SEO reporting dashboard: what to show and what to leave out

A dashboard and a report answer different questions. A report, sent once a month, says what happened and why. A dashboard, opened whenever someone wants, says whether anything needs attention now. Most bad dashboards are reports that have lost their words: every chart the monthly report has, with nothing to say which of them matter this week.

So a dashboard is judged by one test: can its reader tell, in under a minute, whether to act and on what? Every figure on the first screen either helps answer that question or belongs somewhere else. A figure that needs a paragraph of explanation belongs in the report, whose sections are set out in Monthly SEO report: what to include. A figure that only a specialist uses belongs on a detail page that specialist opens.

This is not an argument for fewer numbers overall. It is an argument for putting each number where its reader is. The owner of a business and the person fixing its pages need different screens, and a dashboard that tries to be both serves neither.

Show figures that are tied to how the business earns money, that come from the source closest to the event they count, and that keep their meaning from one month to the next. Six figures cover most sites.

Revenue keeps its label on the dashboard as it does in the report: one of the four kinds, or GA4-reported when only analytics stands behind it, which is none of the four. The labels, why they are never added together, and why branded searches are kept apart from non-branded ones are set out in the guide to SEO reporting in revenue. On a dashboard, the label is a word next to the number, not a footnote.

Each figure appears with its comparison, this period against the one before and against the same period last year, because a number alone cannot say whether to act. Where a figure rests on few events, such as three enquiries in a week, the dashboard says so instead of drawing a trend through it.

Most of what clutters an SEO dashboard is not wrong. It is true, easy to collect and almost impossible to act on. Leave these off the first screen, and some of them off the dashboard altogether.

• Average position shown alone. Search Console reports the topmost position of the site in each result, averaged across all the queries in which it appeared^[1]. A site that starts to appear on page three for many new searches sees its average position number rise, which reads as a worse ranking, while its traffic rises. Drop the lone figure; if position stays on the dashboard, show it next to the ranking spread, how many searches sit in the top 3 and the top 10, and beside the position of the few searches that bring buyers, by name.
• A count of ranking keywords. The total moves with the tool’s keyword database and with every page added, and says nothing about enquiries. Keep it, if at all, as the ranking spread of a frozen keyword set, on a detail page.
• Authority scores from third-party tools. Google states that it does not evaluate or endorse third-party SEO tools, and that these tools do not have access to its internal ranking data^[2]. A tool’s score for a domain is that tool’s estimate, not a figure Google uses.
• Lab speed scores. PageSpeed Insights gives lab data, collected in a controlled environment and useful for debugging, and field data from real users, which captures their real experience^[3]. The Core Web Vitals report in Search Console is based on field data^[4], so that is the figure for a dashboard; the lab score belongs with whoever fixes the page.
• Bounce rate read alone. In Google Analytics, bounce rate is the share of sessions that were not engaged, and a session is engaged if it lasts longer than 10 seconds, has a key event or has two or more page or screen views^[5]. A page that answers a question in eight seconds bounces by that rule. Read the rate beside the key events of the same page, or not at all.
• Impressions without the clicks they produced. Impressions rise whenever a site appears for more searches, including ones it will never win.
• Counts of tasks done. Hours, tickets and articles belong in the report’s list of work done, where they can be matched with what moved.

A dashboard usually has one date control, and that is the source of many wrong readings. Figures do not all move at the same speed, and some are already averages over a period of their own. Choose a period for each figure, and write it next to the figure.

Field data is the clearest case. PageSpeed Insights reports real users’ experience over the previous 28-day collection period^[3], so a fix made yesterday cannot show in it today, whatever the date control says. Write “last 28 days, real users” next to the figure, and nobody will expect it to move the morning after a release.

A dashboard is useful only if its reader knows when a movement matters. Agree, for each figure on the first screen, what change calls for action, and write it on the dashboard or in its definitions. Without that, every dip is a question and every rise a celebration.

• Set the line from the figure’s own history: a movement larger than the usual change from one period to the next over the past year is worth a look; one inside it is not.
• Treat a figure that falls to nothing as a fault first; the guide to automated reports covers the checks that catch it.
• Read a fall in pages indexed against the site’s own changes: pages removed on purpose are a cleanup, pages that dropped out without a change are a problem.
• Watch a Core Web Vitals group that moves to Poor, because it affects every page in the group at once.
• Leave small counts out of the rule. A change from three enquiries to two has no threshold that means anything.

When a figure crosses its line, the dashboard has done its job: someone looks. The explanation, the cause and what is being done belong in a short message the same day and in the next report, not in a note on the chart.

Give each reader the screen that answers their question, built from the same data and the same definitions. Three screens serve most teams.

All three screens read from one definitions page: the key events, the brand terms, the page groups and the keyword set. If the owner’s screen and the practitioner’s screen disagree about a number, one of them has stopped using the definitions. The fields of that page are set out in SEO report template.

Screens for AI search follow the same rule. Mentions and citations are counted for a fixed set of prompts, and a change in the set is dated; how to build that set, and what its numbers can and cannot show, is in the guide to measuring results in AI search.

• Name the source and the period on every chart, so that a reader who opens the source finds the same number.
• Say how each total is counted, and keep one count for each figure; the template’s definitions page sets it, as the guide to the SEO report template shows.
• Label every revenue figure by kind, in words, next to the number.
• Mark a figure that rests on a small count, and draw no trend through it.
• Never use colour as the only cue. A red arrow also says “down”, and a green one “up”.
• Show the date the data was last refreshed, because a dashboard opened the morning after a fix may not show it yet.

Read the sample the way its owner would. Clicks, key events and revenue are up on last year, and nothing there needs action today. Pages indexed fell by 60, which may be a cleanup or may be pages dropping out of search; that is the one figure to check before the report goes out. The AI mentions are marked as a small count, so a change of one or two prompts is not read as a trend.

Leaving a figure off the first screen is not hiding it. The full keyword list, every crawl error and the export behind each chart belong on a detail page or in a linked sheet, where the practitioner who needs them finds them and the owner who does not is never asked to read them.

When someone asks for a figure that was left out, ask what decision it would inform. If there is one, the figure earns a place, on the screen of the reader who makes that decision. If there is none, put it on the detail page, where it is one click away and costs the first screen nothing. Either way, write the answer down, so that the same request is not argued again next quarter.

What to automate in the pipeline behind a dashboard, the limits of each data source and the checks that catch a broken one are in Automated SEO reports: what to automate and what not.

1. 1For each figure on the first screen, name the decision it informs. If there is none, move it to a detail page.
2. 2Check that each figure has its source, its period and its comparison.
3. 3Check that branded and non-branded searches are split.
4. 4Check that every revenue figure carries its label.
5. 5Open one figure in its source and confirm that the number matches.
6. 6Ask each reader what they looked at last time. What nobody looked at goes to the detail page.

A dashboard that passes these checks is smaller than the one it replaces, and it is opened more often.