SEO reporting: how to report results in revenue

A report exists to help someone decide. A board decides whether search is worth what it costs; a marketing lead decides where next quarter’s effort goes; the person doing the work decides what to fix next. Rankings and visits matter to all three, but only as steps on the way to what the business counts: orders, enquiries, bookings and revenue.

That gives every good SEO report the same spine, whatever its length: what was done, what moved, and what it brought in. The first part makes the work visible and checkable. The second shows whether search responded. The third connects those results to money, in figures that carry their source and their label, so that nobody reads an estimate as cash.

A report that stops at the second part is a traffic report. It can be accurate and still useless to a finance team, because it cannot say whether the traffic paid for the work. The rest of this guide is about the third part, and about the definitions that make it believable.

Most arguments about SEO results are arguments about definitions, held months too late. Settle them in writing before the work starts, while nobody has a number to defend.

1. 1Name the actions that count as business: an order, a booked call, a submitted form, a trial that becomes a subscription. In Google Analytics 4 each one is an event marked as a key event, and a standard property can mark up to 30 of them^[1].
2. 2Mark them before the period you will report on. Marking an event as a key event affects reports from the time it is marked and does not change historic data^[1], so a key event set up in March cannot report on February.
3. 3Choose the attribution model and write it down. GA4 offers data-driven attribution, paid and organic last click, and Google paid channels last click^[2]. Changing the reporting model applies to historical and future data^[3], so a quiet switch rewrites every past month of the report.
4. 4Set the lookback window, the time before a key event in which a visit can earn credit: 90 days by default for most key events, with 30 or 60 days as options^[3]. A business whose buyers take months to decide should know which window its figures use.
5. 5Decide where revenue is recorded: the store, the payment system or the CRM. Analytics revenue is a reading of that record, not the record itself.
6. 6Agree how a lead is valued until it closes, and who confirms that it closed.

Keep the definitions in the report itself, on a page at the back, with the date each was set. When one changes, the report says so in the month it changed, and figures from before and after the change are not compared as if nothing had happened.

Organic revenue is the end of a chain, and each link of the chain is counted by a different system. Knowing which system counts what, and where they disagree, is most of the craft of SEO reporting.

Search Console and Analytics will not agree, and should not be forced to. Google lists the reasons: to protect privacy, Search Console omits some queries that are searched very rarely; its processing can remove duplicates; its data usually arrives two to three days after it is collected; it labels its days in Pacific Time; and tools such as Analytics count only visitors whose browsers run JavaScript^[4]. Report each figure from the system that counts it, and never subtract one from the other to find a hidden number.

Linking the two closes part of the gap. With Search Console connected to Analytics, two reports join queries and landing pages to what visitors did and to the key events they reached, but Search Console keeps data for the last 16 months, so the joined reports hold at most 16 months too^[5]. The join works by landing page, device and country^[5], which is also the level at which a revenue report is most useful: the page is what the work changes.

Read the channel definition before reading the channel: in GA4, clicks from Google’s AI Overviews and AI Mode count as Organic Search, while visits from assistants such as ChatGPT have their own AI Assistant channel^[6], a split set out in the guide to AI visibility metrics. A report that says “organic” should say which it means.

Sixteen months allows one comparison with the year before, and little more. Keep the monthly figures yourself from the first report on, in a sheet or a database, so that the second and third years can be compared too, and so that a lost login does not lose the history.

Revenue from search is not one number. A sale can be recorded and traced to a search visit; it can be one that search shared with other channels; it can be an estimate; or it can be the extra that a change added. These are different claims, resting on different evidence, and a report that blends them overstates all of them.

Where only Analytics is connected, the figure is none of the four yet. Label it GA4-reported: it is what the analytics tag recorded and credited by its attribution rule, and it will differ from the orders the business recorded, because consent choices, blocked tags and orders closed offline all stand between the two. It becomes known once the order record confirms it; until then it is GA4-reported.

The rule that assigns a sale to organic search is organic revenue attribution, and the rule changes the number. Under paid and organic last click, 100% of a key event’s value goes to the last channel the customer clicked through before converting^[2]. Under data-driven attribution, credit is shared out between the interactions on the path, so key events and revenue can show fractions for the first time^[3]. And every GA4 model gives direct visits no credit unless the whole path was direct^[2]: a buyer who found you through search and came back days later by typing your address is still credited to search under last click.

Three rules keep the four honest. The label stands beside the figure, in the table and in every sentence that quotes it. The four are never summed into one total, because influenced revenue overlaps the journeys behind known revenue and a modeled figure is not money. And incremental revenue needs a comparison to exist: a group of similar pages left unchanged, a baseline forecast set before the change, or a test. Without one, the report says that it cannot tell what the work added.

Searches that include the brand name come mostly from people who already know the business. They respond to advertising, word of mouth and the product itself more than to SEO, and in many businesses they carry most of the organic revenue. A report that mixes them with everything else credits search work with demand it did not create.

Search Console separates the two: its branded and non-branded queries filter splits the data by whether the search term included the brand name^[7]. Read its limits before relying on it. The filter is not available for sites with a low number of impressions, its history starts from its introduction in March 2025, and Google warns that some queries may be classified wrongly^[7]. Where the filter is missing, a filter on the brand’s names and their common misspellings does the same job; keep that list with the report’s definitions.

Then split the money the same way, as far as the data allows. Analytics does not see the query behind a visit, so the split is made by landing page: revenue that lands on the home page and the brand pages leans branded, revenue that lands on category, product and guide pages leans non-branded. Say that the split is approximate, and keep the rule the same from month to month so that the trend is real even where the level is not exact.

Many businesses do not sell on the page. A clinic books an appointment, a school takes an application, a software company opens a trial, a contractor sends a quote. The report then has two moments to count: the enquiry, which analytics can see, and the sale, which happens later in a system analytics cannot see.

1. 1Count the enquiry where it happens: the form, the call, the booking. Search Console stops at the click, so the business counts its own enquiries at the point of contact.
2. 2Carry the source into the CRM with the enquiry, so that a closed deal can be traced back to its landing page and its channel.
3. 3Until a lead closes, give it a value from your own history, such as the share of past enquiries of that kind that became customers, times their average value. Label that value modeled, and show it as a range.
4. 4When the deal closes, replace the estimate with the known figure, and keep both in the record, so that the model can be checked against what happened.
5. 5Count quality as well as volume. Enquiries that were spam, outside the area served or for a service you do not sell are reported, and then left out of the value.

A month in which enquiries rose and closed revenue did not is not a success yet; nor is it a failure. It is a pipeline whose outcome the report can only state as modeled, with the date on which the known figure will replace it. The next month’s report then closes the loop: of the enquiries search brought last month, how many became customers, and what they paid.

Search results move without anyone touching the site. Google says that positions in its results are not static, that its results change as user expectations and the web itself change, and that this constant change can cause both gains and drops in organic traffic^[8]. A report earns trust by separating the movements the work caused from the ones it did not.

• Compare like with like: the same month last year for anything seasonal, the month before for anything that is not, and always periods of equal length.
• Check the calendar of Google’s updates. After a core update, Google recommends waiting at least a full week after it completes, then comparing with a week before it started^[8].
• Size the move before reacting to it. Google separates a small drop in position, which needs no drastic action, from a large one, which calls for a deeper assessment^[8]; the guide to content decay shows where the line falls.
• Name the cause you believe, and say how sure you are. A cause you inferred is written as an inference, not as a finding.
• Flag a small base. When enquiries go from 6 to 4, say that the numbers are too small to read as a trend.

What a change means for revenue depends on where it happened. A fall in impressions on informational pages that never sold anything is a different event from a fall in clicks on the three category pages that carry most of the known revenue. Report movement by group of pages, with the revenue each group carries beside it, and the reader sees at once which changes matter.

Keep a dated record of every change made to the site, what QueryRev calls the Change Ledger, so that each movement in the numbers can be set against what changed, and when. Pages that lose ground slowly, month after month, have their own diagnosis, set out in Content decay: how to find and reverse it.

The same figures serve different people at different depths. Write the report so that each reader can stop as soon as they have what they need.

For the first reader, the page starts with the answer. One sentence says what changed and what it was worth: “Non-branded organic revenue rose against the same month last year; the gain came from the category pages rewritten in the spring.” The table under it shows known revenue first, then influenced revenue, each with its label, and the cost of the work in the same period. Modeled and incremental figures, where they exist, stand apart with their ranges, so that no reader adds them up.

Leave the ranking chart out of that page. An average position across hundreds of queries can move for reasons that have nothing to do with the quality of the work, such as new queries entering the set at low positions. If rankings matter to the decision, show the few searches that bring buyers, by name.

The monthly version of this report, section by section, is in Monthly SEO report: what to include. The audit that sets the plan a report tracks is in SEO audit report: what it should contain, and which parts of a report a tool can safely produce is in Automated SEO reports: what to automate and what not.

Answers written by AI tools belong in the same report, not in a separate deck: the same buyers, the same pages, the same months. What changes is what can be counted, and therefore which label the money carries.

Visits that AI tools send are observed, in the two channels described above, so revenue from them is GA4-reported as analytics records it, and becomes known once the order record confirms it. A mention in an answer that brings no click is a different thing: whatever it is worth can only be modeled, and the report labels it so, or leaves it as a count of mentions with no money attached.

How to build the set of prompts and count mentions and citations is covered in the guide to measuring results in AI search; which of those counts deserve to become KPIs is in the guide to AI visibility metrics and KPIs.

A report looks back. The same definitions let you look forward. Return on investment compares the profit search brought with what the work cost; a forecast sets out a range of likely results if its assumptions hold. Both go wrong in the same two ways: using revenue where profit belongs, and counting customers who searched for the business by name. Both are worth stating in three cases, low, middle and high, because a single number hides the risk.

The SEO ROI calculator and the SEO forecast template work from your own inputs in those three cases. Keep a forecast and a report apart: a forecast is a range, a target is what you aim for, and an actual is what happened. Report the actual against both, and keep the original forecast unchanged, so that it can be judged when the period ends.

• Adding the four kinds of revenue into one total.
• Changing the attribution model or a key event in the middle of the year without saying so, then comparing months across the change.
• Crediting search with branded demand it did not create.
• Reporting revenue from analytics as if it were the order record.
• Reading a seasonal fall against the month before instead of the same month last year.
• Calling every change that follows a piece of work its result, when a before-and-after on one page shows timing, not cause.
• Hiding the bad news. A decline belongs next to the gains, with its likely cause and what is being done about it.

Each of these makes the next report harder to believe, and a report nobody believes cannot support a decision, which was its only job.