What Is a Sales Changepoint and Why Should You Care?

Every WooCommerce store owner has had this experience: a product that was selling steadily suddenly isn’t. You don’t notice right away — sales data is noisy day to day, and a slow Tuesday doesn’t mean anything on its own. By the time the decline is obvious in your dashboard, it’s been happening for a while. The trend changed days ago. You just found out today.

This is the exact problem changepoint detection was built to solve — and it’s one of the more genuinely clever pieces of the forecasting technology now sitting inside WhoKnew Views™ Pro.

What a Changepoint Actually Is

A changepoint is a specific point in time where the underlying trend of a time series shifts — where the rate of growth or decline measurably changes, not just where a single data point looks unusual. Changepoint detection identifies sudden shifts in time-series data, uncovering moments where the underlying pattern changes abruptly, with applications ranging from financial trading to web traffic to manufacturing sensor data.

The distinction matters. A single bad sales day isn’t a changepoint — it’s noise, the normal variation every product experiences day to day. A changepoint is when the actual slope of the trend changes: a product that was growing 2% a week suddenly starts growing 8% a week, or a product that was flat for two months starts sliding downward in a way that isn’t going to correct itself on its own.

How Prophet Finds the Moment Things Changed

Meta’s Prophet algorithm — the open-source forecasting engine Facebook’s Core Data Science team built and now runs inside WhoKnew the Future™ — handles changepoints in a specific, well-documented way. Rather than guessing where a trend shift happened, Prophet’s own documentation explains that it specifies a large number of potential changepoints across your historical data, then applies a sparse statistical prior that allows the trend to shift at very few of them — only where the data actually supports a real change.

In practice, that means Prophet tests dozens of candidate points along a product’s sales history and asks, at each one: does the slope before this point look meaningfully different from the slope after it? Most candidates get rejected — the model is deliberately conservative, built to use as few changepoints as possible rather than overfitting every wiggle in the data as a “trend change.” The ones that survive are the moments that actually mattered.

Catching a Decline Before It Shows Up in Your Bank Account

The most immediately useful application of changepoint detection is catching products in decline early. A product that’s been a reliable seller for months can start losing momentum for any number of reasons — seasonal shift, a competitor undercutting price, an ad campaign losing efficiency, a trend simply running its course. None of these show up as a dramatic single-day cliff. They show up as a gradual slope change that’s easy to miss if you’re only glancing at total sales numbers.

This is exactly the blind spot most WooCommerce store owners operate in. Standard inventory tools — low-stock alerts, reorder thresholds, back-in-stock notifications — are all reactive by design. They tell you when a number crosses a line you set in advance. None of them are built to notice that the underlying behavior driving that number has already shifted. By the time a low-stock alert fires on a declining product, you’re often looking at inventory you no longer need to reorder at all.

WhoKnew the Future™ AI flags declining products through changepoint and growth trend analysis days before the sales drop becomes obvious in raw numbers — giving you the option to launch a clearance sale, reallocate ad spend, or simply stop reordering before the dead inventory piles up.

Confirming a Spike Is Real — Not Just a Good Tuesday

Changepoint detection works in both directions, and the upside case is just as valuable as the downside one. A single big sales day can be a real trend starting — or it can be a fluke: a one-off mention, an email blast, a coincidence of timing that won’t repeat. Reacting to every spike as if it’s the start of something is how store owners burn ad budget chasing noise.

The same mechanism that flags a genuine decline also distinguishes a real upward changepoint from a single noisy day. If the slope before a point and the slope after it both show a sustained, statistically meaningful shift — not just one unusual day sandwiched between two normal ones — that’s a signal worth acting on. Independent analysis of Prophet’s changepoint mechanics confirms this is precisely the point of the sparse-prior approach: the model is built to ignore noise and surface only the moments where the underlying pattern genuinely changed.

For a WooCommerce store, that distinction is the difference between confidently increasing ad spend on a product that’s genuinely trending and burning budget chasing a one-day blip that was never going anywhere.

How This Plays Out Inside WhoKnew Views™

WhoKnew the Future™ AI — included in WhoKnew Views™ Pro — runs real Meta Prophet forecasting directly against your own WooCommerce sales and view data, with no third-party service and no data leaving your server. Changepoint detection sits alongside growth trend analysis, peak traffic detection, and new-versus-returning visitor tracking as one of several signals the system surfaces automatically.

• 1
  Data collectionWhoKnew Views™ tracks product page views and WooCommerce sales data continuously, building a detailed history with bot traffic and team views filtered out automatically.
• 2
  Trend modelingProphet decomposes each product’s history into trend, seasonality, and changepoint components — the same decomposition Meta uses internally at billions-of-users scale.
• 3
  Changepoint testingThe algorithm tests candidate points across the history and statistically confirms which ones represent a real shift in trajectory, discarding the rest as noise.
• 4
  Alert deliveryWhen a confirmed changepoint is detected — up or down — WhoKnew Views™ Pro surfaces it through Growth and Changepoint Alerts, so you find out at the moment it happened, not weeks later.

Why This Matters More Than It Sounds Like It Should

Most ecommerce forecasting advice talks about averages and totals — last month’s sales, this quarter’s growth rate, year-over-year comparisons. Those numbers tell you where you’ve been. Changepoint detection is one of the few tools that’s specifically built to answer a different question: is the thing that’s been true about this product still true right now, or did it just stop being true and nobody’s noticed yet?

For a store running on tight margins and limited ad budget, that’s not a nice-to-have. It’s the difference between catching a decline while you can still pivot and discovering it after the inventory is already dead weight — and between confidently doubling down on a real trend versus chasing a Tuesday that meant nothing.

Who knew the moment your bestseller started dying and the moment you’d notice were ever supposed to be the same day — and that they don’t have to be anymore.