What does a roadmap look like when it is organized by uncertainty rather than by quarter?

Every day between fares, Marisol Duarte checks whether the bill she asked an app to fight for her actually got fought. Rovengale is a consumer app that negotiates bills and cancels subscriptions on a customer's behalf. Callum Iverson is Head of Product, and the question on the table is what the roadmap should look like once "when will this ship" stops being the same question as "will this ever actually work."

How to answer this, stage by stage

Nobody is scoring whether you can draw three boxes instead of four. They're scoring whether you can say what happens to a real customer once the roadmap format itself starts lying by omission.

Let's learn

Every Tuesday, between fares, Marisol Duarte used to spend about twenty minutes on hold with her internet provider, reading a script off her phone to talk her bill down. It worked, most months, for about forty dollars off.

She signed up for Rovengale eight months ago. The app calls providers on her behalf, and for well-known ones, a negotiation now takes about ninety seconds of her time: approve the ask, wait for a text. For the two smaller, regional providers she and her neighbors use, Rovengale's roadmap listed "negotiate any provider, no matter how new" as a Q3 feature, the same kind of line as every other item on the slide.

This is the workaround the whole anchor is designed to make unnecessary: a person keeping her own ledger because the app won't.

Here's the turn: Q3 arrived, and the "any provider" capability hadn't cleared its own bar. It still only worked reliably on the handful of national providers whose phone trees Rovengale's model had seen thousands of times. For Marisol's regional provider, a negotiation would start, sit in a "processing" state for days, then quietly fail with no explanation. The problem was never the failure rate. It was that the roadmap gave that bet the exact same shape, a box with a quarter in it, as a feature that already worked.

At its worst, a roadmap format can make a hope look exactly as solid as a fact, and the person who pays for that resemblance is the one holding the phone, not the one holding the slide deck.

What I would leave alone: the quarter labels themselves for Committed items. A confirmed, working feature genuinely benefits from a real date, and adding a confidence lane on top of something that's already solid would just be noise.

The lesson: a roadmap doesn't just tell your team what's coming. It tells your sales page, your support scripts, and eventually your users, what's already true. Get the format wrong and all three learn the wrong thing at once.

Now here is the same thing as a story

The short version above is what you'd say scoping this live in an interview. Read this one for how a near miss on a Thursday afternoon turned a slide-format complaint into an actual product decision.

Marisol drives mornings and late nights, six days a week, and checks Rovengale in the gaps: a red light, a slow pickup, the ten minutes before a shift starts. She's careful with money in a way that comes from years of it mattering; she reads every text the app sends twice.

For her first seven months on Rovengale, every negotiation the app started for her national providers, internet, phone, streaming, ended the same clean way: a text saying how much she'd saved, and a line in her account history she could scroll back through any time. She trusted the app the way you trust a scale that's never once given you a strange number.

This is the anchor made concrete: the same routing logic that should decide a lane, applied to one bill at a time.

Then her regional provider raised rates, and she asked Rovengale to fight it, the same way she always had. The app showed "negotiating" for three days. On day four, distracted between two airport rides, she almost approved a full payment on the same bill through her bank's own app, assuming Rovengale's attempt had quietly failed and nobody had told her. She caught it with about ninety seconds to spare, comparing two screens at a red light.

Nothing about that near miss was Marisol's fault. She did exactly what a careful person does: she double-checked. The gap she fell into wasn't a mistake in her judgment. It was a roadmap slide, made eight months earlier, that had put "any provider" in a quarter box with no way to say the model hadn't actually agreed to it yet.

Each step made sense alone. Together, they built the exact private workaround a confidence lane was supposed to make unnecessary.

After the near miss, Marisol started keeping her own running note on her phone, one line per bill, per attempt, per provider, exactly the ledger Rovengale should have been keeping for her all along.

The product-facing version of the same anchor: what Marisol sees instead of a silent "processing" screen.

Rerun the same rate hike with the lane structure already in place: "any provider" never leaves Directional, because its threshold, ninety percent completion across a rotating sample of regional providers, was never actually cleared. Marisol's app shows "still learning your provider, human review within 24 hours" instead of a stuck spinner. No near miss, no private spreadsheet, because the promise she was given matched the promise the model could actually keep.

Same unresolved bet, two very different amounts of honesty about what it actually is.

What I'd tell myself, watching Marisol catch her own near miss at a red light: the quarter-column roadmap wasn't wrong to want a plan. It was wrong to let a hope sit in the same box as a fact, with nothing on the slide to tell the two apart.

SPARK, the lanes that survive their own betNot a slide redesign. SPARK is what turns "organize by uncertainty" into a structure a team can actually run.

The recap, one line per letter: situation is every roadmap item flattened into the same quarter box today, payoff is stakeholders learning to ask "what would move this up" instead of "is it still on track," anchor is the three confidence lanes with a threshold or a date but never both, risk is a bet leaking downstream as a customer promise before it clears its bar, and keep out is a manual monthly review instead of an automated one, at least for now.

And if you want to be sure it really works, try it somewhere elseSame five letters, a public library's cataloguing system instead of a bill-negotiation app. Different flip family entirely, the same undated bet.

Quarrytown Public Library uses an AI assistant that suggests subject headings for new acquisitions, meant eventually to catalogue books without a librarian checking every suggestion. Mapped onto SPARK: situation is Agnes Petrosyan, a cataloguing librarian, checking every suggested heading by hand today, same as before the tool existed. Payoff is Agnes trusting the model enough on well-documented titles to only spot-check, freeing her for the genuinely obscure ones. Anchor is the same three lanes: heading suggestions for mainstream nonfiction sit Committed, suggestions for small-press or self-published work sit Directional with a named accuracy threshold, and suggestions for material in translation sit Exploratory with no promise at all yet. Risk is a Directional suggestion getting auto-applied to satisfy a backlog metric before its threshold clears. Keep out is any plan to fully automate translated-work cataloguing this year; it stays hand-reviewed. The flip here is concealment, not workaround: once a patron complained publicly about a mis-catalogued translated novel, Agnes stopped mentioning to her cataloguing team that she was using the AI's suggestions at all for anything outside the safest, most mainstream titles, quietly cutting off the exact feedback that would have shown which suggestions were actually safe to trust.

The same lane logic, one library shelf at a time: what's solid, what's a bet, and what still needs a flag.

Swap the trigger and it still runs.
Speed: an interviewer caps you at sixty seconds. Say "never let a bet share a box with a fact, and never give a bet a customer-facing date," and stop.
Cost: no time to build named thresholds for every roadmap line this quarter. Say so honestly, and start with the three highest-risk Directional items only, expanding coverage as the habit proves out.
The model gets better, for real: if a Directional item's threshold actually clears on schedule, that's the lane working exactly as designed, and the honest move is promoting it to Committed with a real date, not quietly leaving it in Directional out of habit.

Where people run it wrong.
They add a confidence label to the slide but keep giving every item a date anyway, which defeats the entire point.
They let Exploratory items sit with no revisit cadence, so the lane becomes a polite graveyard instead of a live status.
They let a good demo pull a Directional item's promise downstream to customers before its threshold has actually been cleared.

How to use it live. The moment someone asks what a roadmap should look like under uncertainty, ask yourself: if I deleted every date on this slide, would anyone actually know less about what's real? If the answer is no, the dates were never doing the job you thought they were.

Flashcards (tap any card to flip it)

Check yourself Score: 0 / 0