Date the break in your organic traffic.
Traffic is down. You have two weeks of arguing ahead of you.
Seasonality, a tracking bug, or an algorithm update. Every organic decline starts as that argument, and the argument only ends when someone can name the day it broke. Organic dates the break per content cluster, across classic search and answer engines, and tells you which of the three it was.
Fifteen acquisition series, one cluster losing rankings, detection computed live in your browser, including the site total that barely twitched.
The problem
Nobody can date the break, so nobody can name the cause.
Site-level sessions is the chart everyone opens, and it is the one chart that cannot answer the question. A cluster that is a fraction of organic traffic can lose a third of its sessions and move the site number by less than a normal week does. So the decline shows up as a soft slope over a month, three explanations are equally consistent with it, and the team spends the next sprint building the analysis instead of fixing the pages.
The insight
The cause is in the order the metrics broke, not in the size of the drop.
A ranking loss moves average position first, then impressions, then sessions, and click-through falls on top because a worse slot earns fewer clicks at the same impression count. A tracking bug moves sessions and nothing else, because impressions and position are measured on the search engine side and never touched your tag. Seasonality moves every cluster at once and usually moves paid and direct with it. Those three signatures are trivially separable if you have a changepoint date per metric per cluster, and completely inseparable if you have a monthly line chart of the site total. Dating the break is not a nicer way to present the decline. It is the thing that decides what you do next.
CUSUM on day-over-day percentage changes for every metric × cluster series, Bayesian Online Changepoint Detection (Adams & MacKay 2007) for abrupt reranks, and Benjamini-Hochberg FDR control across the whole run. Average position is scored inverted, since position 4 beats position 9 and a rise in that series is the regression.
How it works
Four steps, no data science team
Search Console, your analytics property, and whatever answer-engine referral data your logs carry. Read-only, no tag changes, no new script on the page.
Pages group by template and intent rather than by folder, so a comparison page and a glossary page do not get averaged together. Sessions, impressions, average position, and click-through per cluster, split by surface.
Each series runs through CUSUM and BOCD. Whatever crosses is filtered by Benjamini-Hochberg across the run, so twenty clusters times four metrics does not produce four alerts a day forever.
The alert names the cluster, the date, and which of the three signatures the ordering matches. Rankings, tracking, or everything at once. Then it lists the URLs that moved.
Who it is for
The growth lead who has to explain the traffic chart on Monday
Growth and content teams at companies where organic is a real acquisition channel rather than a rounding error. Usually 200 to 50,000 indexed URLs, a content lead, an SEO contractor or in-house specialist, and at least one decline in the last year that took two weeks to explain.
Pricing
- –Full detection engine
- –90-day history
- –Weekly email
- –Three clusters
- –Unlimited clusters
- –16-month history
- –Classic and answer-engine split
- –Slack alerts
- –Cause narrowing on every break
- –Multi-property rollup
- –White-labelled client reports
- –API access
- –SSO and audit log
Competition
What exists, and what it does not do
| Who | What they do | The gap |
|---|---|---|
| Google Search Console | The impressions, clicks, and position data, free, straight from the source. | Sixteen months of tables and a comparison mode. No changepoint detection, no clustering, and a three-day data lag you have to remember exists. |
| Ahrefs / Semrush | Rank tracking on a keyword set you choose, plus alerting on big movers. | Keyword-level and threshold-based. A cluster that loses two positions across four hundred URLs never trips a per-keyword alert, and answer engines are barely covered. |
| GA4 anomaly detection | Built-in anomaly flags on session and conversion trends. | Runs on session counts only. It cannot see impressions or position, so it can flag the drop but never distinguish a ranking loss from a broken tag. |
| Algorithm update trackers | Volatility indices and confirmed-update timelines from the industry. | Tells you the weather, not whether it rained on you. Every decline lands near some update, so correlation with a public timeline is a story rather than evidence. |
The honest weakness: dating the break is necessary but not sufficient. If the answer for most customers is that a search engine reranked and there is nothing actionable, then we sold a better explanation of bad news, and better explanations of bad news do not renew. The bet is that a dated, cluster-scoped break changes what happens next, because it points at a specific set of URLs and a specific week of changes to diff against, which is what a team needs to actually recover pages rather than rewrite the whole site. If customers keep detecting breaks and keep not recovering, the churn will be quiet and fast.
Market
Priced beside the SEO tool budget, not instead of it
Companies with real organic programmes already pay $100 to $1,000 a month for rank tracking and crawling. This sits next to that spend as the detection layer those tools do not have. Eight thousand properties at the Growth tier is $29M ARR, and the answer-engine split adds a second surface every one of them now has to watch.