Cohort analysis applied to organic content

Monthly organic traffic is a comfortable lie. It lumps posts published in 2021 with yesterday's launches and hands you a number nobody knows how to read. Cohort analysis breaks that average open: it groups pages by publication date (or last meaningful rewrite) and tracks how each vintage performs in the months that follow. That is how I found, at a B2B SaaS client, that the January 2024 cohort was bringing 38% more qualified sessions than the May cohort despite similar volume. The quarterly average said we were flat. The cohorts said quality had collapsed in May.

Building the first cohort needs only three fields: URL, first-indexed date, and organic sessions per week. I pull it from GA4 crossed with Search Console via BigQuery (I covered that workflow in BigQuery + GSC: queries your agency won't run). Each row becomes a cell in a matrix where the X axis is content age in weeks and the Y axis is the vintage. A healthy pattern shows growth until week 16-20 then stabilization. When a cohort drops before week 12, you are looking at an editorial problem or mismapped intent, which I broke down in Search intent: 4 types and how to map them on the SERP.

The next move is enriching the cohort with quality, not just volume. Sit next to sessions: average scroll depth, assisted conversion rate, and SERP CTR. In a portfolio I audited recently, the March cohort had double the sessions of April but half the dwell time (method in Dwell time: measuring engagement without official data). March was inflated by an informational keyword that did not convert. April, smaller, was feeding the lead base. Without the cohort, that insight vanishes inside the aggregate and you keep publishing the wrong kind of content thinking you are winning.

Cohorts also expose content decay brutally. When an older vintage starts dropping consistently for 3-4 weeks, it is time to decide between rewrite and rebuild - a call I worked through in Rewrite or rebuild: making the call with SERP data. The classic mistake is looking page by page and missing that the entire October 2023 cohort, say, was hit by a Google update that shifted dominant intent. Cohorts turn 47 individual drops into one actionable pattern. In Looker Studio or Hex, a simple red-green heatmap is enough for the editorial team to see where review hours should bleed.

A practical case: e-commerce client, 1,200 posts. I segmented cohorts by quarter and crossed them with page type (PLP, PDP, editorial blog). The Q2 2024 cohort of new PDPs had an average LCP of 4.1s versus 2.3s on the older ones - core web vitals killed the growth curve, exactly as I argued in Core Web Vitals: beyond LCP, what actually moves the needle. We swapped the image server theme, recompiled 340 pages, and the cohort recovered its trajectory in seven weeks. Without grouping by vintage, we would have blamed seasonality or a vague algorithm update. Cohorts force you to look at structural causes, not narratives.

Cohorts also support honest forecasting. If every historical vintage hits 70% of steady-state traffic by week 14, you can project that the 80 posts shipped this quarter should add up to X sessions by year-end - the math I laid out in SEO Forecasting: how to project results with confidence. Combined with position-CTR benchmarks (CTR benchmark by position: updated 2026 data) and segmented dashboards (SEO dashboards: what to show the CFO vs the marketing team), cohorts become the backbone of any serious conversation about content ROI. Stop arguing about gut feel, start arguing about curves.

Practical takeaway: take your last 18 months of content, group by publication month, plot organic sessions per week of life, and flag every cohort below the median at week 12. Those are tomorrow morning's audit priorities - not the pages with the steepest absolute drop, but entire vintages that never took off. Cohort analysis is not a report; it is a diagnosis.