Almost Nobody Goes Back to See If the Project Delivered
Companies build careful business cases to approve spending and then rarely go back to see what happened. The missing feedback loop is why forecasts stay optimistic.
The Gap in the Process
Corporate capital allocation is built around approval. There are forms, thresholds, committees and financial models, all pointed at one decision: should this be funded. Once the answer is yes, the machinery largely stops.
The post implementation review, sometimes called a post audit or a look back, is the step that closes the loop by comparing what a project delivered against what its business case promised. It is a standard recommendation in every corporate finance text and it is inconsistently practised in real organisations.
An organisation that never checks its forecasts against outcomes is not learning from experience. It is accumulating experience without extracting anything from it.
Why It Gets Skipped
The reasons are mundane and they are strong.
Nobody owns it. The finance analyst who built the model has moved teams. The sponsor has been promoted. Ownership of a completed project is genuinely unclear.
The data is hard to isolate. A new production line raises output, but so did a demand upturn and a pricing change. Attributing results to the project requires a counterfactual that does not exist.
The incentives point away. A review can only produce two findings. Either the project delivered, which is unsurprising and generates no action, or it did not, which is uncomfortable for identifiable people. Very few individuals benefit from commissioning one.
It is backward looking. Management attention favours the next decision over the last one, and a review consumes analyst time that could go into a live proposal.
What the Absence Costs
The direct cost is that estimation bias never gets measured, so it never gets corrected. If projects systematically deliver 60 percent of forecast benefit, that is an enormously useful fact. It lets the company discount future proposals by a known factor rather than by intuition, and it identifies which categories of project are most overstated.
Without it, the correction is applied crudely through an inflated hurdle rate, which penalises accurate forecasters and optimistic ones equally and pushes the company away from long lived investment.
The indirect cost is behavioural. When people know a forecast will be checked, the forecast changes. The review does most of its work before it happens.
Doing It Without Turning It Into a Trial
The reason reviews fail once introduced is that they get experienced as blame allocation, after which proposals become vaguer and harder to assess, which is the opposite of the intent.
| Design choice | Effect |
|---|---|
| Review a sample, not everything | Keeps cost proportionate |
| Report patterns, not individuals | Reduces defensive forecasting |
| Fixed timing, set at approval | Removes the implication of suspicion |
| Record assumptions at approval | Makes later comparison possible at all |
That last row is the one most often missed. A business case that states only a return percentage cannot be reviewed, because there is no way to tell whether a shortfall came from volume, price or cost. Capturing the two or three driving assumptions at approval is what makes a later review mean anything.
Separating Bad Forecasts From Bad Luck
A useful review distinguishes between a project that failed because the estimate was wrong and one that failed because the world changed. A logistics investment justified on demand that a recession removed is not the same as one justified on savings that were never achievable.
The first is a risk that materialised. The second is an estimation problem, and only the second should change how future proposals are assessed. Conflating them either punishes people for bad luck or excuses genuinely poor analysis as circumstance.
The Bottom Line
Post implementation review is among the highest return activities in corporate finance and among the least practised, because its benefits are diffuse and its discomfort is specific. The organisations that do it well keep it sampled, keep it about patterns rather than people, and record the underlying assumptions at approval so that a comparison is even possible. The alternative is what most companies have: a rigorous front end, no feedback, and a hurdle rate quietly carrying the weight of every forecast that was never checked.