Email Reporting Dashboards: Building Actionable Views
Three dashboards for three audiences, and the rule that keeps a report from becoming a wall of numbers nobody acts on.
On this page
The short version
- Three audiences need three different views. One dashboard serving all of them serves none, and it is always the operator's version that gets built.
- Every panel needs a decision attached. A panel that would change nothing at any value is costing you the attention of the panels that would.
- State what a number does not include, on the dashboard, not in a footnote nobody opens.
Most email dashboards are built by the person running the programme, for the person running the programme, and are then shown to everyone else. That is why they contain bounce breakdowns and why nobody outside the team reads them.
The fix is not a better dashboard. It is three, each answering a different question, sharing one set of underlying numbers.
Three views, three questions
| View | Question | Contents | Cadence |
|---|---|---|---|
| Operator | What do I do next? | Per-send rates, bounces, complaints by segment, placement by domain | Every send |
| Manager | Is this working? | Three headline numbers, trend, flows versus campaigns | Monthly |
| Executive | Is this worth it? | Attributed revenue, cost, list health as one figure | Quarterly |
The manager view is the one that usually does not exist. It gets replaced by the operator view with some panels hidden, which is why it reads as technical.
The panel test
For each panel, name the decision it supports and the value that would trigger it. "Complaint rate — if this crosses 0.1% we stop and diagnose" is a panel. "Opens by device" is usually not, unless someone can say what they would do differently on any value.
Applying this honestly removes about half of most dashboards, and the remaining half becomes readable. Removing panels is the work; adding them is what happens by default.
Comparisons, not absolutes
A number on its own is uninterpretable to anyone who does not work with it daily. Every panel should carry a comparison: against the previous period, against the trailing average, or against a stated target.
The trailing average is usually the most honest of the three, because period-on-period comparison is noisy on any programme sending less than weekly and a target invites the number to be managed rather than reported.
The sentence at the top
Every view should open with a written line saying what happened and what is being done about it. This is the part people read, and it is the part that gets left out because it cannot be automated.
It also disciplines the dashboard: if you cannot summarise the month in a sentence from the panels present, the panels are wrong.
What to do when a number is unavailable
Some panels cannot be filled — revenue is not attributable yet, placement data is not accessible, the commercial system does not join to the email address. The temptation is to leave the panel out and hope nobody notices the gap.
Show it, empty, with a line saying why. A dashboard with a stated blank is honest and generates the conversation that fills it; a dashboard with the blank quietly removed looks complete and stays incomplete for years.
This is the same reasoning as labelling a caveat on a number. What a report does not know is part of what it is reporting.
Dashboard check
- Three views exist, or you have decided which audiences you are not serving
- Every panel has a named decision and a trigger value
- Every number carries a comparison
- Caveats sit on the numbers, not in a footnote
- Opens are labelled as including automated fetches
- The attribution model and window are stated
- Each view opens with a written sentence
Frequently asked questions
Should the dashboard be automated?
The numbers, yes. The sentence at the top, no — that is the judgement, and automating it produces a report nobody reads because there is nothing in it a person thought about.
How far back should trends go?
Thirteen months, so the current month has the same month last year beside it. Anything shorter hides seasonality and invites explanations for variation that is annual.
What if the executive view is one number?
That is often correct. A quarterly view containing attributed revenue against cost, with one sentence, is a better artefact than a page of charts that gets skimmed.