Data-informed beats data-driven

You have seen both teams.

The first one ships on vibes. The roadmap comes from the founder's gut, the loudest customer, or whatever the competitor launched last week. Data exists somewhere in a dashboard nobody opens, and decisions feel fast right up until the quarter where three of them turn out wrong at once.

The second one cannot move. Every decision waits for one more cohort, one more experiment, one more segment cut. The team is rigorous, well-instrumented, and slower than everyone around it. Certainty is always one analysis away.

Different cultures, same root cause: neither team has decided who is actually making the decision - the people or the numbers.

The 70% rule

Jeff Bezos put numbers on this in his 2016 shareholder letter: most decisions should be made with around 70% of the information you wish you had. Wait for 90% and, in most cases, you are simply being slow. His follow-up matters just as much - being good at recognising and correcting a wrong decision makes being wrong cheaper than being slow.

People read that as a permission slip for speed. It is really a claim about cost. Past a certain point, the marginal piece of information costs more to collect than the mistake it would have prevented, because the market moves while you analyse. The 70% is not a threshold to measure. It is a reminder that the last 30% was never going to arrive anyway.

Driven versus informed

Which brings up the distinction most teams never make, between two phrases that sound identical and are not.

Data-driven: the data makes the decision for you. Whatever the number says, wins.

Data-informed: the data points the direction. You decide.

Data-driven sounds more rigorous, which is why it won the buzzword war. But taken literally it is an abdication. Numbers do not carry context: they cannot see the strategy, the thing the data does not measure yet, or the customer segment too small to move an average but central to where you are going. A team that lets the metric decide has not removed human judgement, it has just hidden the judgement inside whoever chose the metric.

Data-informed keeps the accountability where it belongs. The evidence narrows the field, then a person makes the call and owns it. That is also the honest description of how good product decisions have always worked: evidence has grades, from opinion up to test result, and knowing how strong your evidence is tells you how much weight it can bear. It never tells you what to do. That part stays a decision.

The question that replaces both extremes

The best product teams I have worked with do not ask "are we sure?" They ask a smaller question: do we have enough signal to take the next step?

Not the final step. The next one.

That reframe works because it matches what a product decision actually is: a hypothesis. You assume, you test, you iterate. Under that model, no single decision needs to be right, it needs to be worth the cost of finding out. Five interviews are enough signal to prototype. A prototype in front of eight users is enough to build a slice. A slice with real usage is enough to invest properly. Each step buys the information the previous step could not.

This is also why the size of the decision sets the bar, not a universal standard of proof. A reversible change deserves a 70%-information decision made this week. A one-way door - pricing, platform, a segment change - deserves more evidence, gathered deliberately, and a designed experiment is usually the cheapest way to buy it. The failure mode of the data-driven team is applying the one-way-door standard to everything. The failure mode of the vibes team is applying the reversible standard to the one-way doors.

The reframe also changes meetings. "Are we sure?" produces debates about confidence, which are unwinnable because confidence is a feeling. "Do we have enough signal for the next step?" produces a checklist: name the step, name the signal, check whether you have it. When the answer is no, the meeting output is not another opinion, it is the cheapest action that would produce the missing signal by next week. Teams that make this switch do not become reckless. They become specific about what they are unsure of, which is the actual thing rigour was supposed to buy.

Where the dashboards fit

None of this argues against instrumentation. It argues for a specific job description: data exists to generate and check hypotheses, not to replace the person making the call. A metric moving is the start of a question - which is precisely the role metrics play in the 0-to-1 phase, where the numbers are too small to decide anything and exactly the right size to point somewhere.

Teams usually cannot fix this alone, because the extremes are cultural. A leadership that punishes wrong calls breeds certainty-waiting; a leadership that celebrates shipping breeds vibes. Resetting how decisions get made - what needs evidence, what needs speed, who owns the call - is decision hygiene, and it is a big part of what ongoing product advisory work looks like in practice.

So next time a decision stalls, do not ask for more data by default. Ask what the next step is, what signal would justify it, and whether you already have that signal. About 70% of the time, you do.