You Don't Have A Product Talent Problem. You Have An Operating Model Problem

Rob Versaw, Product Strategy at Dynatrace, bridging innovation and business impact.

gettyEighty percent of features in the average cloud product are rarely or never used, and Pendo put the price of building them across publicly listed software companies at about $29.5 billion. It is not a talent failure. It is the accumulated output of an operating model that keeps sending capable people to build the wrong things.

Walk into most companies complaining about product quality and the brief sounds like: “We need to hire much better product managers.” Too slow. Too junior. No ownership. Not commercial enough. The diagnosis arrives prepackaged with the remedy, usually before anyone has looked at how the team actually operates.

Look closer and the pattern repeats with unsettling consistency. What’s usually there is not a capability gap. It is some combination of product managers who aren’t permitted to speak to customers, priorities that reset three times in a quarter, a P&L nobody has ever seen, four executives requesting four different things and a roadmap set without the PMs. They chose none of it. They are accountable for all of it.

Put a strong product manager in that environment and they will look weak. Give it 18 months and they will be somewhere else, because the strong ones always have options.

Deming’s estimate was that 94% of trouble belongs to the system and is management’s responsibility, leaving 6% to special causes. You don’t need the exact split to accept the principle. Anyone who has run a value stream map knows the pattern: the operator is rarely the constraint. The handoffs, queues, rework loops and approval gates are. Product management is a value stream like any other, and most of us have never mapped it.

Marty Cagan drew the operative line years ago: An empowered product team is given a problem to solve, and a feature team is given a list to build. Most organizations complaining about product caliber are running feature teams and recruiting for empowered ones. They interview for commercial judgment and conviction, then hire someone whose actual job is backlog administration.

Centralized product control now has a respectable name. Since Paul Graham’s essay, “founder mode” has become the standard defense of hands-on executive direction. Founder involvement in the detail is frequently correct, particularly before product-market fit. It is not a decision-rights model. Declaring yourself product-led while operating in founder mode is an unresolved architecture, and your best people will read it as one.

Most of what degrades a product function isn’t dramatic. It is a small set of contradictions, repeated until they function as policy:

1. Withholding The inputs, Then Grading The Outputs: Customer contact is restricted, the data layer is never funded and the analysis that does get commissioned is discarded when it disagrees with the prevailing view. The team is then marked down for not being customer-led or evidence-driven.

2. Accountability Without Authority: Decisions are made above the team, wire frames are approved individually, kick-offs are canceled and the resulting build is criticized. Product becomes the standing explanation for commercial misses it had no hand in shaping.

3. Priority Churn Presented As Responsiveness: The ranked list resets every fortnight, each executive attaches a pet project and the teams are reorganized every six months under the banner of agility. Nothing ships, and the throughput question lands on the PMs.

4. Commercial Expectations Without Commercial Visibility: Revenue, margin, cost-to-serve and retention sit behind an access wall, and the same people are asked why their thinking isn’t commercial enough.

5. Ambition Without The Funding To Match: Bold bets are requested and only safe ones approved. AI-native delivery is mandated while tooling and token budgets are capped. Quality is raised in every review and tech debt is never scheduled.

None of this requires a bad actor. It requires busy people optimizing locally under pressure. That is exactly what makes it systemic, and exactly why replacing individuals doesn’t shift the output.

The 2025 DORA research, drawn from nearly 5,000 technology professionals, landed on a finding worth considering: AI doesn’t fix a team; it amplifies what is already there. Strong teams compound. Struggling teams find their existing dysfunctions magnified, and the return comes from platform quality, workflow clarity and team alignment rather than from the tools themselves.

Read that against an unclear roadmap and an unstable priority stack. You are about to accelerate the rate at which your teams build things nobody uses. Faster feature factories are still feature factories. The bottleneck was never typing speed.

The Australian constraint sharpens it. The technology workforce contracted last year for the first time in the 12 years the ACS has tracked it, with 259,000 additional professionals needed by 2035. Build capacity is the scarce asset. Pointing it at the wrong problems is the expensive failure, not paying over market for a PM.

1. Publish decision rights. One page: Who decides problem selection, solution design, sequencing and pricing? Ambiguity here generates most of the swirl your PMs are blamed for.

2. Restore customer access. Set a monthly floor of direct customer contact per PM, with no sales agenda attached to the conversation. Notify the account owner every time and send the findings back to them first. You are removing the commercial intent from the discussion, not the commercial team from the loop.

3. Open the numbers. Give every PM their product’s revenue, margin, cost-to-serve and retention curve. You cannot outsource commercial judgment to people you keep commercially blind.

4. Freeze the ranked list for a quarter. Changes go through an explicit trade-off, recorded, not through a corridor conversation.

Then reassess. If the same complaints survive two quarters in a stable environment, you have a hiring problem and you should act decisively on it.

Until then, you’re grading people on a test you designed. Fix the system before you replace the people.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?