Short answer
Every custom yes to a big account trades tomorrow’s product for today’s deal, and sometimes that trade is correct. The trap is making it by reflex, one deal at a time, until most of engineering serves five accounts. The way out is a three-bucket decision standard applied before the contract, not after.
Sales has a large deal on the table, contingent on three small features. Two of them exist nowhere on the roadmap and serve no other account. The last four deals closed the same way, which is why this quarter’s roadmap is actually last quarter’s promises being paid off.
Meanwhile the self-serve motion that was supposed to be the growth engine has not shipped anything meaningful since spring. Nobody decided to stop investing in it. It simply lost every individual scheduling argument to a signed contract, which is a much harder opponent than a strategy document.
The three forces that are invisible at signature
None of these appear in the deal review, which is why capable teams walk into this repeatedly. Each operates on a delay measured in quarters, so the feedback that would teach you to stop arrives long after the behaviour that caused it has become normal. By the time margin decay is visible in a board pack, three years of decisions have already been made.
Margin decay
Custom work is build once, sell once. Your pricing assumes build once, sell many. Every bespoke feature moves a slice of the company from the second economics to the first, and because it arrives as revenue rather than as cost, nothing in the reporting flags it. The margin line notices two years later.
Roadmap capture
Each promise is a claim on future engineering capacity. Stack enough of them and the roadmap belongs to the accounts you already have rather than to the market you are trying to win. The single-account version of this is the loudest customer problem; this is the systemic version, and it is harder to see because every individual claim was legitimate.
Valuation drag
Acquirers and investors read revenue quality, not revenue labels. Recurring platform revenue is priced generously and services-shaped revenue is discounted heavily, regardless of how the contract is worded or which line of the P&L it lands on. Companies discover this during diligence, which is the most expensive possible moment.
“Nobody ever decides to become a custom development shop. They decide to close eleven deals, each one obviously worth it, and then look up three years later to find the roadmap belongs to five accounts and the product has not moved since.”
The decision standard that replaces the reflex
Every request goes into one of three buckets before the proposal reaches the customer, not after the contract is signed. The classification is the whole method, because it converts a yes-or-no argument under closing pressure into a question with an evidence standard attached. The bucket call itself needs an owner with customer data and the standing to hold it while a deal is closing.
- Core-accelerating. At least three other accounts, with evidence rather than intuition, need this within a year. Build it: the deal is subsidising your roadmap and you should thank them.
- Configurable. The underlying need is real but the requested shape is account-specific. Build the setting, not the feature, and you have added something every future customer can use.
- Bespoke. Serves this account only. Either decline, or price it honestly as professional services on a services line, scheduled so it never displaces core work.
The third bucket is where the discipline lives. A bespoke yes is not the failure; an unpriced, unlogged bespoke yes is. Price it at what it genuinely costs including the opportunity cost of what it displaces, put it where finance can see it, and record the roadmap lien so somebody can count them later.
What good looks like in 90 days
Ninety days is enough to write the policy, inventory the damage and prove the policy survives contact with a real deal. That third step is the one that matters, because a customisation policy that has never been tested against a closing contract is a document rather than a rule, and everyone in sales knows the difference.
- A written customisation policy that sales can quote inside a deal: what we build, what we configure, what we price separately.
- The existing promise backlog inventoried and honestly rescheduled, so you know the size of the lien before adding to it.
- One large deal closed with the policy rather than around it.
- A services line in the reporting, so bespoke revenue is visible as what it is.
Roadmap owned by five accounts?
Thirty minutes, no pitch deck. You will leave with the promise backlog sized honestly.
Book a Product Strategy SessionFrequently asked questions
We genuinely cannot afford to lose this deal. What then?
Then take the third bucket with open eyes. Price the true cost including what it displaces, put it on a services line rather than burying it in the subscription, and log the roadmap lien so it can be counted later. The trap has never been the yes. It is the yes that nobody priced and nobody wrote down, because that one is invisible until eleven of them have accumulated.
Should the customisation line report to sales or to product?
The policy belongs to product, the exception request belongs to sales, and the decision sits with product leadership using revenue evidence. If product cannot hold that line organisationally, that is the actual gap and no policy document will substitute for it. Our guide to capacity versus direction covers what to do when the authority is what is missing.
How do we know if we are already in the trap?
Count two things. First, what share of the current quarter’s engineering capacity is servicing commitments made to close specific deals. Second, how many of the last ten shipped features are used by more than three accounts. If most of the capacity serves promises and most of the features serve almost nobody, the trap closed a while ago and the work now is inventory and renegotiation.
Does saying no cost us the enterprise segment?
The opposite, more often than founders expect. Large buyers are making a decade-long bet on your survival, and a vendor who folds on every request reads as one that will be a custom shop in three years and a dead one in five. A clear policy, delivered early and consistently, is a maturity signal. Our piece on saying no without losing the account covers the delivery.
The goal is not to stop doing custom work. It is to stop doing it accidentally, one closing call at a time, until the product you meant to build has no capacity left to become itself.