Sales sold a workflow. Product built a field. Three months later the customer is polite about it in the review call and then quietly stops logging in. Nobody lied and nobody was lazy, which is exactly why this argument comes back every quarter with a new deal's name on it.
I have sat in both chairs, and the failure has the same shape every time. The promise was made out loud, in a room, to a buyer who was nodding. It was never written down anywhere product could read it. What travelled across instead was a compression of a compression: a line in a CRM note, a verbal hand-off at a standup, a message that said "they need better reporting".
This is not the question of whether to build the thing at all, which I covered in should you build the custom thing they want?. Assume you said yes and you meant it. This is about what happens to the promise on its way across the company.
Why the promise does not survive the trip
Three translations happen between the buyer's mouth and the backlog, and every one of them loses something. The customer describes a problem in their own language. The salesperson rewrites it as a request, because a request is the thing that closes. Product then reads the request and infers a problem back out of it, usually a different problem from the one that was there at the start.
Underneath that sits an honest conflict. Sales is accountable for a signature by a date. Product is accountable for a system that still works in two years. Both of those are correct, and neither side is being difficult. The argument is a symptom of two scoreboards, not two bad attitudes.
The handover is five lines, not a specification
You do not need a process for this. You need a paragraph, written by the person who made the promise, in the deal record, before the deal is handed on. Five lines:
The customer's own sentence, word for word, rather than your summary of it. The job they are trying to finish. What you committed to, in the words you actually used, and the date you said it by. What you did not commit to but they may still be expecting. And how the two of you will know it worked, stated in their terms rather than yours.
The fourth line is the one that saves you. What was implied in a sales conversation is a commitment too, and it is the only part nobody ever writes down.
How do you do this when sales and product are the same three people?
Most New Zealand tech companies are small enough that the handover is simply a conversation between people who already like each other. Technology Investment Network's 2025 report ranks the country's 200 largest tech exporters, and 67 of them turn over more than $50 million a year, which leaves about two-thirds of even that top group sitting below it. Outside the TIN200 the teams are smaller again.
That closeness is the reason nothing gets written down, and it is also the reason the handover still fails. A conversation cannot be re-read in March by the developer who was not in it, or by whoever picks up the account after the salesperson moves on. Ten minutes and one paragraph is the whole discipline. At this size that is not bureaucracy, it is memory.
What it costs when the translation is wrong
Pendo analysed feature usage across 615 of its own customer subscriptions over a three-month period and found that 80 percent of features in the average software product are rarely or never used, with about 12 percent of features generating 80 percent of daily usage. That is overseas product data from 2019 rather than a measurement of New Zealand software, so read it as the shape of the problem and not as your own number.
The shape is enough to act on. Things get built that answer a request nobody ever made, and the cost lands in three places at once: engineering time you cannot get back, a customer who is still unhappy, and a roadmap that is now partly fiction. No single function can fix that on its own, because the promise crossed all of them on the way through. Fractional marketing fills a seat. Revenue leadership is the view that joins the promise to the build to the renewal.
What changes this week
Take the three deals closing this month and write the five lines for each one, in the deal record, today. Then add a single question to your own deal review: what did we promise that is not written down anywhere? Ask it out loud, with product in the room, and let the silence do the work.
Then run it backwards on the argument you had last quarter. So: what did you promise in the deal you closed last month that product has never once seen in writing?
Not sure where your revenue is leaking?
In about 6 minutes, the free Growth Scorecard shows where your sales are leaking and what to fix first.
Get your free scorecardSources. Unused features: Pendo, "The 2019 Feature Adoption Report". The report states that "80 percent of features in the average software product are rarely or never used" and that "an average of 12% of features generate 80% of average daily usage volume". Its stated method is an aggregation of anonymised Pendo usage data over a three-month period across 615 Pendo subscriptions held by customers who had used Pendo for more than a year, with a feature defined as a tag in Pendo's own product and tagged features assumed to be expected daily, weekly or monthly use. The sample spans banking and finance, HR technology, education, shipping and logistics, healthcare and e-commerce. No New Zealand breakdown is published, the data is from 2019, and the sample is Pendo's own customer base rather than the software market as a whole, so I use it as the shape of the problem rather than as a figure for this market. The same report carries an estimate that publicly traded cloud software companies invested up to US$29.5 billion developing those features; that number is built by taking averaged research and development spend from 54 companies on an index and multiplying it by the 80 percent share, so it is an estimate stacked on an estimate and I have left it out rather than quote it. New Zealand company sizes: Technology Investment Network, "2025 TIN Report" press release, Auckland, 5 November 2025, which reports 67 TIN200 companies with revenue above $50 million and 40 with revenue of $100 million or more, out of the 200 ranked, alongside total TIN200 revenue of NZ$20 billion in FY2025 and exports of $15.31 billion. The "about two-thirds" above is my arithmetic: 200 less 67 is 133 companies below $50 million, and 133 divided by 200 is 66.5 percent. I have deliberately not multiplied the Pendo share by any New Zealand research and development figure, because one is overseas product telemetry and the other is local revenue reporting, and the product of the two would not mean anything. The three translations, the five-line handover, the ten-minute discipline and the deal-review question come from my own work with New Zealand tech founders rather than from published data.