A customer who has not answered is still in the story

## Two denominators, neither a final answer

The question came to me as a choice between two numbers. Which quotation-conversion rate should go into the capacity paper? One calculation counted wins against every offer issued. The other counted wins only among offers whose outcome was known. The request assumed that choosing the correct denominator would settle the matter. In fact, the denominators were describing different populations, and neither population established the eventual result for the whole group of customers. Our reporting had made an unfinished process look finished enough to support a confident investment assumption.

In the fictional review at Zelvani, 350 quotations included 98 wins, 142 losses and 110 unresolved offers. The first calculation produced 28%. The resolved population contained 240 offers, giving 40.8% when rounded to one decimal place. Both were arithmetically defensible descriptions. The problem arose when either was promoted into a statement about where all 350 would end. An unanswered offer had not supplied a final customer decision merely because our management reporting date had arrived. The calendar that closes a report does not necessarily close the commercial process it describes.

Counting every pending quotation as a loss can understate eventual wins. Ignoring every pending quotation can create a different distortion if the resolved offers are not representative of the whole cohort. In industrial purchasing, a repeat component may receive a quick answer while a new application requires specification work and an internal budget decision. That difference is not noise around an otherwise uniform buying process. It is part of the process itself. A report that loses it can appear consistent while quietly giving easier or faster purchases more influence over the planning assumption.

I had helped approve the dashboard, so the correction was not a lecture delivered to somebody else's careless team. We had labelled the measure without doing enough to explain its maturity. Commercial colleagues reasonably used the number they were given. When different departments then preferred different versions, the argument began to sound like a contest over optimism. I needed to return it to the underlying observations. Which customers had decided, which had not, and how much opportunity had each had to reach a decision? Those questions were more useful than defending the first percentage we had published.

## Waiting is part of the commercial process

The next step was to organise the cohort by time since quotation and by relevant commercial characteristics. We did not need to invent a complicated model before showing that a six-week-old request and a six-month-old request were at different points in their lives. We needed consistent definitions of issued, won, lost and unresolved, and a record of when each status changed. Where an offer had been replaced or materially revised, the commercial team had to explain the relationship. A new version could not simply become another independent chance to win the same underlying requirement.

Customer follow-up added context, but it did not turn every conversation into a confirmed outcome. A buyer saying that a decision is expected next month is different from a signed acceptance. A salesperson's judgement about the likely result is useful as a forecast and should be labelled accordingly. The observed status and the judgement can sit beside each other without being collapsed into one field. That distinction also protects the salesperson: an honest assessment should not later be presented as though the customer had supplied a commitment that never existed.

The waiting population may contain more than one explanation. Some offers are still commercially live; others have effectively ended without a recorded refusal. A consistent closure policy helps, but a policy chosen merely to improve a reporting percentage is a poor substitute for understanding the buying cycle. I want the commercial owner to explain why an offer remains open and what evidence supports that classification. I also want old unresolved records visible enough to be challenged. Preserving uncertainty is not permission to keep every weak opportunity alive indefinitely.

There are statistical methods for analysing time to an event and incomplete observation. Their usefulness depends on assumptions and on the quality of the underlying process. A method cannot compensate for a record that confuses a quotation revision with a new customer requirement or an informal expectation with a completed decision. Before commissioning a more elaborate analysis, I ask whether the business can describe its event consistently. If that conversation is unresolved, the method risks lending precision to a disagreement that belongs in commercial operations rather than inside a formula.

## Make a decision without pretending the cohort is finished

Management cannot always wait until every quotation has matured. A manufacturing investment has its own lead time, and a decision deferred can carry real cost. The purpose of showing the unresolved population is therefore not to prohibit action. It is to help management understand which part of the investment case rests on observed outcomes and which part rests on a judgement about the future. Those are different kinds of support. A responsible paper can contain both, provided it does not disguise one as the other to make the recommendation appear less contestable.

Our review cost ₹29 lakh and contributed to deferring a proposed ₹17.5 crore capacity commitment for four months. I would not describe that as a saving of ₹17.5 crore. The business still had to decide whether to invest, and delay could mean losing an opportunity or paying more later. The review purchased a better account of the demand evidence at a known cost. Whether the deferral was commercially wise requires the subsequent decision and its consequences, not a retrospective claim that a cautious analyst protected the whole investment amount.

The revised discussion used a range of demand assumptions and retained the unfinished cohort explicitly. Commercial colleagues could argue for a more ambitious view, but the argument had to explain what they expected from the pending offers. Operations could explain the consequence of underestimating demand, while finance could test the commitment under less favourable assumptions. My role was not to choose the company's appetite for risk. It was to prevent the reporting convention from choosing it accidentally by making one incomplete snapshot look like a settled customer verdict.

I would also agree when to revisit the assumption. A cohort review becomes more useful when the next observation is specified before results arrive: which pending offers will be followed, what status will count as resolved and what change would materially affect the investment case. That does not mean mechanically reopening every decision whenever a number moves. It means preserving a route by which the company can learn whether its assumptions were reasonable. Otherwise a revised dashboard may simply become the next static number people repeat until somebody finds a reason to dispute it.

For a board, the useful question is often modest: what has not yet happened in the evidence behind this recommendation? A director does not need to rebuild the analysis to ask it. The answer may expose an unresolved customer decision, a period that has not elapsed or a population that was excluded for convenience. In our case it exposed 110 offers still waiting for a final answer. Giving those offers a visible place in the story made the recommendation less tidy and more intelligible. That is the standard I want our data work to meet.