This isn’t about slogans or tools. Start with business goals, then hardwire a few habits so people reach for data by default. Use maturity as the primary lens — the number of people who regularly work with data — while noting company size as a rough proxy.

Start with business goals

Write down the top three company priorities with an accountable owner and the metric each will move. Trace every analytics task to one of those priorities. Define how impact will be measured before work starts, whether that’s a test, a counterfactual, or a clean before/after with controls. When leaders do this in public, “use data” stops being a slogan and becomes muscle memory.

Maturity Stage 1: Foundation

(≈≤10 regular data users; often ≤50 people)

Make data usage visible every week. Run a 15-minute metrics review that shows three core numbers, explains shifts, and takes “why” questions until the logic holds. End all-hands with one chart, one insight, and one decision it changed.

Create one source of truth and remove access friction. Schedule a report by Mail, Slack or Teams and open it in meetings. Publish a two-page KPI glossary with names, formulas, and sources so definitions don’t drift.

Put quality first from day one. Name owners for the core tables. Keep a public defect log. Fix issues in the open so trust grows with each correction.

Lower the cost of asking questions. Add an AI data agent in chat so people can ask “What happened to signups yesterday?” and get a consistent answer without a ticket. Dot works well here. Feed the agent real business context — your KPI definitions, allowed filters, time windows, segment labels, revenue rules, and synonyms — so it can translate between business and data. The value comes from the context you provide.

Enable the highest-impact users first. Make sure the CEO and senior leaders are regular users of the live dashboards and the AI agent. If the top decisions run on the same numbers everyone sees, quality and adoption rise quickly.

Maturity Stage 2: Expansion

(≈10–50 regular data users; often 100–500 people)

Require a “data backing” section in every plan. A campaign, a feature, or a hiring request should state the current baseline, the expected impact, and the decision rule. Once this is expected, preparation and outcomes improve without extra process.

Turn BI into a teaching function. Hold open office hours where people leave with a saved query or reusable dashboard and one new skill. Keep a lightweight, quarterly hygiene pass that archives dead reports, fixes broken filters, and merges duplicates. Set default expiry on new reports so owners re-affirm relevance.

Establish a simple experimentation rhythm. Define when to test, minimal guardrails for sample size and duration, and a weekly review where product and marketing walk through live tests with a decision date. Tooling can mature later; the cadence matters now.

Let the AI agent handle translation at scale. Continue feeding Dot with business context — metric definitions, table lineage, canonical joins, and approved calculations — so chat answers match your source of truth.

Prioritize executive enablement. Give the C-suite a clean, relevant daily report and the same agent access. Ask for assumptions, confidence, and links to the live source in staff meetings. Optimize for the decisions that move the company.

Explain where QBRs fit. A Quarterly Business Review is the standing forum where each team shows progress against the goals defined up front, the analysis or experiment that drove a decision, the result, and what changes next. Link each claim to the live source. Use the same QBR rhythm to retire vanity metrics and reports.

Maturity Stage 3: Scale

(≈50+ regular data users across multiple teams; often 500–2,000+ people)

Eliminate “two sets of numbers.” Have executives open meetings on the same live dashboards everyone sees. If a figure looks wrong, fix it in the source so the correction benefits all.

Standardize the spine. Maintain a single KPI catalog, a governed semantic layer or metric store, and a searchable catalog with owners and freshness. Assign domain ownership and simple data contracts for high-value tables so upstream changes are announced and monitored.

Keep governance enabling. Default to broad read access on non-sensitive data with role-based permissions and basic anonymization where needed. Track quality with named owners and a public defect backlog. The path to the right answer should get shorter each quarter.

Run a role-based training ladder. Front-line managers learn to interpret trends and ask better questions. Analysts deepen methods and storytelling. Power users get extra tooling. Keep training short, regular, and built on your real datasets.

Use incentives sparingly but clearly. Recognize teams that contribute reusable analyses or improve shared datasets. If you experiment with storytelling or light gamification, tie it to business outcomes, not vanity usage.

How you’ll know it’s working

More people outside BI use BI weekly. Time to first answer for common questions drops from days to minutes. Proposals routinely include a real data backing section. Core tables have named owners, freshness checks, and fewer defects quarter over quarter. QBRs show targets set before work began, the analysis that drove the decision, and measurable outcomes. Most importantly, you can point to decisions each quarter where analysis changed the call and moved a KPI.

If you lead analytics, start here next week

Publish a crisp KPI glossary and schedule one relevant report by Mail, in Slack or Teams to the C-suite. Announce that every proposal includes a data backing section and link a one-page analysis template. Open BI office hours and name one business-context owner per function to feed your AI agent with context. Make sure the CEO uses data publicly.