An SEO report can be accurate and still be useless. A dashboard says impressions rose, average position fell, and organic enquiries stayed flat. The client asks whether the work is paying off. The dashboard has no answer because it mixes three different questions: what changed in search, what the agency delivered, and what the business gained.
A better report keeps those questions connected without pretending they are interchangeable. It explains a movement, shows the evidence behind that explanation, and ends with a decision someone can act on. This guide builds that report around an explicitly hypothetical service business. The arithmetic is reproducible; the numbers are teaching examples, not Serse or client results.
Start with a reporting contract
Before choosing charts, write down what the business wants the report to help decide. For a local service business, that might be which service page deserves a rewrite. For an agency selling retainers, it might be whether a new industry guide attracts qualified enquiries. These require different evidence even when both teams call the work “SEO”.
Use a short scope note at the beginning of every report. Include the Search Console property, search type, periods, countries, devices, the landing pages being evaluated, and the date of the latest complete observation. Keep the note even if it feels repetitive: it prevents next month's report quietly becoming a different analysis.
Choose comparable periods
Suppose you compare two consecutive 28-day periods. Matching their length keeps the weekday mix consistent, but it does not remove seasonality, public holidays, an advertising campaign or a change in demand. If those matter, add the same period last year as context. Do not replace one comparison with whichever makes the graph look better.
Keep partial days out of the primary comparison. Search Console can show preliminary data, and its page and property aggregations differ. A page breakdown is not necessarily a partition of the headline chart. Google's Performance report guidance explains those reporting boundaries.
Define the business outcome before counting it
“Leads” needs a definition. A button click, submitted form, sales-qualified enquiry and paying customer are separate events. Choose the outcome the team can actually verify. If qualification happens in a CRM, record the rule there: for example, a genuine enquiry for an offered service in an area the business covers. That definition is an operational choice, not a Google metric.
Do not silently replace a missing outcome with a weaker one. If completed-enquiry tracking broke halfway through the period, report the break. A row labelled “unavailable” is more useful than a tidy but misleading zero. Our Search Console integration guide shows how to check the search side of this contract.
Explain the movement before assigning credit
Consider this simplified dataset. The two query groups are disjoint, use the same reporting dimensions, and contain no hidden rows. Those assumptions make the calculation suitable for teaching; a real Search Console export may be incomplete.
| Query group | Earlier impressions | Earlier position | Later impressions | Later position |
|---|---|---|---|---|
| Established service queries | 1,000 | 12 | 1,200 | 9 |
| Newly visible research queries | 0 | Not available | 800 | 55 |
| Combined visible cohort | 1,000 | 12 | 2,000 | 27.4 |
The established queries improved from position 12 to 9. Yet the combined average deteriorated to 27.4 because the site also began appearing for a new group at lower positions. The later weighted average is (1,200 × 9 + 800 × 55) ÷ 2,000. An unweighted average of 9 and 55 would be 32 and would describe neither group's actual exposure.
Now add clicks. Suppose established-query clicks rose from 30 to 48, while the new group produced four clicks. Total clicks rose from 30 to 52. The combined CTR fell from 3% to 2.6%, even though the established group's CTR improved to 4%. Both the position headline and the CTR headline look worse while the established group performs better.
What this example establishes
It establishes a composition effect under the stated assumptions. It does not show that a real site's falling average is harmless. Established queries could also be declining. Split the observations before choosing the story, and retain the original query set so you can repeat the comparison next month.
Google's definition of clicks, impressions and position is the reference for the actual metric. In practice, do not reconstruct property totals from overlapping page rows, assume omitted queries had zero impressions, or treat a top-result position as every result's position.
What it does not establish
The new queries may be irrelevant. More impressions for a topic the business cannot serve are not necessarily valuable. Review a sample of the actual queries and landing pages. Ask whether the reader's next sensible step connects to the business. If not, the report should describe broader exposure, not better acquisition.
This is where a good analyst earns their space in the report. The chart identifies movement; the analyst checks whether the movement matters. “Visibility expanded into early research questions, but their commercial relevance remains unproven” is a more useful sentence than “SEO visibility increased 100%”.
Build a scorecard with four layers
A single score cannot represent delivery, discovery, engagement and revenue without hiding a tradeoff. Keep four layers in a compact table, with one observation and one limitation per layer.
| Layer | Evidence to show | Question it answers | What it cannot prove |
|---|---|---|---|
| Delivery | Changed URL, date and live verification | Did the promised work reach the website? | That rankings improved |
| Search | Fixed page/query cohort, clicks and impressions | Did relevant search exposure change? | That the edit caused the change |
| Business | Verified enquiries or completed product actions | Did useful activity happen? | Which channel deserves all the credit |
| Economics | Qualified opportunities, customers and attributable costs | Is the work commercially worthwhile? | A reliable return when attribution is missing |
Keep the raw counts next to the percentages. Two enquiries becoming four is a 100% increase, but the difference is two enquiries. That may matter to a small business; it still deserves a very different interpretation from 200 becoming 400. Percentages communicate scale poorly when the denominator is tiny.
Do not colour every increase green. More support searches can mean existing customers are confused. More impressions can come from irrelevant queries. More enquiries can include spam. Choose the success definition before looking at the direction of the line, then explain exceptions in ordinary language.
Connect each important change to a page
The most useful unit in an agency report is often a short page-level note. It gives enough context to turn a metric into work without drowning the reader in exports. Use the same fields each time, but let the reasoning change with the evidence.
Example: emergency plumbing service page. Assume a hypothetical business serves one city. Its page receives relevant impressions, but the opening paragraph lists services without saying where the team operates or whether it handles urgent jobs. The proposed work is to clarify those facts, show the actual response process and answer booking questions. It is not to repeat the city name more often.
Record the observation, alternative explanations, proposed change, owner, release date, verification rule and later review date. The alternative explanations matter. A low click count might reflect weak search visibility, a poor snippet, an irrelevant query mix or low demand. A content rewrite is only sensible after deciding which explanation it addresses.
Write acceptance criteria that can fail
“Improve the page” cannot fail a review because it says nothing precise. A useful acceptance criterion might be: “The live page states the service area, explains how urgent bookings work, links to the booking form, and the form submits successfully on mobile.” These are delivery checks. Keep the later search outcome in a separate field.
For search, choose observations such as whether the retained relevant query group receives more clicks over a comparable period. Record the raw values. If the page remains too low-volume to interpret, say so. Do not convert an unclear outcome into success because the new page reads better.
Keep a change log that explains interference
Log changes outside the SEO ticket when they affect interpretation. A pricing update, a paid campaign, a new phone number, a migration or a broken form can change enquiries independently of search rankings. A sudden sales improvement after a form repair may be real and valuable without being evidence of stronger search visibility.
The log should let a colleague reconstruct the timeline later. It needs dates, URLs and concise descriptions, not every commit or chat message. Serse's reporting workflow is one place to organise the work; the underlying discipline also works in a shared document.
Decide what to do with contradictory evidence
Reports are easiest to write when everything improves. Their value is clearer when signals disagree. Use a diagnostic table to choose the next question instead of smoothing the disagreement away.
| Observed pattern | Plausible explanation | Next useful check |
|---|---|---|
| Clicks up, qualified enquiries flat | New traffic is less relevant, or conversion is weaker | Query intent, landing page and enquiry qualification |
| Impressions up, CTR down | Exposure expanded into lower positions or different intent | Established versus new query groups |
| Position improves, clicks flat | Demand changed or the visible query mix shifted | Impression counts and the same query cohort |
| Enquiries up, search clicks flat | Another channel or conversion improvement contributed | Campaign timeline, form changes and source evidence |
| Tool totals disagree | Scope, dates, aggregation or collection differ | Reconcile one property, period and dimension at a time |
Treat this table as a set of hypotheses, not a lookup table for certainty. Several explanations can coexist. Start with the check most likely to change the next action. If a broken form explains the business problem, fix and verify it before spending another week debating title wording.
A useful report also says what the team is choosing not to do. For example: “We are deferring another article on the same topic until we know whether the service page answers the current queries.” That is a concrete resource decision. It is more persuasive than a list of twenty tasks with equal priority.
Finish with a recommendation someone can challenge
The executive paragraph should contain the finding, the confidence level, the decision and the review condition. Here is a completed example using the hypothetical data above:
“Clicks increased from 30 to 52. Established service queries improved, while new research queries pulled the combined position average down. We will improve the emergency-service page's booking explanation before commissioning another general guide. The search evidence supports better visibility in the retained cohort; the commercial effect remains unproven because qualified enquiries did not increase. We will review the same cohort and verified enquiry count after the next comparable period.”
That paragraph is deliberately easy to challenge. A colleague can inspect the cohort, question the proposed page, or show that the enquiry definition is wrong. A report that cannot be challenged usually cannot guide a decision either.
Before sending it, perform a final reading without the charts. Can the recipient tell what happened, what the agency actually delivered, what remains uncertain and what happens next? If not, add the missing reasoning. More screenshots will not supply it.
A practical monthly review sequence
Begin by checking collection freshness and the reporting contract. Reconcile obvious breaks before writing explanations. Next, examine the retained page/query group, then the newly visible queries. Only after that should you look at the all-query headline and ask whether it fairly represents the work.
Review the change log beside the search observations. Pull in verified business outcomes with their definitions and limitations. Write a page-level decision for the few items that deserve action. Keep the rest of the data available as an appendix rather than forcing every number into the main story.
Finally, preserve the report and its scope. Next month should begin with the previous recommendation: did the team do it, what happened afterwards, and did the evidence change the decision? This creates a learning record rather than a sequence of disconnected success slides.
Our open SEO improvement study applies that distinction to Serse's own small baseline. Its value depends on retaining disappointing or inconclusive observations as faithfully as promising ones. That is the reporting standard an SEO company should be willing to use on itself.