The replacement interval in the model was shorter than the history

## Ask where the fourteen months came from

The three-year replacement scenario showed fifty-six crore rupees for a defined customer cohort. Beneath the number sat a blanket fourteen-month replacement interval. Once that assumption entered the model, the subsequent arithmetic was consistent and the presentation was easy to follow. The problem was the origin of the interval. I could explain what the model did with it more confidently than I could explain why it represented the customer history. That should have been a reason to pause before debating the attractiveness of the final revenue figure.

Replacement demand can appear more predictable than new-equipment demand because the installed base already exists. But an existing product does not create a known date for the next purchase. Use, application, customer decisions and technical condition all matter. A strategist can examine commercial history without becoming the person authorised to prescribe replacement. I wanted our account to distinguish those responsibilities. The model needed a defensible assumption about buying events, while qualified technical owners continued to assess the actual condition and suitability of equipment in its particular use.

We commissioned a review costing twenty-nine lakh rupees. Sixty eligible completed replacement intervals were examined, producing a median of twenty-two months. The group had limits. It was not the full population of customers, and completed intervals did not describe every item still in use without a subsequent recorded replacement. The review therefore did not reveal a universal product life. It gave us more relevant evidence for the cohort under discussion and a reason to question the blanket assumption that had been doing so much work in the revenue scenario.

I asked for those limitations to travel with the observed median. Without them, we could have replaced one overbroad number with another and called the model improved. The point was not that twenty-two months should become the new rule for every customer. It was that the earlier scenario had presented a common interval without an adequate account of whose behaviour it represented. The reviewed history made the scenario less comfortable and more specific. Both effects mattered. A number should not earn its place merely because it is conservative or because it came from a spreadsheet containing actual records.

## Let the sample change the scenario within its limits

The revised three-year revenue scenario for the same cohort was thirty-nine crore rupees. The earlier fifty-six-crore figure and the revised figure were both forecasts. The change did not mean seventeen crore of actual sales had disappeared, and the research had not caused customers to buy less. We had changed our account of a possible future. That distinction may sound obvious when written out, but forecast revisions often acquire the language of a loss or a saving as they move through a management conversation.

Some colleagues wanted to keep the earlier case as the main ambition. They argued that stronger commercial engagement could bring replacement discussions forward and that a strategy should not merely extrapolate past behaviour. I agreed with the second point. A business can influence its future, and a historical median is not destiny. But a claim about what a new commercial effort might achieve is different from an assumption presented as the customer's existing pattern. If the faster case required changed behaviour, the proposal needed to name that change and the evidence supporting it.

We therefore used the revised account in the central strategy discussion and kept a separate explanation of what would have to change for a faster scenario. That preserved ambition without disguising its conditions. I did not want a model so cautious that it excluded any possible improvement. I wanted one whose reader could distinguish observation, judgement and intended intervention. Those categories can all belong in a strategy, but they should not borrow credibility from one another without being identified. A forecast is more useful when the board can see which assumption it is being asked to accept.

The research expense remained a cost. The review did not automatically release capital, reduce staffing or change the work required for existing customers. It changed the evidence available for a decision. That can be valuable without a fabricated financial benefit attached to it. I resisted describing the exercise as preventing a particular quantified loss, because we did not know what would have happened under a different decision. The bounded result was a revised scenario, a clearer account of the cohort and an explicit statement of the work needed to justify a different pace of demand.

My own role in the earlier account mattered. I was not an outside critic arriving after somebody else's forecast failed. I had been advocating the aftermarket direction and was responsible for ensuring its supporting assumptions deserved examination. It would have been convenient to treat the review as a technical correction for an analyst. Instead, I needed to explain why the less ambitious figure should be heard in a discussion whose momentum I had helped create. The strategic recommendation could remain attractive, but it could no longer rely on the old interval as though the new evidence had not arrived.

## Separate commercial ambition from observed behaviour

An industrial replacement model has several layers that deserve their own explanation. There is the population being considered, the events visible in its records, the way incomplete histories are handled and the commercial interpretation placed on the result. A model may also contain an intended change in customer engagement. These are not instructions for another company's technical maintenance process. They are questions about how a commercial scenario is assembled. The qualified people responsible for product condition and safe operation must retain their own standards and decisions regardless of what a revenue model would prefer.

The review also changed how I hear a request for one simple assumption. Simplicity can help directors understand a paper, but it can conceal variation that materially affects the decision. I do not think every forecast needs to display every customer record. It should explain enough about the cohort and the judgement for a reader to assess the conclusion. When a common interval carries most of the growth story, its basis is not a minor detail. It belongs near the recommendation, along with the range of evidence that could cause the sponsor to revise it.

In a board role, I would ask management to identify the behaviour it has observed and the behaviour it hopes to change. I would want to know whether the same cohort is being compared and whether the revised number describes a forecast or an actual result. Directors should be able to examine those distinctions without rebuilding the model line by line. Their contribution is to test the business inference and the proposed commitment. If a scenario depends on customers acting differently, the board should understand why that change is plausible and which responsible executive will examine the evidence as the plan proceeds.

The lesson was not to replace fourteen months with twenty-two everywhere. It was to stop treating an inherited planning number as though familiarity made it evidence. Our revised scenario was smaller, the research cost was real and the future remained uncertain. That was still a better foundation for the next decision. I want a strategy function to be able to say that its preferred direction needs a less flattering account, and then continue the discussion. A business gains more from that correction than from a confident forecast whose central assumption nobody can explain.