Diagnosing a Sales Drop: A Step-by-Step Investigation

A sales drop is really two different problems wearing the same coat. Either fewer people are seeing your listings, or the same number are seeing them and fewer are buying. Those have almost nothing in common as fixes — one is a visibility problem, the other is a listing problem — so the whole job is telling them apart before you start changing things.

First, what Etsy actually gives you

This part explains everything else, so it's worth getting straight up front. Etsy's API publishes a counter, not a history.

What exists is a single lifetime views number per listing, tabulated once a day, for active listings only. There is no views-over-time series, no daily breakdown, and no “visits” figure at all — that field doesn't exist anywhere in Etsy's API, and there's no traffic or analytics endpoint to ask.

So when a tool shows you “views fell 79% week on week”, that comparison did not come from Etsy. It can't have. It came from something recording that counter repeatedly over time and measuring the gaps.

One consequence worth internalising: because the counter is cumulative it never goes down. “Views fell” never means the number decreased — it means the rate it was climbing at slowed. That's why a real comparison needs two periods, not one.

Step 1: Let Atlas collect the history

This is the actual first step, not a delay before the first step. Atlas snapshots that counter for every listing as it syncs, and those snapshots are the entire basis of everything below. Until enough of them exist, there is genuinely nothing to compare.

Both tools say so plainly rather than showing you a number built on too little. On a real shop that had been syncing for four days, Stale-Listing Detector showed exactly this and no listings at all: “Not enough view history yet — this fills in once your shop has synced for about a week.”

• Stale-Listing Detector needs about 7 days — it only compares one window against a threshold.
• Root Cause Radar needs about 14 days, because it compares two 7-day windows against each other. It shows a “warming up” state with the days it has and the days it needs.

If you're setting up during a slump and want answers today, this is the honest answer: you can't have them yet, from Atlas or anywhere else, because the underlying history didn't exist before you started collecting it. Connect the shop and come back in a week — that week is the tool working, not stalling.

Step 2: The wide pass (Starter)

Once you have a week, Stale-Listing Detector gives every active listing a simple engagement level from real view and favorite activity. It's deliberately blunt: how many views a listing gained in the last 7 days, against flat thresholds.

• Suppressed — 0 views in the window. The clearest possible signal, and it needs no comparison to be meaningful.
• At risk — 1 to 4 views. Some traffic, but little enough to watch.
• Healthy — anything above that.

Use this as the wide pass. It won't tell you why, but it will tell you which listings stopped being seen at all, and that's usually where a shop-wide drop is concentrated. More detail in Using Stale-Listing Detector.

Step 3: The diagnosis (Growth)

Root Cause Radar is the same underlying data read properly: this week's views against last week's, per listing, alongside what you changed in the same window. It's Growth-tier.

Its most useful panel is Traffic vs. Conversion, and its logic is narrow on purpose. It only considers listings whose orders are currently down week on week — a listing selling as well as last week has nothing to diagnose, whatever its views did. For those that qualify, it splits them:

• Likely a traffic problem — orders down and views down with them. Fewer people are arriving.
• Likely a conversion problem — orders down but views held up. People are arriving and not buying, which points at price, photos, or the listing itself.

Each listing also gets a What changed (last 14 days) section — your own edits to title, tags, price or photos, labelled by week. The changes window is wider than the views window on purpose, since Etsy's reindexing can lag behind an edit.

Read that section as a timeline, not a verdict. The panel says so itself: Atlas can't tell you whether any change caused the view move — it shows you what changed and when. And where nothing was edited, it says the cause is outside anything it can see, because Etsy publishes no ranking data either.

One more guardrail worth knowing: listings under roughly 15 views a week are left out of the percentage comparisons entirely. At that volume a single extra visitor swings the figure several percent, and a number nobody should act on is worse than no number. See Using Root Cause Radar for the full module.

The two tools side by side

If it's a conversion problem

Traffic problems send you toward search — tags, titles, how findable the listing is. Conversion problems send you toward the listing itself, and price is usually the first suspect.

Before you cut it, though, it's worth knowing what you're actually keeping. Etsy's fees are a larger share of a sale than the headline 6.5% suggests, and a price cut comes out of a margin that's already thinner than the price tag implies — our breakdown of what Etsy really charges goes through the whole stack.

A note on data and plans

For transparency about how this guide was written: the Stale-Listing Detector behaviour above was checked against a real shop — which is how we could show you its genuine not-ready state rather than describing it. Root Cause Radar was read from the product demo, whose figures are sample data, since the real account was on Starter and hadn't accumulated 14 days of history either way.

Which is also the plan situation in short: the wide pass works on Starter, the traffic-versus-conversion diagnosis needs Growth. On Starter you can still find which listings went quiet — you just won't get the split between why.