Gives you the exact day it changed
Not the day the chart happens to look lowest — the day your traffic settled at a new level, with the figures either side of it. That date is what you take to your release log.
Your traffic is down and everyone has a theory. Upload your Search Console or GA4 daily export and find out exactly when it changed, by how much, what it has cost you, and whether a Google update was rolling out that week. Nothing leaves your browser.
Drop a Search Console or GA4 daily export here, or
One row per day, a date column and a count. At least 42 days. Nothing is uploaded.
The file is read and analysed in this browser tab. It never reaches our server, so there is nothing for us to keep.
A line chart shows you that something happened. It will not tell you which day it happened, whether it was real or just a bad week, or what it has cost you. This does all three, and is honest about what it cannot tell you.
Not the day the chart happens to look lowest — the day your traffic settled at a new level, with the figures either side of it. That date is what you take to your release log.
Every confirmed Google update and when it rolled out. If one overlaps your date you get its name — and a straight warning that overlapping is not the same as causing.
Almost every site is quieter at weekends. That weekly rhythm is taken out before anything is measured, so it cannot be mistaken for a decline.
What your traffic was on course to be, against what it actually was. The gap is the cost, counted in clicks rather than in adjectives.
Outages, spikes and tracking gaps are listed on their own, because a single strange day needs a look at your logs, not a change of strategy.
The dates it found with their matching updates, and every day of your data with the expected figure beside it. Check the working or chart it yourself.
In Search Console open Performance, set the date range as wide as it goes, hit Export and take the Dates sheet. A GA4 daily export works too. One row per day is all that is needed.
The date and traffic columns are found for you, and days missing from the file are marked rather than quietly closed up. Everything runs in your browser; the file is never uploaded.
You get the date, the size of the drop, any Google update rolling out that week, and what it has cost. Check your deploys and releases for that date before you accept any explanation.
Nearly every organic traffic decline is one of the following. The point of running the numbers first is that the date narrows the list before you start guessing.
This is the part that is hard to do by hand. Most people answering “was it an update?” open somebody else’s volatility chart, squint at two lines and decide. Here, your own date comes out of your own data first, and only then is it checked against 51 confirmed Google updates — every core update, the spam updates, the helpful content and reviews updates, and dated launches like page experience and AI Overviews. If one was rolling out around your date, you get its name, its dates and how close it was.
Now the uncomfortable bit, which this page will keep repeating because every core update checker on the internet skips it. An overlap in dates is a coincidence in time. It is not evidence of cause. Google ships a core update roughly every quarter and rollouts run two to six weeks, so a meaningful share of any calendar year is inside some window — which means a match is easy to find and cheap to believe. Treat a traffic drop after a core update as a hunch, then test it: did the loss spread across the whole site or concentrate on one template? Did impressions fall with clicks, or only clicks? Did a deploy go out that week?
The list ends at 22 September 2025. Anything Google has shipped since is not in it, and an empty update column means “nothing in this list” rather than “nothing happened”.
Five things come out of a run. They are different findings with different fixes, so they are reported separately rather than blended into a score.
Start by working out whether the traffic fell off a cliff or slid down a hill, because the two have different causes. A step — one date, a new level, and it stays there — points at something that happened on that date: an algorithm update, a migration, a robots.txt change, a broken tracking tag, or a large referrer disappearing. A slide over months points at competitors gaining ground, content ageing, or demand falling. Find the shape first and you have halved the list of suspects.
Pin down the date the level changed, then check what happened on that date. Work through deploys and template changes, robots.txt and noindex tags, redirects and migrations, analytics or Search Console property changes, and whether the loss is concentrated in one country, device or page template. Losses spread evenly across a whole site suggest an algorithm update; losses on one template suggest something you shipped. Always confirm the same window in a second metric before accepting any explanation.
A core update is a broad change to how Google's ranking systems assess content, released a few times a year. Rollouts typically take two to six weeks, during which rankings move about before settling. Core updates are not penalties and there is nothing to fix in the technical sense: sites lose visibility because other pages are now judged a better answer, not because something was marked against them.
Find out what you lost before changing anything. Compare the queries and pages that fell against those that held, and look for a pattern: a topic, a template, a content type, a search intent. Then improve those pages on their merits — depth, accuracy, first-hand evidence, a clear author, a page that answers the query rather than circling it. Recovery usually arrives with a later update rather than within days, so expect weeks or months, not a quick reversal.
That is a results-page problem rather than a ranking problem. You are still being shown as often, but fewer people are clicking through. The usual causes are an AI Overview or featured snippet answering the query above you, more ads or a shopping carousel pushing you down the page, a competitor taking the position above, or a title and description that no longer match what people want. Rewriting titles for intent is the fastest lever here.
Compare like with like. Check the same weeks last year, and remove the weekly pattern before reading the trend, because almost every site has a weekday rhythm that makes any Saturday look like a collapse. A seasonal dip recovers on the same schedule every year; a real drop sets a new level and stays there. If the line is still below last year's equivalent weeks once the weekly cycle is taken out, it is a drop.
Changepoint detection is a statistical method that finds the dates where the underlying level of a series changed, rather than the dates where the chart happens to look low. It works by asking whether splitting the series into segments explains the data well enough to justify the split, with a penalty that stops it cutting at every wobble. For traffic analysis it answers one very practical question: on exactly what day did the new normal begin?
Enough history before the drop to establish what the old level was. About 42 days of daily figures is the practical floor, and six months or more is where the answers become solid. The reason is simple: separating a weekly pattern from a genuine level shift takes several repeats of that pattern, and a trend estimated from four weeks is mostly noise. Search Console keeps sixteen months, and all of it is worth exporting.
A date narrows the search; it does not end it. If the break lines up with a migration or a template change, crawl the site and look at what moved: the free SEO audit tool checks canonicals, noindex tags, titles and sixty-odd other things that break quietly during a release. If pages started 404ing, the broken link checker will find them and tell you which pages they sit on.
The same method works on anything with a date column and a daily count, which is why the review sentiment analyser sits alongside it — a drop in traffic and a turn in reviews often share a cause, and both files usually arrive as a CSV nobody has time to open.
All of these are pieces of Meelu, a desktop app where an AI marketing agent runs your marketing on your own machine — it watches the series for you, so the next drop arrives as a message rather than a surprise in a quarterly review. Join the waitlist.