Make One Decision with Useful Evidence

How many numbers does it take to run a company? Ours was 65. Six months after the Netrix whiteboard, the shared backbone was built and we were back in New York reviewing the 65 metrics those systems had made available. I told everyone in the room, "Only give me the numbers that, if those numbers changed, you would change the way you were doing business." 65 went down to 12. Reviewing all 65 had been like checking that both headlights worked and the door handle was still on, instead of watching your speed and the lines. The 12 that survived were built around the customer: downtime, and what we did about every failure to deliver. Then sales: close rates, next steps, and why we were winning and losing. Then product timelines. And beside every number, the reason it moved.

At Infrascale, the data protection company where I was CEO, the dashboard was beautiful. It was a picture of the business at one moment, and it could not answer the only question that mattered: so what? When the product performed in a crisis, the company recovered the customer and moved on. Nothing asked who calls this customer and what the account team does now. The people who had chosen the system had made the right call, and nobody told them so. The dashboard missed competitors too: no one tracked Zerto or Veeam alongside our own numbers, and a good internal trend was not evidence that we were keeping pace. Once the dashboard included market share, it showed a market significantly larger than the company had capacity to scale into, and that changed where we invested in sales and marketing.

Data can support almost any story if you do not begin with a question. First identify the decisions the business needs to make. Then collect the evidence that could change the answer, including the evidence you do not like. Otherwise, the data becomes a defense of the position you already hold. A metric earns its place only when it reveals an outcome, predicts a result, or triggers a management action by a named owner. For each one, agree what change would prompt a response and when the evidence will be reviewed. Jeff Bezos described one-way and two-way doors in his 2015 shareholder letter: choices that are difficult to reverse deserve deeper analysis, and reversible ones can use a lighter process. My rule follows from it: classify the door first, then size the evidence. When you follow a number to its cause, five whys is a reminder to go deeper, not a quota: stop when you have a cause you can act on and a way to test whether the correction works.

The Decision Brief is written before the review. The Metric-to-Decision Map audits the dashboard backward: what decision does this number inform, and who may respond. Customer Recovery Evidence is three questions asked after the product performs under pressure. The book's worked example returns to the illustrative support record from Module 5: median resolution time against the 60-minute starting point and the 30-minute target due on day 90, repeat lockouts and reopened tickets, and the backup's demonstrated coverage, each with a decision attached. A failed identity check never waits for the weekly dashboard.

If the review is yours, begin with the decision, its owner, the evidence needed for its risk, and the date it closes, and keep only the measures that help make that choice. If you assemble the pack, write beside each number the decision it would change, and move the rest to the back. If you report upward, lead with the decision your number informs, and ask what further evidence would settle it and by when. If a decision waits until you are satisfied, say what would satisfy you and by when. Otherwise the company learns to wait for one person instead of learning to decide.

A dashboard that answers no decision is expensive reassurance.