Giving leaders one screen they actually look at

The situation

The client was a wholesale distributor with a dozen warehouses and a leadership team that felt simultaneously over-informed and under-informed. They received roughly 60 reports per week — daily operational, weekly management, monthly financial, quarterly strategic. The common complaint: "We have all this data and no idea what matters."

The weekly ops meeting took 90 minutes and was dominated by someone walking the room through spreadsheets. Decisions rarely came out of it. Engagement was low.

What we found in discovery

Two weeks of conversations with the CEO, COO, CFO, and warehouse leads clarified three things:

• Most reports were generated, but not read. A quick survey found that of 60 recurring reports, fewer than 10 were consistently read by their intended audience.
• The same numbers appeared on three reports in three formats. Revenue by region was on the daily op report, the weekly management summary, and the monthly financial — all computed slightly differently.
• The leaders agreed on fewer metrics than expected. When asked individually "what's the one number you look at?" we got overlapping but not identical lists. The act of forcing alignment was half the engagement.

The approach

We ran the project in three phases:

Phase 1 — The 5 questions workshop (week 1). We put the leadership team in a room and walked every proposed KPI through the five-question filter: does it change a decision, who owns it, is it leading or lagging, is it auto-measurable, is it hard to game? 42 candidate metrics collapsed to 7 weekly KPIs and 5 quarterly ones.

Phase 2 — Data plumbing (weeks 2–4). For each KPI, we traced the underlying data — where it lives, how current it is, how it's calculated — and fixed the gaps. Three KPIs required new instrumentation. Two required agreement on a single canonical definition. One was retired when the team realized they'd never really agreed what it measured.

Phase 3 — The dashboard (weeks 5–6). A single executive dashboard, live-updating, with the 7 weekly KPIs prominent and the 5 quarterly ones secondary. Every metric has a named owner, a trend line, and a threshold that triggers a flag if crossed.

We explicitly didn't build a BI platform. The client had three already. We built one screen on top of the warehouse that existed.

The outcomes

Three months after go-live:

• 7 weekly KPIs on the wall, all auto-generated, all with named owners.
• Weekly ops meeting dropped from 90 minutes to 55 minutes. More importantly, the meeting time is now spent on decisions ("the west region is flagged — what are we doing?") instead of walkthroughs.
• Data latency dropped from "last week's numbers on Wednesday" to live. Leaders check the dashboard on their phones.
• Report sprawl: 60 → 18. The 42 retired reports had a combined preparation cost of ~15 hours per week that went back to the teams that were producing them.
• Decision velocity: leadership now makes operational calls midweek based on the dashboard instead of waiting for the Friday debrief.

What didn't happen

• No new BI platform. Dashboard sits on their existing warehouse. Budget impact near zero beyond the engagement itself.
• No data team expansion. The owners of the KPIs are also the owners of the underlying definitions and plumbing — responsibility is distributed, not centralized.
• No retainer. The dashboard is owned, documented, and extensible by the client's internal team.

Why it worked

Two disciplines.

First, we spent a full week on agreement before writing a line of SQL. The hardest part of KPI work isn't the engineering — it's getting a leadership team to agree on what matters. Skipping that phase produces beautiful dashboards nobody trusts.

Second, we built the fewest metrics we could justify. "What else should be on here?" is easy to answer. "What should come off?" is the question worth asking. Seven is enough. Seventy is noise.

The dashboard isn't the product. The weekly conversation is. We built the smallest artifact that reliably produces the right conversation.