Branch performance dashboards that regions and head office read the same way

Branch reviews in most retail banks share a pattern. Head office has one number for new accounts, each region has its own, and the first twenty minutes of the meeting go on working out why they differ. One region counts joint accounts twice, another excludes accounts opened online but verified in branch, and a third uses last month’s staffing figures.

The dashboard above is built to remove that argument. It is an Agent Serve dashboard on shared definitions, so the North region and the head of distribution are looking at the same metric, filtered differently, rather than at two versions of it.

Start with the definitions, not the charts

The semantic model describes the business once: branches, regions, products, customers, appointments and staff, mapped onto the tables in your warehouse through data sources, with the joins between them and the synonyms people actually use ("site" for branch, "advisers" for FTE).

On top of that sit governed metrics. Each one has a single definition that every panel, report and AI answer reads:

• New accounts per FTE is a ratio metric: accounts opened, counted once per customer and product, divided by average full-time staff in the period.
• Average appointment wait is the mean number of days between booking and the appointment.
• Complaints per 1,000 customers is a ratio over the branch’s active customer base.

Metrics do not yet carry time-intelligence such as month-to-date or year-on-year. Period comparisons happen on the dashboard instead, which is where the "vs previous period" notes and the last-year line in the chart come from.

The KPI row and the trend line

The four tiles answer the first question in any branch review: is the network doing better or worse than last time? Each tile uses the dashboard’s comparison setting, set to the previous period, so the change note is calculated, not typed.

The trend chart compares this year’s weekly new accounts per FTE with the same weeks last year, using the same-period-last-year comparison. Annotations mark events that explain a step change, such as a branch refurbishment or a product launch, so nobody has to remember them in the meeting.

The map, the pivot and drill-down

A map panel plots every branch with its watchlist status. Regional managers usually start here because geography explains a lot: two neighbouring branches on Watch often share a staffing problem.

A pivot panel shows region by product, so the head of distribution can see whether a dip in new accounts is concentrated in one product line. Drill-down goes from region to branch, and date drill-down goes from year to quarter, month and day. Filters at the top (period, region, product) apply to every panel at once.

The watchlist rule

The table at the bottom of the dashboard is driven by one rule. Rules in Agent Serve are plain conditions over row fields that return a flag, a label or a score. This one returns a label:

• Review when new accounts per FTE is below 10 and appointment wait is above 5 days.
• Watch when either one of those conditions holds, or complaints per 1,000 customers is above 2.5.
• On track otherwise.

The thresholds above are illustrative. The point is that they are written down, visible on the rule, and the same for every region. When the network team changes a threshold, the change applies everywhere at once.

The monitors

Two monitors watch the dashboard so people do not have to:

• A rule monitor fires when three or more branches carry the Review label, at warning severity.
• A KPI alert fires when the average appointment wait across the network goes above 4 days.

Alerts appear in the app and by email to the head of distribution. If your team wants them in Slack or Microsoft Teams, they can go to a channel through a workspace webhook. An alert resolves on its own once the condition clears, so the alert list reflects the current state of the network.

Reports each region can trust