The customers you're winning aren't the ones you're selling to
The deck still said mid-market. The closed-won list was mostly enterprise. The team ran the wrong playbook and called the gap a tough quarter.
At a Series B company, the head of revenue described the ideal customer to me without hesitation: mid-market, fifty to two hundred employees, software or fintech. The deck said it. The sales playbook said it. The marketing site was built for it, and the SDRs sourced their lists against it. Then I asked her to pull the last 20 customers the company had closed. She pulled them. The median was two hundred and fifty employees. Eleven of the twenty sat in industries the deck never mentioned. The customer base had drifted into enterprise and into adjacent verticals, and no one had touched a single document.
The sales team was still cold-calling companies of fifty to two hundred employees in software and fintech. Pipeline conversion had been sliding for two quarters. Marketing was generating leads against the old profile, and the leads weren't converting, because they weren't the customers the company was winning. Real revenue was still coming in. It was coming in almost by accident, through inbound and referrals, against a buyer the prospecting motion no longer described.
This is the most predictable form of go-to-market entropy. The early ICP, the one you defined when you were first working out who to sell to, produces a handful of customers. Those customers refer companies that look a little different from them, and the slightly different companies close too. The product evolves toward their use cases. Customer success builds playbooks for them. Pricing bends around their procurement. Six rounds of "slightly different" later, the customer base has nothing to do with the original profile, and nobody has noticed, because the team is still running the playbook that produced the first cohort.
The drift compounds through three mechanisms at once. Inbound moves first. Word of mouth follows whoever the existing customers naturally evangelize, so if the base has tipped toward enterprise, the referrals arrive enterprise. Sales treats those leads as edge cases instead of reading them as the buyer signal they are.
Product moves second, on a lag. The roadmap accommodates the customers who are in the product today, not the ones on the slide. After two years the software fits the drifted buyer better than the original one, and the sales team is selling a product that no longer matches the profile it's pitching. Conversion decays in ways the team files under "execution."
Pricing moves third. The customers the company now wins have different willingness to pay, different procurement, different decision cycles. Pricing built for the old ICP either overshoots the new buyer or leaves money on the table. The deals that close are the ones where a discount can bridge the gap. The deals that stall are the ones where the gap is structural.
The cost shows up in three budgets. Prospecting efficiency, because reps are dialing companies that don't match the real pattern. Marketing spend, because the site and the campaigns are tuned to a buyer who's decreasingly likely to purchase. And product, because the roadmap keeps serving assumptions about a customer who has already moved on. This is the same measurement gap that shows up when a feature request is a churn signal in disguise: the label on the data stops matching the data.
The diagnostic is cheap, and almost nobody runs it on a schedule. Once a quarter, the head of revenue pulls the last twenty closed-won accounts and clusters them by segment, size, industry, buyer role, and trigger event. Compare the median to the written ICP. If it has moved, and after a year it almost always has, update the profile, re-aim the prospecting, adjust the positioning. The exercise takes two hours. It runs four times a year. The output is a one-page memo naming the real ICP and the changes needed to line the team up behind it.
The reason it doesn't happen is political, not analytical. Updating the ICP means admitting the version in the deck, the one the board keeps quoting and the marketing team is proud of, is wrong. It means retraining sales and redoing positioning work people were proud to finish. Every function has built something against the old profile, and every function eats some cost to realign. The path of least resistance is to keep the old documents and let the drift run.
And the drift produces the decay. The team explains it away as a tougher market, or an execution gap, or a need to spend more on pipeline. Each story defers the real diagnosis, which is that the team is aiming the wrong positioning at the wrong audience and grading itself against a conversion baseline that no longer applies.
Read your own customer list every quarter. The deck is a hypothesis. The closed-won is the data. When the two disagree, the deck loses, every time. Before your next leadership review, ask who's running the closed-won cohort analysis and when it last ran, what the process is for updating the ICP once drift shows up, which conversion trend the team has been blaming on the market, and which function will fight the update hardest. If nobody owns the analysis, the team is flying on the ICP it wrote down, not the one it's living.