Kill the Roadmap

The lie everyone is performing together

I've made roadmaps at Microsoft, Adobe, Salesforce, and four startups. Every one of them was fiction within six weeks of the date I put on the title slide. The customer need shifted. A competitor shipped something. The model changed. A top-10 account asked for something I hadn't seen coming. The plan was already wrong.

But I'd still show it at the board meeting. Sales would still forecast against it. Engineering would still plan capacity against it. We all knew it was stale. Nobody said so. Instead we'd have a "re-alignment" meeting two weeks later, which is corporate code for "we all knew it was wrong but needed permission to admit it."

That's not a planning problem. That's a group performance. And it's the single most expensive piece of theater in our job.

Why the roadmap keeps breaking

Three things about roadmaps are load-bearing and broken at the same time.

It rewards conviction theater. The PM who writes the most confident roadmap wins the most trust. That PM isn't more correct. That PM is more practiced at the performance. We reward the wrong skill.

It fossilizes old signal. The roadmap you're building today is based on customer conversations from last month, competitive data from the last quarter, and adoption curves from before the last model release. By the time it's published, it's already a fossil. And then you hold yourself accountable to the fossil.

It creates a re-plan tax. Everyone knows the roadmap is wrong. But changing it requires a whole process: exec review, stakeholder comms, a Slack thread about "the changes." So you don't change it. You execute against a plan you know is wrong because the cost of changing the plan is higher than the cost of being wrong.

Before AI, this was survivable. A quarterly plan was close enough. Now? Your competitor can ship a prototype in an afternoon. Your customer's context shifts weekly. A locked roadmap is a commitment to being slow.

What I run instead

I stopped publishing roadmaps about 14 months ago. I run a one-page live bet portfolio that updates every Monday. Three sections. That's it.

Active bets (5 to 7 maximum). Each bet has: the hypothesis, the surface we're testing it on, the signal we're watching, the kill condition, the next decision date, and cost-so-far. Not a feature list. A set of things we believe and are testing right now.

Recently killed (rolling 30 days). What we stopped, why, what we learned. This is the most important section. It's also the section most orgs refuse to make public because admitting failure feels expensive. Make it public anyway. It earns more trust than ten successful launches.

Standing queue. Things we'll pick up if the signal shifts. Not things we're working on. Things waiting for a trigger. The queue has fewer than 20 items. If it has more, you're rebuilding a backlog.

The document updates every Monday morning. It's visible to the whole company. Same version goes to the board, to sales, to CS, to engineering. If I ever feel the urge to "clean it up" for a board meeting, that's the signal my internal version isn't honest enough yet.

---

Preparing your next board deck? I review board decks for CPOs in one week, against the 2026 CPO framework, with specific rewrites and a walkthrough call before you present. See the Board Deck Review →

---

What a bet looks like

Here's a real one from a quarter I ran at Smartcat last year (details sanitized):

• Hypothesis: Mid-market buyers will choose us over the incumbent if we can show 70%+ cost reduction with comparable quality.
• Surface: The new pricing landing page plus the inline quality compare widget on the product dashboard.
• Signal we're watching: Win rate in competitive deals tagged "migration from Incumbent X," plus the 30-day retention of accounts acquired that way.
• Kill condition: Win rate stays below 25% for 6 weeks with no identified lever to pull.
• Next decision: Week 8, against Week 1 baseline.
• Cost so far: ~1 sprint of eng, ~2 weeks of PM/design, ~$3k in Claude API to run the comparisons.

That's the whole bet. One card. I could publish 100 of these in the time it takes to make one slide deck that would be dead in two months.

The stakeholder conversation you're about to have

A senior stakeholder will push back. It usually sounds like: "We need predictability for planning purposes."

That's not a bad-faith ask. They're trying to do their job. The answer isn't "planning is dead." The answer is:

"We'll commit to the bets we're making this quarter. We won't commit to the specific features we'll have shipped by the end of it. Those are different commitments. The first one is honest. The second one is a lie we were all performing together."

If that lands, you have a convert. If it doesn't, run the bet portfolio alongside the roadmap for one quarter. At the end, compare: which document predicted what actually shipped? The portfolio wins every time. Now run it without the roadmap.

The hardest part isn't the process

It's emotional. A roadmap that claims certainty feels safer than a portfolio that admits bets. You want the false comfort. Resist it.

The roadmap does not protect you. It delays the moment you have to admit the plan was wrong, and the delay costs you trust when the truth arrives. The portfolio puts the uncertainty in the open, in writing, where everyone can see what you're betting on and why. That feels exposed the first few weeks. Then it feels normal. Then the idea of going back to quarterly roadmaps feels insane.

The shift I had to make in my own head: from "here is what we will do" to "here is what we are learning." I'm not losing authority by making this move. I'm trading theatrical authority for the real kind, the kind that comes from being the person in the room who actually knows what's happening right now.

Pick one thing this week

You probably can't delete your company's roadmap on Monday. You probably can build a bet portfolio alongside it. Here's the 30-minute version.

1. Open a single page in Notion, Linear, or whatever you use. Title it "Active Bets."
2. Write down the 5 things your team is actually working on right now, as bets. Each one: hypothesis, surface, signal, kill condition, next decision date.
3. If you can't write the kill condition, the bet isn't a bet. It's a feature you're committed to regardless of signal. Note that as a finding.
4. Share the page with your team. Ask if the bets are accurate. If anyone says "wait, I thought we were doing X," that's the gap between what the roadmap claims and what's actually happening.

That's the first honest artifact. Keep doing it weekly. Within a quarter the roadmap will feel like a PR document and the bet portfolio will feel like the real plan. At that point you can have the conversation about retiring the roadmap. The evidence will already be on your side.