Business Intelligence: Produce Decisions, Not Dashboards

Most BI projects succeed technically and fail in practice. The dashboards get built, the charts are colourful, data flows — and six months later nobody opens them. The reason is usually that the dashboards were designed to display available data rather than to answer questions.

Start From the Decision

The right order is: which decision will you make, what question does that decision require answering, which metric answers it, and from what data is that metric calculated. Working the other way — "we have this data, what shall we show" — produces dashboards nobody uses.

A concrete example. Decision: which product group deserves inventory investment. Question: which group turns quickly and holds margin. Metric: inventory turnover and gross margin. Data: sales movements and stock cost. Once that chain exists, the dashboard may be a single table — but it gets opened every week.

Choosing Metrics

• It connects to an action. If you know what you will do when the number worsens, it is a good metric. If you do not, it is just information.
  
• It has one owner. A metric without an owner does not get followed.
  
• It is hard to game. If the team can improve the metric by damaging the business, the metric is wrong. Cutting call handling time by rushing customers is the classic case.
  
• It is comparative. A number alone means nothing; it acquires meaning against a prior period, a target or a peer unit.
  

Data Quality Comes First

• Shared definitions: What does "sales" mean — orders, dispatches or invoices? Are returns deducted? If departments use different definitions, dashboards will contradict each other.
  
• Completeness: If category, region or sales-rep fields are left blank, segmented reporting is impossible.
  
• Timing: The lag between when a transaction happens and when it is recorded distorts period comparisons.
  
• One source: The same metric producing two different values in two systems destroys trust in the entire report.
  

Dashboard Design

• It fits on one screen. A dashboard that requires scrolling is a dashboard whose lower half is never seen.
  
• The most important metric sits top left. That is where the eye goes.
  
• Colour is reserved for exceptions. When everything is coloured, nothing stands out.
  
• Chart type matches the data. Lines for time series, bars for category comparison. Pie charts stop being readable beyond two slices.
  
• Freshness is visible. A dashboard without an "as of" timestamp does not inspire confidence.
  

Turning Reports Into Decisions

• A short weekly review covering only the metrics that moved.
  
• An owner and an action defined for each deviation.
  
• Last week's actions reviewed at the start of the next session.
  
[*]Metrics that have driven no decision in three months are removed from the dashboard.

Automated Alerts

The most efficient reporting is reporting that does not wait to be looked at. Threshold-based alerts — critical stock levels, overdue receivables, sales below target — should reach the relevant person on their own. Waiting for someone to open a dashboard is waiting for the problem to grow.

In Short

BI is not a visualisation project; it is a decision-discipline project. Work backwards from the decision to the data, keep the metric count low, and attach an owner and an action to each one. Removing unused metrics matters as much as adding new ones.

We can work out which metrics you should be tracking with Mekjoy Business Intelligence & Reporting.