PaperBoat Media · A closer reading of search
Which searches
did you lose?
Organic traffic is down. But the searches that matter to your business may be moving in the other direction.
Open the worked exampleThe report says organic traffic fell 30 percent. That is usually enough to change the mood of a meeting.
Before anyone starts rewriting service pages, I want to know what disappeared. Searches for the work the business wants? Searches for something it mentioned once in an old article? A temporary burst of attention that was never likely to become an inquiry?
A website can be useful to several audiences. Someone reading about a tool, a historical figure or a favorite book may have found exactly what they needed. That does not make the visit a useful measure of demand for a marketing engagement.
If a large unrelated audience shrinks while a smaller relevant audience grows, the total falls. The report is arithmetically correct. The business interpretation still needs work.
This is a close examination of that particular situation. The numbers below are invented so the calculation is inspectable. They are not PaperBoat results, client results or a forecast.
An illustrative report · Two comparable 28-day periods
The total fell.
Look inside it.
One property. Google Web search. The same geography, device coverage and classification rules in both periods.
- Relevant nonbrand
- Brand
- Outside the service objective
- Undisclosed query
| Query group | Earlier | Later | Change |
|---|---|---|---|
| Relevant, nonbrand service searches | 100 | 180 | +80 |
| Brand searches | 100 | 100 | 0 |
| Disclosed searches outside the service objective | 600 | 220 | −380 |
| Clicks without a disclosed query | 200 | 200 | 0 |
| Whole property | 1,000 | 700 | −300 |
+80 relevant clicks − 380 outside the objective = −300 overall
The relevant group grew 80 percent in absolute clicks. It did not merely become a larger percentage of a smaller total. Meanwhile, the other disclosed searches lost 380 clicks, more than enough to outweigh that gain.
The business has a reason to protect the relevant growth. It also has a reason to understand the decline. Perhaps the lost readers supported a separate publishing objective. Perhaps some later became buyers. Calling them outside this service-acquisition objective is narrower, and more honest, than calling them worthless.
The 200 undisclosed clicks stay unclassified. We have not quietly assigned them to the favorable side of the story.
The judgment happens before the filter
“Relevant” needs a definition you would keep in a bad month.
Imagine a consultancy selling SEO, paid-media management, marketing analytics, AI implementation and business storytelling. A search for “marketing analytics consultant” has an evident connection to an offered engagement. “Best cordless drill” does not, even if an interesting article on the same website happens to answer it.
Other queries require judgment. “Marketing dashboard examples” could be research by a future client or an assignment by a student. “AI tools” is broad. “Museum marketing agency” describes a service need; “museum opening hours” describes a visitor’s task. Industry vocabulary alone cannot tell you whether a query belongs in the business group.
I would write the classification against the actual offering: what the firm sells, who it serves, where it can work and which questions reasonably precede an engagement. Then I would apply that definition to both periods, including every relevant query that declined.
The uncertain group matters. A report does not become more rigorous because every ambiguous phrase has been forced into a confident category. If an analyst cannot explain why a query belongs, leave it unresolved until the landing page and business context make the decision clearer.
Brand searches deserve separate treatment because an existing reputation and discovery of a service are different questions. Both can be valuable. Mixing them makes it harder to see whether more people are finding the firm without already knowing its name.
Once the definition is recorded, keep a version of it. If the company adds a service, update the rule and recalculate the comparison where possible. Otherwise a larger category can masquerade as growth.
Visibility and visits
An impression is not a visit.
A falling impression total can hide relevant visibility growth. To say relevant traffic grew, the click comparison must support it too.
| Measure | Earlier | Later | Change |
|---|---|---|---|
| Whole-property impressions | 100,000 | 62,000 | −38% |
| Relevant nonbrand impressions | 10,000 | 18,000 | +80% |
| Relevant nonbrand clicks | 100 | 180 | +80% |
| Relevant nonbrand click-through rate | 1% | 1% | Unchanged |
Here the relevant group received more exposure and more clicks at the same click-through rate. That is a different finding from “the snippet became more persuasive.” Nothing in these numbers establishes an improvement in click-through rate.
For the impression totals to reconcile, the other groups contain 2,000 brand, 68,000 outside-objective and 20,000 undisclosed impressions in the earlier period; then 2,000, 22,000 and 20,000 respectively in the later period. Those numbers are illustrative too.
Google distinguishes impressions from clicks. Throughout this example, “traffic” means Google organic clicks recorded in Search Console. It does not mean Analytics sessions, unique people, inquiries or revenue.
If the relevant impressions rose while relevant clicks stayed flat, I would report greater visibility. I would not describe that as more visitors. And if clicks rose but suitable inquiries did not, I would investigate what happened after arrival before declaring a commercial improvement.
Your second question
Is the growth getting broader?
One successful phrase can lift a total. Growth across more of the work you sell is a different development.
“We rank for more keywords” is not enough. Singulars, plurals, reordered words and one-impression appearances can inflate the count without adding a new business opportunity.
For this example, group searches by the service need they express. Define all five groups before the comparison, including the two with no disclosed activity in the earlier period. Within each group, count disclosed queries once and assign ambiguous phrases consistently. A phrase must not land in two groups and double the apparent growth.
| Service need | Impressions Earlier → Later | Clicks Earlier → Later |
|---|---|---|
| SEO advice | 4,000 → 5,000 | 40 → 50 |
| Paid-media management | 3,500 → 4,500 | 35 → 45 |
| Marketing analytics | 2,500 → 4,000 | 25 → 40 |
| AI implementation | 0 → 2,500 | 0 → 25 |
| Business storytelling | 0 → 2,000 | 0 → 20 |
| Relevant nonbrand total | 10,000 → 18,000 | 100 → 180 |
Service groups meeting the example’s observation threshold: at least 500 disclosed impressions and 5 clicks per 28-day period.
That threshold is an explicit reporting choice for this invented dataset, not an industry benchmark or a statistical significance test. Its job is to stop one fleeting appearance from being counted as an established area of visibility.
The existing three groups gained 35 clicks, from 100 to 135. The other two contributed 45. That tells us more than a single “80 percent growth” label: the result combines deeper activity in established services with disclosed activity in additional services.
Zero here means zero in the disclosed group for the selected period. It does not prove that nobody searched, that Google had never shown the site, or that every later click came from a newly acquired audience.
Before calling this sustained breadth, I would repeat the same view across later comparable periods. I would also inspect the landing pages. If storytelling queries are arriving on a general biography instead of a useful explanation of the service, the visibility needs a closer look.
Taking the example into Search Console
Keep the unfiltered report on the desk.
Start with the same property and Web search type. Compare equal, completed periods, keeping country and device settings consistent. For a seasonal business, add the appropriate prior-year or season comparison. Do not keep adjusting dates until the chart tells a more comfortable story.
Save the whole-property totals first. Then examine the disclosed queries from both periods together, apply the recorded classification, and compare each group’s impressions and clicks. A shortlist drawn only from this month’s winners will miss the losses you need to understand.
Search Console supports query filters, including regular expressions. A pattern can help collect candidates for review, but matching a word is not the same as identifying intent. “Marketing jobs” can match a marketing filter without describing a prospective client. Google documents the filtering and comparison controls here.
For a small report, an export with a query, service group, inclusion reason and both periods’ counts can be enough. Preserve the unclassified rows and the original export. Someone else should be able to reproduce the totals without having to guess which phrases were quietly removed.
The example’s undisclosed row is a deliberately simplified accounting device. In a real export, a gap can involve disclosure and retrieval limits; check those before naming it. Query rows and page rows can also use different aggregation. A page-group total is useful corroboration, but it should not be forced into a property-level reconciliation.
A second view of the intended service pages helps answer whether relevant attention is reaching the right destinations. It cannot recover the intent of an undisclosed query. Keep the two views labeled rather than presenting a page filter as an exact commercial-query filter.
The conclusion has to survive a skeptical reader
What would make this a bad explanation?
The relevant group shrank too.
Then unrelated traffic loss does not explain away the business problem. Investigate the relevant decline, including affected services and pages. A smaller share of a worse total is not protection.
The definition changed between periods.
Adding new phrases only to the later group makes the comparison unreliable. Apply the revised definition to both periods and show the effect of the change.
The gain came from one narrow pocket.
A single query, location, page or short burst may account for most of the increase. Show that concentration before describing broader demand. Several spellings of one need are still one need.
The useful inquiries went the other way.
Search relevance is evidence about discovery, not proof of buying intent or commercial value. Check suitable inquiries and their progress using consistent definitions and enough time for the sales process.
These checks are not an argument for distrusting every positive result. They make the positive result more useful. “Relevant disclosed clicks increased across five service groups” is a claim a reader can inspect. “The traffic is better now” leaves almost every important word undefined.
If the decline extends beyond query composition, use the broader organic-traffic diagnostic guide to examine availability, intent and the arrival experience. This example deals with the interpretation of the search mix, not every possible cause of a traffic drop.
What I would put in the report
“Total Google organic clicks fell 30 percent. Within the same definition of relevant nonbrand searches, disclosed clicks rose from 100 to 180. The loss was concentrated outside that service-acquisition group. Growth also extended from three to five service groups meeting our stated observation threshold. We still need to confirm persistence and the quality of the resulting inquiries.”
A model conclusion for the illustrative data above. Not a statement of PaperBoat performance.
That gives leadership something to act on: preserve the relevant gains, understand the lost audience in its own context, and check whether the increased attention becomes worthwhile conversations.
The biggest number on the report can still be useful. It just cannot tell you which searches you would pay to earn back.
Bring the report
that needs a closer look.
My marketing analytics and reporting work examines the definitions and evidence behind the conclusion. KPI and dashboard design makes that context visible when the next report is opened.
Talk through your search report