How to measure whether SEO is working beyond rankings

A practical framework for judging SEO through implementation, qualified visibility, important landing pages, demand quality, commercial outcomes and assisted influence.

“We are investing in SEO, but what would count as evidence that it is working?”

That is a more useful question than “What position are we ranking in?” Rankings can move in the right direction while the wrong audience arrives, important pages remain invisible or commercial results stay unchanged. The reverse can happen too: a business may be creating useful demand before that progress appears clearly in revenue data.

A sensible answer is to assess SEO as a chain of evidence. First, confirm that the agreed work was implemented. Then examine qualified search visibility, visits to important landing pages, demand quality, commercial outcomes and the role organic search played in recorded journeys.

Each layer answers a different question. None proves the next one on its own.

Start with the work that was meant to happen

Before asking whether an SEO change improved performance, check that it reached the live website in the way it was intended to.

This sounds obvious, but it is easy to skip. A recommendation may have been approved but not deployed. A template change may work on one URL but fail in another market or page type. A release may go live, then be overwritten by a later CMS update.

An implementation record can include:

• the agreed SEO change;
• the affected templates, URLs, markets or page groups;
• the deployment date and release version;
• quality-assurance checks on representative pages;
• known exclusions, errors or follow-up work.

A ticket or deployment timestamp establishes delivery. It does not establish that the change was technically correct, indexed, visible in search or commercially valuable. For larger releases, checking one URL is not enough to show that a template-based change worked consistently across the affected site.

Implementation evidence belongs at the beginning of the measurement chain because it prevents the business from evaluating a result that was never properly delivered. Our guide to when to expect evidence after an SEO change covers the timing question in more detail.

Measure qualified visibility, not just more visibility

Once implementation is confirmed, the next question is whether the site is becoming more visible for searches that matter to the business.

Google Search Console reports impressions, clicks, click-through rate and average position. It can also break performance down by queries, pages, countries, devices, dates and search appearance. These measures describe performance in Google Search, not lead quality, sales or revenue. See Google’s Search Console performance report documentation for the scope of these measures.

The important word is qualified. A rise in total impressions may come from brand searches, low-value topics, an irrelevant country or a page group the business does not want to grow. It may be positive, but it is not automatically commercial progress.

Instead, segment visibility around questions such as:

• Are commercially relevant topics gaining impressions and clicks?
• Are the priority markets and audiences appearing in the data?
• Are the pages designed to support those searches receiving visibility?
• Is growth coming from new non-brand demand, or simply from people who already knew the business?

This classification is partly a judgement call. Search intent is not always obvious, and Search Console query data is subject to aggregation and anonymisation. Detailed query groupings will not represent every search perfectly; Google explains some of these reporting limits in its performance data deep dive.

Average position also needs care. It is an aggregate Search Console measure across the impressions represented in the report, not a single universal ranking for a keyword. Treat it as one visibility signal, rather than a definitive description of where every searcher saw the page.

Follow demand to the pages the business actually wants to grow

Sitewide organic traffic is convenient. It can also hide the detail that matters.

Imagine a software company improving its comparison pages and pricing information. If overall organic sessions rise because an old glossary attracts more visitors, the sitewide number may look healthy while the pages closest to commercial demand remain unchanged.

Analyse important landing pages and page groups separately. Depending on the business, these might include:

• service or solution pages;
• category and product groups;
• pricing, comparison or demonstration pages;
• location or market pages;
• high-value educational content that introduces a buying journey.

This connects search demand with the experience the business wants customers to reach. It is an inference from combining Search Console’s page and query dimensions with analytics landing-page data, rather than a claim that page-group analysis proves causation. Search Console and Analytics measure different parts of the journey, so their totals should not be expected to match exactly. Differences can result from scope, time zone, canonicalisation, consent, attribution and other data-processing details; Google outlines the relationship between the platforms here.

The useful question is not simply “Did organic traffic increase?” It is “Did qualified users reach the pages where the business can help them, and did that change after the agreed work was implemented?”

Separate leading indicators from outcomes

Some measures show that a user has moved closer to a commercial action. Others show that the action actually happened. They should not be placed in the same bucket.

Leading indicators might include:

• visits to priority landing pages;
• pricing-page views;
• comparison-tool interactions;
• trial or enquiry starts;
• downloads associated with a known buying process.

Outcome measures might include:

• qualified enquiries;
• accepted sales opportunities;
• completed bookings;
• new customers;
• revenue or margin.

The distinction matters because an analytics event is a configured measurement construct. It does not automatically mean that a user was a good prospect or that revenue followed. Google’s documentation on key events in Analytics explains the platform concept, but the business still needs to define what makes an enquiry qualified and connect it to downstream records where possible.

For a short sales cycle, a completed booking may be available quickly. For B2B software, education or high-value services, useful evidence may first appear as an accepted enquiry or sales opportunity, with revenue arriving much later. A short reporting window can therefore undervalue genuine progress or overstate the importance of an early-stage action.

Use commercial data when the decision requires it

Analytics can show what happened after a user arrived. CRM, ecommerce, booking or revenue systems can show what happened to the business afterwards.

That does not make CRM data automatically perfect. The conclusion depends on reliable identifiers, sufficiently complete records, consistent definitions and careful handling of consent and source fields. Google supports importing offline events into Analytics, but the technical ability to join data does not guarantee that every offline outcome has been captured correctly. The relevant limitations are set out in its offline event and conversion documentation.

A useful commercial measurement set might connect:

• organic search visibility for priority topic groups;
• organic visits to important page groups;
• enquiry or trial starts;
• qualified opportunities in the CRM;
• customers, bookings, revenue or margin where the sales cycle allows it.

The stronger the connection between these systems, the more confidently the business can discuss commercial progress. Even a well-joined data set does not remove other influences such as pricing, promotions, sales capacity, product availability, paid media, brand activity, competitor behaviour or website changes. Google’s guidance on investigating search-traffic changes also illustrates why external factors should be considered alongside the measured trend; see its traffic-drop debugging guidance.

Include assisted journeys without giving SEO all the credit

Search is often part of a longer journey. Someone may discover a company through an organic result, return through a branded search, attend a webinar and then speak to sales. A last-click report would usually assign the conversion to the final recorded interaction. It would not describe the whole path.

Assisted-journey reporting can show that organic search appeared in recorded paths associated with later key events or opportunities. Google Analytics describes these reports as a way to examine how channels contribute across conversion paths; see its documentation on attribution paths and models.

That is useful evidence of contribution. It is not proof that SEO deserves full credit for the outcome.

Recorded journeys may omit cross-device activity, offline conversations, untagged campaigns, users who did not consent to measurement and incomplete CRM information. Attribution models can also allocate credit according to rules or statistical modelling. Modelled credit should be labelled as modelled or directional, rather than presented as observed revenue ownership.

This is the distinction worth keeping in view:

• Correlation: SEO-related measures and commercial outcomes changed during the same period.
• Contribution: organic search appeared to play a role in a recorded or plausible customer journey.
• Causation: there is credible evidence of what would have happened without the SEO change.

Attribution can help describe contribution. It cannot, by itself, establish the counterfactual: what would have happened if the work had not been done. That is a general causal-inference limitation, not a reason to abandon measurement. For an accessible explanation of the counterfactual problem, see Miguel Hernán and James Robins’ work on causal inference.

A worked example: measuring SEO for a B2B software company

Consider a fictional B2B software company selling workflow software to operations teams. Its agreed SEO work includes improving solution pages, clarifying internal links and publishing comparison content for three priority use cases.

After the release, the evidence looks like this:

• Two of the three solution-page templates were updated. The third is still waiting for a development release.
• Search Console shows more impressions and clicks for the two completed solution groups, particularly for non-brand use-case searches.
• Organic visits to those pages have increased, while the glossary has remained broadly flat.
• More visitors are reaching the pricing page and starting a demo enquiry.
• Several organic-assisted paths are associated with new opportunities in the CRM.
• Revenue data is still immature because the company’s sales cycle is around four months.

A weak report might say: “SEO worked because rankings, traffic and conversions increased.” That compresses too many claims into one sentence.

A more proportionate conclusion would be: the release is only partially implemented, but the completed page groups show improved qualified visibility and stronger movement towards commercial actions. Early CRM evidence is encouraging, although revenue impact cannot yet be judged confidently. Organic search appears to be contributing to some recorded journeys, but the available data does not isolate its causal effect.

That conclusion is commercially useful without pretending the evidence is stronger than it is. It also gives the team a clear next action: complete and validate the third template, continue observing opportunities through the sales cycle and check whether the early-stage improvement persists.

The measurement chain in one view

The following framework keeps each layer tied to the question it can reasonably answer:

• Implementation: Did the agreed SEO change reach the live site correctly? Use deployment records, release notes, crawls and representative QA checks. Do not treat delivery as proof of performance.
• Qualified visibility: Is the site more visible for commercially relevant searches, markets and audiences? Use segmented Search Console queries, pages, countries and dates. Do not treat total impressions or average position as commercial success.
• Important landing pages: Are relevant users reaching the page groups the business wants to grow? Use landing-page analytics joined, where practical, with Search Console page data. Do not assume a rise in sitewide traffic reflects priority-page growth.
• Demand quality: Are users showing useful signs of intent? Use validated actions such as trial starts, qualified enquiry starts, pricing interactions or calls. Do not assume every event represents a valuable prospect.
• Commercial outcomes: Did qualified opportunities, bookings, customers, revenue or margin change? Use CRM, ecommerce, booking and finance data where definitions and joins are reliable. Do not ignore sales-cycle lag or other business changes.
• Assisted influence: Did organic search appear in recorded journeys connected with later outcomes? Use path and attribution reporting carefully. Do not present allocated credit as proof of causal revenue ownership.

What a sensible conclusion looks like

SEO measurement rarely produces a perfectly isolated return figure in routine business conditions. It usually produces a reasoned judgement whose confidence depends on the quality of the implementation evidence, segmentation, data joins, observation period and comparison evidence.

Stronger causal claims may be possible with controlled page-group comparisons, matched controls or interrupted time-series analysis. Those approaches need a defensible design and explicit assumptions. Evidence from other marketing contexts also shows why observational attribution should not automatically be treated as experimental evidence; for example, see this study of advertising attribution. Its findings are not direct evidence about organic search, but the methodological warning is relevant. These methods are not necessary for every business, and a sophisticated model built on incomplete data can create more confidence theatre than insight.

For most organisations, the practical discipline is to avoid asking one metric to do six jobs. Rankings can help describe visibility. Search Console can show search performance. Analytics can describe behaviour. CRM and revenue systems can show commercial outcomes. Attribution can describe recorded influence. Implementation records can establish whether the intended change happened.

Together, those sources can support a much better business conversation than “traffic was up” or “we moved from position nine to position five”. They can also reveal where the evidence breaks: perhaps the work was incomplete, the wrong page group improved, the enquiries were poorly qualified or the CRM join is not trustworthy.

Before the next SEO review, ask three practical questions:

1. Which agreed SEO changes were completed and validated?
2. Which commercially relevant page groups improved?
3. What evidence exists that qualified demand or business outcomes changed?

If Search Console, analytics, CRM or revenue data and implementation records do not line up, the problem is not solved by adding another dashboard. It needs diagnosis, clear definitions and careful joining of the evidence. That is where specialist support can be worthwhile for larger or more complex organisations. For a simple site with a short sales cycle, a small set of validated measures may be entirely sufficient.

The practical aim is not to force every SEO effect into a single return figure. It is to make the evidence chain explicit, show where confidence is strong or limited and choose the next measurement step that will reduce the most important uncertainty.