Which SEO Metrics Actually Matter?

The four that survive scrutiny

Search Console alone exposes dozens of numbers, and most SEO reports use that as permission to include all of them. Four carry nearly all the signal:

• Impressions. How often Google showed your pages. It moves first, which makes it the earliest evidence that work is landing.
• Clicks. The outcome, and the slowest thing to move. Reporting it alone makes every early month look like failure.
• Average position, per page. Diagnostic rather than a goal, and genuinely misleading as a site-wide average.
• Organic conversions. The only one the business actually bought, and the only one that lives in GA4 rather than Search Console.

Everything else, and there is a great deal of everything else, is detail underneath those four.

The one ratio worth more than any single number

Impressions against clicks. Watching the two together diagnoses more than either alone, because the direction of the gap names the problem:

The second row is the interesting one and the most commonly misread. It is not a failure, it is a CTR problem, and the fix is the title and meta description rather than more content.

Different metrics for different audiences

The most common reporting failure is showing everyone the same dashboard. Three audiences want genuinely different things:

• Executives: organic revenue or qualified pipeline, and its trend. One chart. Rankings in an executive report invite a question you cannot answer with more rankings.
• Marketing leads: clicks, conversions, and which content is producing them, so they can decide where the next quarter of effort goes.
• You: everything else. Impressions by query, position per page, indexation coverage, crawl findings. This is the working layer, and it does not belong in anyone else's report.

The metrics to stop reporting

• Site-wide average position. It gets worse when you gain impressions on new low-ranking queries, which is progress reported as decline.
• Third-party authority scores. Domain Authority and its equivalents are vendor models, not Google metrics, and they move for reasons unrelated to your work.
• Total indexed pages. More indexed pages is not better. Fewer, better pages frequently outperform.
• Composite SEO scores. Every tool computes them differently, so they are comparable only against themselves.
• Keyword counts. Ranking for 4,000 keywords means nothing without knowing whether any of them are searched.

The metric nobody has yet

One genuine gap in this list, and it is growing. None of the four metrics above capture whether you appear in AI-generated answers, because Search Console does not report it and no equivalent console exists.

That means a growing share of your search presence is invisible to your reporting. A site can gain or lose substantial visibility in ChatGPT, Claude, Perplexity or AI Overviews without a single number in this article moving.

The current substitute is manual and works better than most tooling: run a fixed set of 20 buyer-intent prompts across the major assistants monthly, and record three outcomes separately.

• Mentioned: your brand appears in the answer.
• Cited: the answer links to you, which is the outcome with traffic attached.
• Recommended: the answer actively suggests you over alternatives.

Keep the questions identical between runs. The value is comparability across months, and a rewritten prompt set silently resets your baseline. Record which competitors appear when you do not, because that turns a gap into a brief.

Why the baseline matters more than the metric

A metric without a recorded starting point cannot deliver good or bad news, which makes it decoration. Before any programme starts, record twelve months of organic conversions by month, the impressions and clicks for the pages you intend to work on, and the cost of the work including tools and time.

None of that can be reconstructed later, and Search Console only holds a rolling 16 months, so a second year-over-year comparison depends on capturing history yourself starting now rather than deciding you want it in eighteen months.

If you can only show one number

Sometimes the format forces it: a weekly stand-up, a board slide, a single line in somebody else's deck. The right choice is organic conversions, not traffic.

Traffic invites the question of whether it was the right traffic, and you cannot answer that with more traffic data. Conversions answer it in advance, and they move in the same direction as the work when the work is good.

Pair it with one sentence of context naming the leading indicator, so a flat month has an explanation attached rather than becoming a question you answer defensively later.

The sentence matters as much as the number. "Conversions flat, impressions up 40% on the pages we reworked" is a report. "Conversions flat" is an invitation. Same data, and the difference decides whether the next quarter gets funded.

Write that sentence before you look at the number, based on what you expect. If the result surprises you, the surprise is the report.

How often to look at each

• Weekly: impressions and clicks, for direction only. Weekly ranking data is mostly noise.
• Monthly: position and CTR on active pages, organic conversions, indexation coverage.
• Quarterly: year-over-year comparison, content performance by cluster, and a full technical audit.

For turning these into something a client will read, see how to create SEO reports for your clients, and for the executive framing, the SEO KPI report.