath it, including how recently the costs behind your quotations were checked.
Frequently asked questions
What is cost variance analysis?
Comparing what delivery was planned to cost with what it actually cost, and breaking the gap into causes that each have an owner: scoping, scope change, staffing mix and execution in a services firm; price, usage, rate and efficiency in manufacturing. The comparison is the easy part. The breakdown into causes is the point.
Should a standard cost be set at perfect performance?
No. A plan set at perfect produces adverse variances every month, and people stop reading it. Set it at what a competent team actually delivers under normal conditions, so that a variance means something changed.
How large should a cost variance be before anyone discusses it?
A threshold of 3% to 5% of the cost element is a workable starting point. Below it, the discussion costs more than it finds. Above it, look for a trend: one adverse month is noise, three in the same direction is a change in the business.
How often should standard costs be reset?
Once a year for most businesses, and whenever a material input moves. Resetting more often than that makes the plan follow the actual, which removes the variance and the signal with it.
How useful was this article?
One tap. It tells us what to write more of.
About the author
SRF Capital Studio
The next one
Get what we publish next, by email.
Working notes on raising, borrowing, protecting, growing and structuring capital in India. One email a week at most, and you can leave any time.
We use your address only to send this. See our privacy policy.
We store your address to send you these emails and nothing else. See our privacy policy.
Related reading
