Dashboards that answer questions nobody asked
Many dashboards are a wall of numbers with no decision attached. A useful one starts from the decision and works backwards to the metric, and it usually ends up smaller and calmer than expected.
Key takeaways
- A metric is only useful when a plausible decision depends on it.
- Three numbers describe most content businesses: qualified sessions, capture rate and value per subscriber.
- Event definitions written once, next to the event, prevent most reporting disagreements.
A useful dashboard starts with a decision and works backwards to the number that would change the answer. Many dashboards do the opposite: they list everything a tool can measure and leave the reader to invent a question. The result is usually impressive and rarely consulted.
Start with the decision
Writing the decision first, as a sentence, makes the metric selection straightforward: “Should budget stay on the comparison page or move to email?” Then ask which number would change that answer. Usually there is one, occasionally two. Everything else is context that can be looked up when the question changes.
The three-metric test
If a screen cannot be summarised in three numbers, it is closer to a data dump than a dashboard. For a content-led site, the three that describe the machine are acquisition, capture and value, and they fit on one line.
- Qualified sessionsTraffic from people who could plausibly buy, rather than all traffic. A post about a competitor’s outage is not a qualified session, even when it is popular.
- Capture rateThe share of qualified sessions that end in an email address or an account. This is the number most content sites ignore.
- Value per subscriberWhat an average subscriber is worth over ninety days. It turns audience growth into a business number.
A number without a decision attached is entertainment.
Making the boring parts automatic
Definitions drift. Two people reading “signup” will usually count two different events within a month unless the definition lives in one place, next to the event itself. A short document with the exact event name, the trigger and one example removes most of that ambiguity, and it is a pleasant thing to write once.
Reviewing the dashboard itself
A monthly pass that deletes anything unused for four weeks keeps the dashboard short. A dashboard people trust is one short enough to read, and removing unused metrics is the fastest way to keep it that way. Nothing is lost by doing it, because anything genuinely needed finds its way back.
Frequently asked questions
- How many metrics should a dashboard have?
- Three at the top level is a comfortable starting point. Acquisition, capture and value cover a content-led business. Everything else can sit one click deeper and be consulted when the question changes.
- Why is bounce rate often left out?
- Because few decisions follow from it. A doubling prompts a look at the page regardless, and a halving rarely changes a plan. Metrics that do not change behaviour take attention without returning much.
- How can definitions be kept consistent between people?
- Keep one short document with the exact event name, the trigger and an example. It takes about twenty minutes to write and prevents a whole category of recurring arguments.
Start here
Read a guide, or open a tool and try it.
Every guide is tested on a live project before it is published, and every tool runs in your browser without an account.
Keep reading
The Eisenhower matrix: a kind way to sort a long list
Two questions, four boxes, and a list that finally feels manageable. The matrix is a gentle triage tool, and it works best when it is used with a little patience.
Read articleIncrementality: is paid search actually adding revenue?
Attribution shows who received credit. Incrementality asks what would have happened without the activity. The difference matters most where SEO and paid search compete for the same click, and it can be tested without a data team.
Read article