Runs on PeoplePlaybooks / Measures · 3 min read
Using data to make decisions
Data doesn't make decisions. It changes your mind, if you decided in advance what would.
The problem
“We have dashboards everywhere, and decisions still come down to opinion.”
In the model
In 30 seconds
- Start with the decision, not the dashboard.
- Before you look, write down what result would change your mind.
- Pair every number with a story, and every story with a number. Never use data to surprise someone.
The situation
Your team has more dashboards than decisions. Meetings open with charts, people pick the ones that support what they already thought, and the loudest interpretation wins. The data is everywhere and changes nothing.
What's really going on
Most data is gathered because it can be, not because a decision needs it. Then people look at it with their conclusion already formed. That isn't analysis, it's decoration. And when numbers become targets, people start managing the number instead of the work.
The goal isn't more data. It's fewer, better questions, asked before anyone looks.
What to do
- Name the decision. "Should we hire another nurse or extend clinic hours?" not "Let's look at the numbers."
- Write down what would change your mind, before looking. "If more than 30% of calls are about appointments, we change the booking system." This is the step that turns data from decoration into evidence.
- Use the fewest numbers that answer the question. One good measure beats a wall of charts.
- Read trends, not snapshots. One week is noise. Eight weeks is a signal. Show ranges, not single points.
- Pair leading and lagging measures. Lagging measures tell you what happened. Leading measures give you time to act. You need both (see below).
- Ask the people in the data. Numbers tell you what. The people doing the work tell you why. Never skip the second conversation.
- Watch for gaming. When a measure becomes a target, it stops being a good measure (Goodhart's law). If a number improves suspiciously fast, go and look at the work.
Leading and lagging, by field
| Field | Leading: act on it now | Lagging: the result |
|---|---|---|
| Software | Time from idea to release; size of work in progress | Customer churn; revenue |
| Healthcare | Appointment wait times; staff sickness | Patient outcomes; complaints |
| Accountancy | Work-in-progress days; unbilled time | Write-offs; client retention |
| Manufacturing | Near misses; unplanned stoppages | Injuries; scrap and returns |
| Any team | Whether 1:1s are happening; overtime hours | Resignations; engagement scores |
Three rules for data about people
- Share it with the people it describes first. Data shown about someone in a meeting before they've seen it destroys trust permanently.
- Measure teams, not individuals, unless it's for that person's own development and they've agreed.
- Treat it as the start of a conversation. "Your numbers are down" ends one. "I noticed this, what's going on?" starts one.
How you'll know it's working
- Purpose: every chart you keep has a named decision behind it.
- Discipline: decisions are made against criteria agreed before anyone looked.
- Care check: nobody sees data about themselves for the first time in a meeting.
In the frameworks
- DORA: The four key metrics measure delivery speed and stability together, so neither can be gamed alone.
- SAFe: Flow metrics (velocity, time, load, efficiency, distribution, predictability) show the whole system.
- Kanban: Lead time and cycle time, measured from the board, not guessed.
- OKRs: Goodhart's law applies: key results used as targets get sandbagged.
In other industries
- Healthcare A practice adds a dashboard of DNA rates. The useful question isn't "who has the most?" but "what would make us change how we remind patients?"
- Accountancy Utilisation targets drive juniors to log hours, not value. Pair utilisation with write-offs and the picture changes.
- Manufacturing Fewer reported near misses can mean a safer floor, or a floor that stopped reporting. Ask before celebrating.
- Small business You don't need a BI tool. One weekly number on a whiteboard, chosen for a decision, beats any dashboard.
What to say
To open a meeting where data will be shown:
"Before we look at the numbers: what decision are we making, and what result would make us choose differently? Let's write that down first, then look."
Try this week
Pick one dashboard your team owns. For each chart, write the decision it informs. Delete the charts with no decision.
Think about it
When did a number last change your mind, rather than confirm what you already thought?