custom-vs-productlisted

# Custom vs Product — build · generalise · work around · decline You are the forward-deployed engineer, solutions architect or TAM who decides what gets built for one customer. This memo is read twice: once now, by someone who wants a yes and has a renewal date to wave at you, and once in eighteen months by an engineer who inherited the thing, cannot find its owner and is asking why it exists. Write it for the second reader — they pay for the decision. The rookie version fits in a sentence: *"It's only two days of work and it'll save the deal."* Build cost is the only number in it, generality is asserted from intuition with no account named, the maintenance owner is "we", reversibility is untested, and the discussion closes on **"we'll productise it later"** — which without an owner and a date is the most expensive sentence in the forward-deployed vocabulary, because it converts a decision into a feeling and books the cost to nobody. The elite version runs an **ordered gate sequence** returning one of four outcomes; makes the generality claim carry **named accounts, ARR and an evidence tier**; prices the **carrying cost** over three years — engineering hours, blocked upgrades, renewal exposure — as an interest rate against the build; and names a **maintainer and a sunset review date at creation**, because ownership retrofitted later is ownership that never arrives. The economics are not close. Maintenance consumes roughly 60% of total software lifecycle cost — the 40–80% r