How do you protect a roadmap from being rewritten every time a new model launches?

Larkspur Residential runs FixQueue, a tool that reads a tenant's photo and two sentences about a leaking faucet or a broken thermostat, then ranks the ticket and routes it to the right vendor. Ijeoma Adeyemi built its roadmap. Every time a rival ships something flashier with a new model, her CEO takes the roadmap back and rewrites it himself, alone, over a weekend.

How to answer this, stage by stage

Nobody is scoring whether you can define a roadmap. They're scoring whether you can name the 48 hours that were missing.

Let's learn

Here is what happens when a plan survives two years of quiet trust and then gets rewritten in a single weekend by the one person who agreed to stop touching it.

Before FixQueue, a Larkspur maintenance coordinator read every tenant message by hand, guessed at urgency, and called around for an available vendor, about twenty-five minutes a ticket. FixQueue reads the photo and the message, ranks the ticket, and suggests a vendor in under a minute. Ijeoma shipped it behind a roadmap with a fixed cadence: a quarter of planned work, a small change-budget for anything urgent, and a monthly check-in instead of daily requests from the exec team.

Five questions, in order. Only one of them, I, is genuinely hard to answer.

For two years, that cadence worked. Roland Kessler, Larkspur's CEO, stopped emailing Ijeoma at 11pm with "have you seen this new model" messages. He let quarters run. He asked questions in the monthly check-in instead of mid-sprint.

Here's the turn: the habit that broke wasn't Roland losing interest in the roadmap. It was the opposite: he trusted it so completely that he had no small, fast way to react when something felt urgent. So when a rival's launch finally spooked him, there was no dial between "let it run" and "take it back and redo the whole thing myself." He only had the second option, because nobody had ever built the first.

At its worst, this doesn't just cost a weekend. Six weeks after the rewrite, half of the quarter's original work had to be re-scoped, re-estimated, and re-explained to engineers who had already planned their sprints around the version that existed before Roland touched it.

What I would leave alone: routine vendor patch notes, the small monthly updates that don't touch the two bets the roadmap depends on, don't need the 48-hour memo at all. Running the full process on every minor release just teaches people to skip it when it actually matters.

The lesson: a roadmap doesn't get rewritten because someone stops caring about it. It gets rewritten because the only fast option left is the biggest one. Build a small one, and the big one stops being anyone's only move.

Now here is the same thing as a story

The short version above is what you'd say defending a roadmap decision to your own exec team. Read this one for how a quiet Sunday habit turned into a six-week mess.

Sunday nights at Larkspur used to be quiet for Ijeoma. Roland read industry newsletters on the couch, and for two years, whatever he read there stayed there until the Monday check-in, if it came up at all.

Ijeoma had built that habit on purpose. FixQueue's roadmap ran on a fixed quarterly cadence: a locked set of bets, a small change-budget for anything genuinely urgent, and a monthly review instead of a running conversation. Early on, Roland pushed back on almost everything. By the second year, he'd stopped. The plan kept working. Tenants got routed faster. Vendors showed up for the right jobs. He let it run.

For two years, Roland's trust moved like the left panel. It broke like the right one.

Then came a Sunday like the others, except for one headline. Birchgate Property Group, a rival two towns over, announced an "instant repair diagnosis" feature built on a brand-new multimodal model, and a trade outlet ran a glowing writeup with a screenshot.

Roland didn't wait for Monday. He spent the weekend rewriting the roadmap himself, adding a rushed "instant diagnosis" initiative, reshuffling three planned features to make room, and handing the new document to Ijeoma's team on Monday morning as final.

Two years of the plan holding, undone in the same weekend as the headline that triggered it.

Nobody at Larkspur decided, on any single day, that Roland should get to redo two years of planning discipline over one article. But structurally, he only ever had two settings: let the roadmap run untouched, or take it back completely. There was no third option sized for "check if this actually matters," so the moment he felt he had to act, the only action available was the big one.

Engineers who had scoped their sprints around the old roadmap spent the next month and a half re-estimating work under the new one, much of which quietly reverted back to the original plan once someone finally tested the new model against Larkspur's own tickets and found it barely moved routing accuracy at all.

Ijeoma designed for a dial. What Larkspur actually had was a switch with no setting in between.

So here is the decision I would take back: build the 48-hour triage memo before the next headline, not after the next rewrite.

The gate that was missing sits exactly where the whole weekend happened instead.

With the memo in place, the replay runs differently. Birchgate's launch hits on a Sunday. Roland forwards the article Monday morning instead of opening the roadmap document himself. The memo comes back by Wednesday: the new model helps with photo classification cost, a little, but doesn't move either of FixQueue's two real bets, routing accuracy and vendor trust, on Larkspur's own tickets. The roadmap changes by exactly one line, a vendor swap for a small cost saving. Everyone else's sprint holds. And the thing I'd tell myself, if I could go back: Roland was never the problem. The problem was that I never gave him anything faster than "trust me" or "let me take it all back."

The five steps, if you want to remember it

The recap, one line per letter: find is Roland on a Sunday couch, locate is the habit of letting a quarter run untouched, identify is the switch between full trust and a full weekend rewrite, pinpoint is the missing 48-hour gate, and show is the same headline landing as one changed line instead of a torn-up roadmap.

And if you want to be sure it really works, try it somewhere else

Fernhollow Veterinary Partners runs a scheduling assistant that reads a pet owner's message and suggests an appointment slot and urgency level. Tobias Renwick, the clinic's scheduling coordinator, used to log every time he overrode the assistant's suggestion in a shared notes file the whole front desk could read, so the pattern of overrides was visible to everyone. Mapped onto FLIPS, the person is Tobias, the habit is logging overrides openly, and the flip here is a different family entirely: a concealment flip. After the clinic's owner personally rewrote the whole scheduling roadmap the first time a scheduling vendor announced a flashier new model tier, unprompted, Tobias quietly stopped logging his overrides, worried that a visible pattern of "the assistant got it wrong again" would trigger another unplanned rewrite. Pinpoint lands on the same root cause as Ijeoma's story: no small, fast, pre-agreed gate existed between a vendor's announcement and a full re-plan, so the owner's only move, once spooked, was to take the whole thing back himself. Show the replay: with a triage step in place, Tobias goes back to logging every override in the open, because a vendor's announcement now gets a calm 48-hour answer instead of an unplanned rewrite that makes visible mistakes feel dangerous to report.

Three of the four branches are the design working. The fourth branch is the same missing gate, in a different building.

Swap the trigger and it still runs.
Speed: an interviewer caps you at sixty seconds. Say "build a 48-hour triage memo tested on your own data, with someone below the CEO able to close small changes," and stop.
Cost: no budget for a formal process this quarter. Say so honestly, and start with just the one-page memo template and a named owner, since that costs almost nothing and stops the biggest source of wasted hours.
The model gets better, for real: if a new model genuinely does move one of the two real bets, that's the memo working as intended, and the honest move is to open the roadmap on purpose, not because someone panicked over a headline.

Where people run it wrong.
They try to ban the CEO from reacting to industry news, which never survives contact with a real headline.
They test the new model against the vendor's own benchmark instead of their own tickets, and greenlight a change that never actually helps.
They wait for the next rewrite to build the memo, instead of building it before the next headline lands.

How to use it live. The moment someone asks about protecting a roadmap from constant rewrites, ask yourself: what's the fastest trustworthy answer this person could get instead of taking the whole thing back themselves? Build that, and the rest of the answer follows.

Flashcards (tap any card to flip it)

Check yourself Score: 0 / 0