The right metrics for a business intelligence dashboard are the few numbers that change a decision someone actually makes each week. Start from the decisions, not the data: for each, pick one clear metric, define it precisely, name an owner and show a comparison. If a number would not change anyone's behaviour, leave it off.
Key takeaways
- Begin with decisions and questions, then choose the metrics that answer them.
- Fewer metrics, clearly defined, beat a crowded screen of charts.
- Every metric needs a written definition, an owner, a data source and a comparison point.
- Pair leading indicators (early signals) with lagging results (outcomes) so you can act in time.
- Review dashboards regularly and remove metrics nobody uses.
Why most dashboards fail
Many dashboards are built the other way round. Someone connects a data source, sees what is available and drops every number onto a screen. The result looks impressive and answers nothing.
Typical symptoms:
- Twenty charts, and nobody can say which three matter.
- Two reports show different values for "revenue" or "active customers".
- People export to a spreadsheet and rebuild the numbers themselves, because they do not trust the dashboard.
- Metrics measure activity (emails sent, tickets opened) instead of results (deals closed, issues resolved).
A BI dashboard earns its place only if it changes what people do on Monday morning.
Start with decisions, not data
Sit down with the person who will use the dashboard and ask:
- What decisions do you make every week or month?
- What do you wish you knew earlier?
- What would make you act differently, such as reorder stock, call a customer, change a budget or hire someone?
- How do you check this today, and what is painful about it?
Write each answer as a question: "Are we on track to hit this month's sales target?" or "Which products will run out in the next two weeks?" Each question points to one or two metrics.
Split by audience
A founder, a sales manager and a warehouse lead need different views. One dashboard for everyone usually serves no one. Create a small number of focused views, each with a clear audience.
Pick metrics that are useful
Leading and lagging
Lagging metrics tell you what already happened, such as monthly revenue. Leading metrics move earlier, such as the number of qualified leads entering the pipeline this week. You need both: lagging metrics confirm results, leading metrics give you time to react.
Rates and ratios over raw totals
A total can rise simply because the business grew. A rate, such as the share of leads that become customers, shows whether you are improving. Use totals for scale and ratios for performance.
Avoid vanity metrics
Page views, followers and downloads can look good while the business stands still. Ask: if this number doubled, would we make more money, serve customers better or save time? If not, it is probably decoration.
Match metrics to the business model
- E-commerce: conversion from visit to order, average order value, repeat purchase rate, return rate.
- Services and agencies: pipeline value, win rate, utilisation of staff time, invoices overdue.
- Subscription software: new and cancelled subscriptions, churn, expansion revenue.
- Operations and supply chain: stock cover, order fulfilment time, supplier delays.
- Support: first response time, resolution time, backlog, customer satisfaction.
These are examples to adapt, not a universal list.
Define every metric in writing
Most disagreements about numbers are really disagreements about definitions. For each metric, record:
- Name and plain-English definition. "Active customer: placed at least one paid order in the last 90 days."
- Formula. Exactly how it is calculated, including what is excluded, such as refunds or test orders.
- Source. Which system is the single source of truth.
- Time window. Daily, weekly, monthly, rolling or calendar.
- Owner. The person who explains the number and fixes it if it looks wrong.
- Target or comparison. Last period, same period last year, budget or target.
A short metric glossary linked from the dashboard saves countless arguments.
Design for fast reading
- Put the most important metric at the top left and give it a clear label.
- Show a comparison next to every number, since a figure alone has no meaning.
- Use simple chart types: lines for trends, bars for comparisons, and plain numbers for headline figures.
- Use colour sparingly, for example to flag values outside the target range.
- Show when the data was last refreshed, so nobody acts on stale numbers.
- Let people drill down from a summary to the underlying records.
Data quality and access
A dashboard is only as trustworthy as the data behind it. Before building, check that the source systems capture the information consistently, that duplicates and test records are handled, and that data from different tools can be matched, for example by customer ID. If your data is scattered across spreadsheets and apps, a data pipeline that collects and cleans it in one place usually comes before the dashboard.
Also think about who may see what. Finance numbers or individual performance data may need restricted access. Look for row-level or role-based permissions so a branch manager sees their branch, not every branch.
A short worked example
Imagine a regional retailer with a small online store and three physical outlets. The owner asks for a dashboard and the first draft contains thirty charts.
The team goes back to decisions. The owner says: "Each Monday I decide what to reorder, which outlet needs help and whether to spend on ads." That reduces the dashboard to three questions. Which items are close to running out? Which outlet is behind its weekly sales target? Is advertising spend producing orders at an acceptable cost?
For each question, one headline metric is chosen: stock cover in days for the top items, weekly sales against target by outlet, and cost per order from advertising. Each is defined in writing, compared with the previous week, and given an owner. The thirty charts become one page. Detailed charts stay one click away for analysts. Three months later the owner reviews which tiles are used and removes two that nobody opens.
Tools and next layers
You can build dashboards in many tools. If you want something ready to use, Grocito's Analytics & BI Dashboard lists drag-and-drop dashboards, connectors for databases, Sheets and APIs, KPI alerts, scheduled PDF and email reports, row-level permissions and forecasting and trend detection. For custom reporting, pipelines and warehousing, see data analytics. Alerts and scheduled reports are worth using once your metrics are stable, because they bring the number to people instead of waiting for them to look.
FAQ
How many metrics should a business dashboard have?
There is no fixed number, but fewer is better. A focused dashboard for one audience often works with a handful of headline metrics and drill-downs behind them. If people cannot tell which numbers matter most, there are too many.
What is the difference between a KPI and a metric?
A metric is any measurement. A KPI, or key performance indicator, is a metric tied to a specific business goal and usually has a target and an owner. Every KPI is a metric, but most metrics are not KPIs.
How often should we review our dashboards?
Check the numbers at the rhythm of the decision: daily for operations, weekly for sales and finance reviews, monthly for strategy. Review the dashboard itself every few months to remove unused metrics and add new questions.
Why do my reports show different numbers for the same metric?
Usually the definitions differ: different date ranges, different exclusions or different source systems. Write a single definition, pick one source of truth and calculate the metric in one place.
Next steps
Take a blank page and list the five decisions you or your team make most often. Write the question behind each one, then choose one metric per question. If you would like help connecting your data and turning those questions into a working dashboard, contact us and we can review your sources together.
