Reading who's slipping away: churn risk and predicted next order

What this is for

Every week you lose customers you could have kept. Usually not because of the product, but because nobody reached out at the right moment, and by the time it was obvious, they were already gone.

Churn risk and predicted next order, read together, give you a ranked list of exactly who to contact this week and who to leave alone. Here's how I read it, because getting this right is one of the highest-return habits you can build.

---

What churn risk is actually measuring

This isn't a flat "hasn't bought in 90 days" rule. SendLens scores churn against each customer's own rhythm, and that distinction is the whole point.

The logic is simple. Take how many days it's been since their last order, divide by how long they normally go between orders, and that ratio tells the story:

• Someone who reorders every 30 days and it's been 35? Barely worth noting.
• Someone who reorders every 30 days and it's been 75? That's a real problem, even though 75 days sounds fine on its own.

Same 75 days, different meaning, because we grade against their normal instead of a fixed number. SendLens turns that ratio into four levels:

The High / At Risk bucket is where you focus. Low is noise, Lost is a lost cause, and the recoverable dollars sit in that overdue-but-not-gone middle. That's your worklist.

---

One setting decides whether any of this is true

Don't skip this. For customers who've only ordered once, there's no personal rhythm to measure, so SendLens falls back to your expected purchase cycle. The default is 90 days. If you never set it, those scores are being graded against a guess.

Go into Settings and set that number to your real reorder window. A consumables brand might be 30 to 45 days. A big-ticket brand might be 180-plus. Get this right first, or you'll chase people who aren't actually late and miss people who are. It takes a few minutes and it makes every churn number on the screen trustworthy. Do it before anything else.

---

Now bring in predicted next order

Churn risk tells you who's late. Predicted next order tells you when each person is due and, just as important, how much to trust that prediction. Those two columns together are what make this useful.

Here's how the prediction works so you read it correctly. It needs at least two orders to see a pattern. We take the average gap between their orders, add it to their last order date, and that's the predicted date. Then confidence is scored on how consistent they are and how many orders back it up:

• A customer who orders every four weeks, ten times running, gets high confidence. When SendLens says they're overdue, believe it.
• A customer with exactly two orders gets their confidence held down on purpose, because a single gap can't prove a pattern. It takes around five or more orders before confidence can get near the top.

So read the two columns as a grid:

• Overdue and high confidence: a reliable repeat buyer just broke their pattern. Top of the list, every time.
• Overdue and low confidence: could be a fluke, or a two-time buyer we're not sure about yet. Worth a light touch, not an expensive win-back.
• Upcoming and high confidence: don't discount these people. They're about to buy anyway, so a reminder is plenty.
• Upcoming and low confidence: background noise. Ignore for now.

The point is that confidence keeps you from wasting discounts. Anyone can blast the overdue list. The better approach is to spend real incentive dollars only where the prediction is solid.

---

What healthy looks like, what should worry you

Read these as ranges, since every brand is different, but here's what I look for:

Looking good:

• Your At Risk bucket is a manageable slice you can work through in a week, not a mountain.
• Most of your predicted-overdue customers are low confidence (two-time buyers), not your proven regulars.
• The high-confidence, upcoming group is full. That's a stable, predictable revenue base.

Worth worrying about:

• Your best customers showing up as overdue with high confidence. This is the one to act on fast. Reliable repeat buyers breaking their pattern is an early sign something's off, whether a product issue, a competitor, or a broken flow.
• The At Risk pile growing week over week. You're acquiring fine but retention has a hole in it. Fixing the flow that should catch these people beats chasing them by hand forever.
• Almost nobody has enough order history to predict. That's a repeat-purchase problem, which is the subject of the next guide.

---

Build the worklist, then work it

This is a weekly habit, not a one-time thing. Each week, read the screen in this order:

1. Overdue and high confidence, sorted by value. Proven, reliable buyers who just went quiet. Best return on your time, because you're reminding a regular, not convincing a stranger.
2. The rest of At Risk, highest value first. Overdue, worth real money, still reachable. Work down the list as far as your week allows.
3. Overdue and low confidence. A light, cheap touch. No deep discounts on people who may just be occasional buyers.
4. Skip Lost. Past 3x their cadence, they're gone. One automated attempt if you want, then move on.
5. Leave the upcoming group alone, aside from a no-discount reminder for the high-confidence ones.

Then make it real. Pull the customers you decided to chase out of SendLens, push them to a Klaviyo list or export the CSV, and send a win-back flow to that group: a reminder, then a reason, then a last call. Don't send it to the whole database. The tight targeting is the advantage.

---

Check yourself next week

You worked the list. Come back and read it again:

• Did the people you contacted move from At Risk back to Low or Medium? That's a recovered customer.
• Is your At Risk bucket shrinking, or at least holding while your customer count grows? Retention is improving.
• Are your high-value regulars staying in the upcoming bucket instead of sliding into overdue? Your base is stable.
• Is the win-back flow actually pulling revenue? If people open it and buy, keep going. If not, the offer is the problem, not the list.

Do this every week and you'll catch most of these customers while there's still time to keep them.