Projection vs Impact: Why Business Cases Die After Approval

Business cases have a strange lifecycle. Enormous care goes into building them — assumptions debated, scenarios modelled, numbers defended in committee. Then the investment is approved, and the document that justified it is never opened again. The project ships, the team moves on, and nobody ever checks whether the promised value arrived. The projection dies at the moment of approval, which is precisely the moment it should start doing its second job.

Nobody measures, and the evidence says so

This is not a cynical caricature. Deloitte's 2022 intelligent automation survey of 479 executives across 35 countries found that more than half of organisations implementing automation had never calculated their actual cost reductions. Those who did measure reported a 32% average reduction — but the headline is the denominator: the majority approved investment against a projected benefit and then simply never looked.

The same survey shows why this matters commercially, not just intellectually: average pilot payback lengthened from 16 months in 2020 to 22 months by 2021/22. Benefits are arriving later than the business cases assumed. If you never measure, you never find out — and your next business case inherits the same optimistic timing, unchallenged.

Why the loop stays open

The reasons are mundane. The person who built the case has moved roles by the time results are measurable. The baseline was never recorded, so there is nothing to compare against. The projection lived in a spreadsheet with assumptions that were edited during negotiation, and nobody is sure which version was approved. And there is a quiet incentive problem: measuring realised impact creates the possibility of being visibly wrong, and no one volunteers for that.

But the cost of the open loop is paid later, with interest. Every unverified business case makes the next one harder to approve. Finance teams that have watched three projections evaporate unmeasured start applying an informal haircut to everything that crosses their desk — 50% off your number, on principle. Your model is no longer being evaluated on its merits. It is being evaluated on the unmeasured history of every model before it.

Variance is credibility currency

The counterintuitive point: being wrong, with a measured variance, is worth more than being unaccountable. A leader who returns to the committee and says "we projected £240,000 of annual benefit, we realised £185,000, and here is which assumption drove the gap" has done something rare. They have demonstrated that their numbers are falsifiable — which is the only reason numbers deserve trust in the first place.

Variance analysis also compounds. If your realised deflection rate came in below projection but your cost-per-contact assumption held, you have learnt something specific and portable: next time, tighten the adoption assumption, keep the cost model. Over three or four measured projects, your assumptions stop being estimates and become calibrated priors. That is a moat no template can give you, and it shows: your conservative case starts landing within a few percent of actuals, and approvals get faster because finance has watched it happen.

Closing the loop in practice

Four disciplines make the loop closeable, and none of them is difficult if you commit before approval rather than after:

Freeze the projection. The approved version of the business case — inputs, assumptions, scenario chosen — must be preserved exactly. If the projection can be silently edited after the fact, the comparison is theatre.

Record the baseline. Realised impact is a before-and-after claim. Capture the "before" numbers (contact volumes, handle times, error rates, costs) at projection time, because reconstructing them eighteen months later is somewhere between painful and impossible.

Measure on the same definitions. If the projection counted deflected contacts, the impact measurement must count deflected contacts — not sessions, not containment, not a metric that happens to look better. Changing definitions mid-loop is how organisations fool themselves politely.

Report the variance, whichever way it points. Publish realised-versus-projected as a routine artefact, not a confession. The habit matters more than any single result.

Tooling the loop instead of willing it

This is the reason LeadersToolset treats Projection and Impact as two linked journeys rather than one calculator. A Projection report captures the business case before the change — inputs, scenarios, the full assumptions audit trail — and stays fixed. When results are in, an Impact report measures what actually happened, and links back to its source Projection, so the variance between promised and realised is computed against the frozen original, not against memory. The comparison you would otherwise have to reconstruct from old spreadsheets and older emails is simply there.

That structural link is the whole argument of this piece in software form: a business case is not finished when it is approved. It is finished when it has been checked.

If you want to see what a closed-loop report looks like before building your own, the sample report shows the full structure — scenarios, assumptions trail, the lot. Or start your own projection at the calculator library; your first calculator is free, with unlimited re-runs, and the methodology behind every formula is public on the methodology page.