How I diagnosed a 31% traffic drop that wasn't what the SEO team thought
A root-cause analysis from the Hackney Digital ecommerce project.
The Brief
During my data analytics internship with Hackney Digital (UK remote), I was asked to investigate a significant decline in a client's ecommerce store performance. The numbers were alarming: sessions had dropped by 31% over six weeks.
The SEO team's hypothesis: checkout conversion had fallen. They suspected a broken payment flow, a UX degradation in the cart, or a change to the checkout funnel that was causing users to abandon.
My job was to validate or refute this hypothesis using the data.
What the Data Actually Said
The first thing I did was segment the traffic decline by source.
Organic search sessions had dropped 48%. Direct and referral traffic were stable.
This immediately shifted the framing. If checkout conversion were the problem, you'd expect sessions from all sources to decline proportionally — or you'd see a spike in cart abandonment rates while session counts remained stable. Neither was happening.
The sessions weren't converting badly. They simply weren't arriving.
Tracing the Source: Google Search Traffic Collapse
I pulled the Google Search Console data alongside the analytics export. Over the same six-week window, impressions had dropped significantly for the site's top-performing product category pages. Click-through rates were stable — the issue wasn't that ads or search snippets had become less compelling. The pages had lost ranking positions.
Cross-referencing with a crawl report revealed several category pages had recently been tagged with noindex during a site migration. The pages were still accessible via direct URL but had been excluded from Google's index.
The SEO team had been looking at the bottom of the funnel. The collapse was at the top.
How I Reported This
Communicating this to non-technical stakeholders required reframing the entire story.
Instead of saying "the pages are noindex'd", I said: "Google has effectively stopped showing these pages in search results. Potential customers searching for your products aren't seeing the site at all. This accounts for the full session decline — the checkout is working fine."
I built a simple visual showing the timeline: when the site migration happened, when impressions dropped, when sessions declined. The causal chain was clear and non-technical.
The recommendation was equally simple: audit and correct the indexing directives on the affected pages, request recrawling via Search Console, and monitor recovery over the following 4–6 weeks.
What I Learned
In analytics work, the stated hypothesis is rarely the right starting point. The checkout hypothesis was plausible — it fit the visible symptom. But symptoms point to failure modes, not root causes.
Methodologically: always segment before you dive. If a metric drops across all cohorts by similar proportions, the cause is upstream of behaviour. If it drops in one cohort and not others, the cause is cohort-specific.
This is a distinction that most dashboards obscure and most analytics briefs skip. The interesting work is in the segmentation before the modelling.
Tools used: Google Analytics 4 export, Google Search Console, Screaming Frog crawl data, Python (pandas) for aggregation and reporting.