In short:Your SEO dashboards lie to you. Here’s how to find the real data. — 1975. Economist Charles Goodhart poses a simple law.
Goodhart’s Law applied to SEO
1975. Economist Charles Goodhart poses a simple law.
« Any observed statistical regularity tends to collapse once pressure is exerted upon it for control purposes. »
Translation: the moment a metric becomes a target, behavior shifts to hit the metric — not to improve what it originally measured.
This is the core problem in SEO reporting.
Real example. An agency commits to « average position » in Google Search Console. To hit it, they publish dozens of articles on long-tail queries — zero competition, easy top placement. Average position climbs. Traffic on commercial queries stagnates or drops. Target met. Business performance vaporizes.
Not a theory. The majority of SEO audits on sites that worked with KPI-driven agencies reveal this pattern.
74 % of SEO teams report having optimized for a KPI that improved the dashboard with zero positive impact on revenue (source: Moz State of SEO, 2025)
74 %of SEO teams report having optimized for a KPI that improved the dashboard with zero positive impact on revenue (source: Moz State of SEO, 2025)
The 4 metrics that mislead the most
Metric 01
Average position — the smoothing that hides regression
Average position is calculated across all queries that generated impressions. If you publish content on low-competition queries where you rank 1st easily, average position climbs mechanically — even if your core commercial queries drop from 3rd to 7th place. The metric smiles. The business suffers.
Metric 02
Bounce rate — the metric that depends on the definition
Bounce rate in Google Analytics 4 doesn't measure the same thing as Universal Analytics. In GA4, a "bounce" is a session with no engagement event (under 10 seconds, single page view, no clicks). A user who reads your article for 4 minutes and leaves without clicking counts as a bounce. The metric measures interaction, not usefulness.
Metric 03
Impressions — visibility that costs nothing
Impressions grow every time you publish — even content that ranks on page 4 and nobody actually sees. A +47 % growth in impressions over six months can easily correspond to flat traffic. An impression is counted the moment the page appears in search results — even at the bottom, even with no click.
Metric 04
Domain Authority / Domain Rating — private scores that depend on the source
DA (Moz) and DR (Ahrefs) are proprietary estimates. They measure the likelihood a domain will rank well — not its actual performance. Sites with DR 15 rank ahead of sites with DR 60 on specific queries. These scores depend on tool crawl completeness, not Google. Optimizing them as KPIs means optimizing for Ahrefs, not Google.
Confirmation bias in AI reporting tools
AI-enriched SEO reporting tools add another layer to the problem.
An auto-generated report selects metrics that progress and highlights them. The writing algorithm is trained to produce readable reports — which means reports that tell a coherent story. A coherent story is usually a positive one.
Result: the report says "average position up, impressions up, new keyword coverage." It doesn't say "the 12 queries generating 67% of your commercial traffic dropped 2 positions on average."
Real case. A fashion e-commerce reports SEO progress over 6 months based on automated AI dashboards. Impressions +38 %, average position +1.4. Manually examining segmented GSC data: transactional query traffic (purchase intent) dropped 17 % over the same period. All visible growth came from informational queries with zero conversion.
How to build a trustworthy SEO dashboard
A trustworthy dashboard doesn't show only gains. It shows reality — even when it stings.
Principle 1 — Measure a fixed panel of queries
Define the 20 to 50 queries that actually convert. Not high-volume ones. Revenue ones. Track these queries over time. A fixed panel doesn't lie. Global average position can mask regressions. Panel position never does.
Principle 2 — Cross GSC data with conversion data
Organic traffic crossed with conversions immediately reveals whether growth is commercial or cosmetic. The right metric: revenue per organic session, segmented by entry page type. Traffic growth without growth in this ratio signals a quality problem.
Principle 3 — Systematically include a decline list
An honest dashboard has a dedicated section for indicators that dropped. Which panel queries lost positions? What content saw CTR decline? Which pages lost inbound links? Automated reports rarely ask these. They're the early warning signals.
Principle 4 — A monthly review without tools
Once monthly, open Google Search Console directly. Browse raw data without third-party tools. Tools filter, round, normalize. Raw GSC data shows what actually happened. It's the necessary counterweight to auto-report confirmation bias.
Do your SEO KPIs reflect business reality?
A measurement audit identifiés misleading metrics in your current reporting and rebuilds a tracking system anchored in commercial indicators. 30 minutes for a first diagnosis.
True data exists. You just have to know where to look.
An honest SEO dashboard won't tell you everything is fine when it isn't. It also won't tell you everything is broken when the fundamentals are solid.
It will tell you the truth: what's actually progressing, what's flat, and what's declining under flattering metrics.
That truth is uncomfortable. It demands you measure conversion data, not traffic data. Measure value generated, not volume produced.
SEO teams that outperform in 2026 have one thing in common: they measure what's hard to measure, not what's easy to optimize.
Goodhart was right in 1975. He's right today. The only protection against his law: choose metrics you can't game.
Building a layered SEO dashboard: raw data, interpreted data, decisions
Most SEO dashboards blend everything. Raw data (impressions, crawl budget, Core Web Vitals), interpreted data (average positions, organic traffic), and decision indicators (opportunities, alerts). Result: dashboards nobody truly understands. Everyone interprets differently.
Layered architecture solves this. Three distinct levels. Each with its own audience.
Layer 1: raw data
Access reserved for technical profiles. No interpretation. Factual. Sourced.
Typical content:
Crawl logs (URLs crawled, HTTP response codes, crawl frequency per section)
Index Coverage Report GSC (pages indexed, excluded, with errors)
Core Web Vitals raw (LCP, FID, CLS) per URL, not aggregated
Raw impressions and clicks per query and per URL (no smoothing, no comparison)
Raw backlink data (new links, lost links, referring domains)
This data triggers no action by itself. It validates or invalidates hypotheses from upper layers.
Layer 2: interpreted data
The layer of trends and comparisons. Same raw data, but aggregated, compared over time, contextualized.
Typical content:
Organic traffic evolution by segment (branded / non-branded / long-tail)
Average position by topic cluster (not individual keywords)
Click-through rate by position (to detect CTR anomalies)
Indexed vs. crawled pages evolution (coverage ratio)
HTTP response code distribution trends (200/301/404 patterns)
This layer is read weekly by the SEO lead. No recommendations. Just trends.
Layer 3: decisions
The most important layer. The most often missing.
Content: concrete actions triggered by Layer 1 and 2 observations, with status (to do, in progress, done, result measured).
Format:
Signal observed: "CTR on branded queries dropped 18% between March 15 and March 28."
Hypothesis: "Google shows more SERP features (knowledge panel, sitelinks) capturing clicks before our result."
Action: "Add structured data Organization + Website to homepage. Verify knowledge panel."
Timeline: "Live April 2. Measure at day 21."
Expected outcome: "Recover 8-12% of branded CTR."
3 layers
— raw data (technical), interpreted data (strategy), decisions (action) — this is the architecture letting a 2-person SEO team manage 50,000 pages without endless troubleshooting.
The decision layer is also the antidote to Goodhart's law. Every action is logged with its trigger signal and expected outcome. Metrics stay navigation tools — never ends.
Minimal tooling for this dashboard
No fancy platform needed. A Google Sheet with 3 tabs and charts linked to GSC via API covers 80% of projects.
Structure:
"Raw" tab: auto-imported data from GSC API and crawl tools
"Trends" tab: comparison formulas and charts from the Raw tab
"Actions" tab: manual kanban updated each analysis session
Data import automation (Google Apps Script or Looker Studio connector) takes 3-4 hours to set up. It saves 2-3 hours weekly afterward.
The 6 SEO KPIs that actually measure business value
Organic traffic measures visibility. Not value. A site with 1 million monthly organic visitors may generate less business than a site with 50,000 targeted visitors.
Here are the 6 KPIs measuring what SEO actually returns. Not what it generates in volume.
KPI 1: qualified organic traffic
Definition: organic visitors whose on-site behavior signals intent — session time over 2 minutes, pages viewed over 2, or a micro-conversion trigger (form started, contact page visited, product added to cart).
Why it beats raw traffic: viral content poorly targeted can triple organic traffic. Zero qualified leads out. Qualified traffic stays stable or grows. Right people arrive.
31%
is the average qualified organic traffic ratio on optimized B2B services sites. Moving from 31% to 38% typically outpaces doubling total traffic in business value.
KPI 2: revenue attributable to SEO
Not traffic. Revenue.
In Google Analytics 4: create an "organic source" segment and link conversions — purchases, leads, calls — generated from it.
Refine by attribution type. Last-click overstates SEO on short paths. Data-driven attribution shows true SEO contribution on multi-touch journeys.
This KPI enables real SEO ROI calculation: attributed revenue / SEO investment (time + tools + vendors). Ratio under 3? Strategy needs rework.
KPI 3: cost per organic lead
Especially relevant in B2B and services. Divide total SEO budget (internal + external) by leads generated from organic over the same period.
Compare against paid CPL (Google Ads, LinkedIn). Organic CPL 3-5x lower than paid CPL in your space? SEO is under-invested. Organic CPL higher than paid? Content strategy is targeting wrong.
KPI 4: visibility on high-value queries
Forget global average position. Measure position on a commercial query subset identified as revenue sources.
Building the subset:
Extract all GSC queries generating at least 1 conversion in the last 90 days
Enrich with queries that generated qualified traffic (GA4 segment)
Keep 50-150 most representative queries
Track average position evolution on this subset only. This is your true SEO business health indicator.
KPI 5: Discover coverage rate
In 2026, editorial traffic increasingly flows through Discover. Measure editorial traffic share from Discover vs. classic organic vs. social.
Rising Discover coverage signals your content is validated as trustworthy by Google — which correlates strongly with GEO performance (LLM citability).
KPI 6: LLM presence rate
Essential new KPI in 2026. Measure monthly the percentage of representative questions in your domain for which your site is cited or paraphrased in major LLM responses (ChatGPT, Claude, Perplexity, Gemini).
This KPI isn't in standard dashboard tools yet. Build it manually or via tools like Profound or Otterly. Time investment: 2 hours monthly. Information value: irreplaceable for sectors where conversational search grows fast.
The monthly audit protocol: challenge your own data
A dashboard, however well-built, reflects your assumptions. The monthly audit is the mechanism detecting when assumptions break.
Step 1: cross-source verification
The three main SEO data sources never align perfectly. Google Search Console, Google Analytics 4, and crawl data (Screaming Frog or Semrush) show three different reality views.
These divergences are informative. Not problematic.
GSC shows traffic as Google sees it — impressions and clicks from SERPs.
GA4 shows traffic as your site measures it — sessions and behavior.
Crawl tools show the site as bots see it — structure and accessibility.
Calculate the GSC clicks / GA4 sessions ratio monthly. A stable ratio (0.85 to 1.15) indicates alignment. A drifting ratio (below 0.75 or above 1.30) signals a tracking, redirect, or GA4 configuration issue.
1 in 3
analyzed e-commerce sites have a GSC/GA4 ratio outside normal range — revealing degraded tracking or redirects losing sessions in transit.
Step 2: the simple rule test
For each major SEO decision made last month, ask: "What if I'd applied a simple rule instead?"
Examples:
Decision: "We updated 15 pages because they lost 8% traffic." Simple rule: "Update pages losing 15%+ over 2 consecutive months." Would the decision be the same? If not, the simple rule beats your interpretation.
Decision: "We added internal links to this page because traffic was flat." Simple rule: "Any page ranking 6th-15th with fewer than 3 internal links gets a link optimization." How many pages fit the rule? Big gap? Your interpretation was biased.
Step 3: investigating anomalies
Monthly, flag the 3 biggest data anomalies. A page overperforming unexplainably. A query collapsing without cause. A traffic segment diverging. Spend 30 minutes on each.
Investigation protocol:
Verify the anomaly in 2 different sources. If only one shows it, likely a tracking artifact.
Hunt for external correlation: algo update, competitor move, page change.
Form an explanatory hypothesis and log it in the dashboard "Decisions" layer.
Schedule verification at day 30.
Why anomalies matter: the most valuable insights about Google's algorithm come not from stable trends but unexplained anomalies. A page jumping 14th to 3rd position with no visible change is a lesson in what Google actually values. Understanding it beats 10 generic SEO guides.
Step 4: update your value model
Monthly, recalculate organic CPL and qualified traffic ratio. Compare to last month and same period last year.
If qualified traffic ratio rises but CPL stays flat, on-site conversion optimization is the focus. SEO performs well. The site needs work.
These two diagnoses never appear when mixing all metrics in a single view. They emerge only with separated layers and independently measured business KPIs.
Audit your site in 30 minutes
Get a live diagnostic of your SEO + GEO + AI Search visibility.
