A roadmap is a treaty, and Productboard sold it as a forecast

The quarterly roadmap review, in the product organization I joined after leaving consulting, ran on a slide with three columns: Now, Next, Later. Forty minutes in, a sales leader arrived carrying a spreadsheet of customer requests he had tagged himself, forty of them, each one tied to a renewal he was defending or a competitor he was losing. He did not need the roadmap changed. He needed the roadmap to contain his tags, and by the end of the meeting his requests had been allocated a column and he left visibly lighter. Everyone in that room understood that items placed in Later were there to die. He understood it too. The column satisfied him anyway, and I watched him leave before I understood why: the slide was never a statement about the future. It was a signature page, and he had gotten his name on it.

A roadmap is not a forecast. It's a treaty, a political settlement among sales, engineering and the board about which promises each faction can live with for one more quarter, and Productboard, a San Francisco and Prague company founded in 2014 and launched at TechCrunch Disrupt in September 2016, has built an entire business on making that settlement look like evidence. The company then spent 2026 proving my point about itself, on itself, in public, which is more intellectual honesty than most of its marketing ever managed.

The counted anecdote is the ammunition, and the counting is the product

The mechanism is elegant. Sales and support drop customer requests into a portal, the requests pile up in an insights inbox, each note gets linked to a feature, each feature accumulates a score, and out the other end comes a prioritized roadmap with a number attached to every row. What the count changes is the grammar of the argument. Before it, a sales leader argues from an anecdote, and anecdotes lose to opinions of equal volume. After it, he argues from a total, and a total sounds like data even when the tagging was done by whoever shouted first.

Look at who pays, because the design is candid about it. Pricing is per maker: the free plan is $0 and includes the company's new Spark agent, Plus is $19 per maker per month billed annually or $25 monthly, Business is $59 or $75 with a two-maker minimum, and Enterprise is custom. The people being counted, the sellers, the support team, the customers filling in portals, are all free. Consider the incentive. The payer is the person who wants the count to exist, and the counted supply their own data without ever seeing an invoice, which means the system's politics are fixed by its pricing: it will always look like evidence from the chair that paid for it, and like a backlog of other people's opinions from every chair that didn't.

The treaty seller renegotiated its own treaty

Then the market tested the product's premise, and the premise lost. Productboard raised a $125M Series D at a $1.725bn valuation in February 2022, employed more than 500 people at its peak that spring, and its chief executive told Bloomberg an IPO was planned for 2025. Instead:

The April announcement came from the chief executive's own keyboard: Hubert Palan wrote that the company was becoming an AI-only organization built around Productboard Spark, the agent that reached general release on June 22, 2026 and drafts the plan the old product merely organized. In July, co-founder Daniel Hejl told the Czech tech site CzechCrunch that on the old platform "it was clear we wouldn't survive." The Mighty Nine European Unicorn Stress Report 2026 estimates the valuation at about €600M, down roughly 60 percent, a figure Palan calls speculation, and the legacy product still earns most of the revenue. I've written before that a tool reveals its maker's theory of the world. The maker of the notarized treaty could not afford the notary, and the buyers of roadmaps were the first budget line cut, which tells you exactly how load-bearing the count turned out to be when the quarter got tight.

Without the count, the loudest voice wins

The case for Productboard deserves to be made at strength, because in most organizations the alternative to a counted trail is not Marty Cagan's empowered team. It is a hallway. Cagan's 2015 essay against feature roadmaps argues for continuous discovery and outcome-based teams, and it is the correct advice for companies with trust, patience and product leadership already in place. From inside a quota, the essay reads differently. A product manager with a sales VP across the table, a renewal at stake and thirty minutes of airtime has exactly one weapon, and it is the tagged trail, because the trail converts his anecdote into a number the VP must argue against in front of witnesses. The inbox also creates a memory the meeting cannot revise: the forty tagged requests from my opening scene are still in the system, dated, even the ones that died in Later.

I'll grant the rebuttal to this defence, too, since I lived it. Tagging is free and unpriced, so the count can be manufactured, and the sales leader with forty tags and an afternoon can out-vote the support inbox that accumulates forty over a year. The instrument that protects the product manager from the loudest voice can be captured by the same voice with better spreadsheet habits. A treaty is only as good as the enforcement nobody has.

It depends, in the end, on what the count is for. As armor for a negotiation the tool is honest work at $19 a maker; as a substitute for the negotiation it is a notary stamping a treaty that was going to be signed anyway. The variable worth watching is whether Spark's drafted plans change what the room ratifies or only who holds the pen. Later still means never. Now the column writes itself.