Field notes again. Short, and opinionated.

The product manager wants a feature in a week. You think it is three. This is one of the most common awkward conversations in software, and most people handle it badly in one of two ways. They say yes, then miss the date and lose trust. Or they say no, sound obstructive, and the work gets assigned over their head anyway.

Both fail because they treat the conversation as a negotiation about a number. It is not. The date is a symptom. The real subject is what the business needs and by when.

First, check your own number

Before pushing back, make sure three weeks is honest and not a reflex. Break the work down. A rough example for, say, adding UPI autopay mandates to a subscription product: change to the data model, integration with the payment gateway’s sandbox, handling of failed and pending debits, notification emails, admin screens for support, tests, and deployment with a rollout plan. List each with a best case and a worst case.

A trick I like is the three-point estimate: optimistic, likely, pessimistic. If the likely case is 12 working days, the optimistic is 8 and the pessimistic is 20, then “one week” is not even in the optimistic range, and you can say so with evidence. Estimates for software are uncertain; saying “likely 12 days, could be 20” is more truthful than a single figure, and PMs generally respect the honesty once they see the breakdown.

Be sceptical of yourself, too. Developers tend to estimate the coding and forget code review, testing, environments, waiting for another team, and the production bug that will interrupt the week. If your three weeks is padded with unnamed “buffer”, say so, or trim it.

Ask what the deadline is attached to

Then ask a question that changes the conversation: “What happens if it lands in three weeks instead of one?” Sometimes there is a real, immovable reason: a regulatory deadline, a launch with a partner, a festive sale. Sometimes the date was picked because a senior person mentioned a week in a meeting. You will not know until you ask, and the answers lead to completely different plans.

Ask, too, what the minimum useful version is. A PM who says “we need everything” often means “we need a version we can demo to the customer”. Those are different.

Offer options, not a refusal

The strongest response is a short menu. Using the same example:

  • Reduce scope. In one week we can deliver mandate creation and the successful-debit path for one bank and one plan, with the failure handling done manually by support for now. Full version follows in week three.
  • Move the date. The full feature, tested, in three weeks.
  • Add people or drop other work. Honestly, adding a developer rarely speeds a one-week task, because onboarding takes time. Pausing another committed item can help; say which one it would be.
  • Accept the risk, knowingly. We can try for one week with less testing, but the likely outcome is bugs in production affecting real money. If you choose this, I would want that decision recorded.

Look at what has happened. The PM now has choices that connect cost to outcome, and the decision is theirs, which it should be. Your job is to make the trade-off visible, not to win.

What not to do

Do not quietly cut quality to hit the date. Skipping tests and review is borrowing from the future without telling anyone, and it is your name on the incident report. If scope is truly fixed and time is fixed, quality is the thing that bends, and the person who set the constraints must know that.

Do not say “it depends” and stop there. Say what it depends on, with figures.

Do not agree to “try” and then work weekends to make it. Heroics teach the organisation that your estimates are negotiable. The next request will be shorter still. My view: a single emergency push for a real business reason is fine; a pattern of them means your estimates are being overruled.

Keep the numbers honest over time

Track your estimates against actuals for a few months. If your “likely” numbers are consistently 30% low, that is useful, and you can calibrate: a measured history is the best answer to “why does this take so long?”. Splitting work into pieces of a few days each also helps, because small tasks are estimated more accurately than large ones, and slipping on a small piece is an early warning, not a late surprise.

When you must commit, give a range with a confidence level and re-forecast openly at the halfway point. “On track for the 12-day likely case; the gateway sandbox is the main risk” is the sort of sentence that builds the relationship in which a “no, three weeks” is believed the next time.

Figures and names in the examples are illustrative.

Leave a Reply

Trending