Smart Timing is deliberately cautious: it changes your pacing only when it is at least 90% sure the change is better and expects at least a 3% lift in pick-ups. Most "it isn't doing anything" reports are it doing exactly that. Here's how to tell the difference.
The card is on the campaign's page (Campaigns, then open the campaign). It starts collapsed as a single band with the status chip (Learning, Ready to apply, Optimizing or Off); click it to see the detail. Smart Timing's switches are in the ⋮ menu → Campaign settings.
It's stuck in "Learning"
Work down this list.
1. Not enough calls yet. Read the line under the headline: Learning from N calls is the practice-wide count across your AI campaigns. With a few hundred calls or fewer, the starting point still carries a lot of weight, so it will rarely be confident enough to apply. The progress bar and "About N more calls until it can show …" tell you how far it is to the next level of detail (time of day, then hour of day, then weekday by time of day, then day and hour). There is no shortcut except more calls.
2. The pattern is real but small. If your patients pick up about as well at 10 a.m. as at 3 p.m., there is nothing to optimize, and Smart Timing correctly does nothing. The status line shows its confidence ("Confidence so far: 62%", or "62% so far" when it applies itself). A number that hovers rather than climbs usually means a flat pattern, not missing data.
3. Your window only has one narrow block. If the campaign calls 10 a.m.–12 p.m. only, there are two hours to choose between. Smart Timing only re-balances inside your windows; it never adds hours. Widen the block or add a second one in Campaign settings and it will learn inside the larger fence.
4. Apply automatically is off. In that mode the chip parks at Ready to apply and waits for you; the card shows the expected lift and how sure it is. Press Apply on the card. To let it apply on its own, turn Apply automatically when ready on in Campaign settings.
5. You pressed Undo recently. An Undo (or any apply) starts a one-week pause before Smart Timing applies automatically again, so it can't immediately redo what you just reversed. Show changes on the card lists each change with its date and time.
6. The campaign is in staff mode. Staff campaigns have no auto-dialer, so there is nothing to pace and the Smart Timing card does not appear. Only AI-agent campaigns learn.
7. Auto-apply is paused. If Smart Timing had to undo its own change three times, the card says auto-apply is paused and a ready change waits at Ready to apply. To restart it, press Apply when it shows, or turn Apply automatically when ready off in Campaign settings, save, then turn it back on.
The best times look generic
If the top three picks show estimate instead of "N of your calls", you are looking at the starting point, not your own data. The note near the bottom of the card says where it came from: "Started from what works across N other … practices" (practices in your specialty), "Started from what works across FrontDesk practices", or "Started from industry defaults (midweek, late morning and late afternoon)." That is expected until your own call count grows.
If you have plenty of calls and it still looks generic, check that they are landing where you think:
- Timezones. Everything is measured in each patient's local time ("In each person's local time" under the grid). A list spanning several states spreads the same hour in the campaign's calling time zone across several patient-hours, which flattens the picture. Nothing to fix, but it explains a softer pattern.
- A short window. With a 9-to-5 block the grid can only light up 9-to-5. Faded cells are outside your windows and never get calls.
An applied schedule feels wrong
- Check the observed lift first. After an apply, the card shows "measuring the real effect (N of 100 control calls)". Once enough calls have gone out on both sides, it shows the measured difference ("+9% observed vs. the control slice"). A positive number means it is working even if the schedule looks unfamiliar.
- Press Undo. Pacing goes back to what it was before the last apply (even pacing, if that was the first apply). Nothing else changes.
- Look at the numbers. Hover the grid cells it leaned into and away from. Each shows the pick-up rate, how many of your calls back it, and the pacing multiplier. A cell with three calls and a 67% rate is noise; Smart Timing weighs that appropriately, but the raw number can look alarming.
- Check the log. Show changes records what was applied or undone, when, and whether it was automatic.
- Turn it off with Turn off on the expanded card, or in Campaign settings, if you'd rather dial until the daily limit is reached. Your call history is kept for when you switch it back on.
The grid is empty or all one shade
- No grid at all. The hour-by-hour grid appears once calls have gone out; before that the card says the picks are estimates.
- All one shade. Every open hour has about the same estimated rate (a flat pattern, or early days when the starting point dominates). Normal.
- Everything faded. The campaign has no open hours between 8 a.m. and 8 p.m. on the days it calls. Check the calling days and blocks in Campaign settings.
- The card doesn't update. Expand the card and press the refresh icon to re-run the analysis now. If it keeps failing, contact support with the campaign name.
Callbacks aren't being counted
A callback is credited when the same phone number calls the practice within three days (72 hours) of an AI campaign call. It won't be counted if the patient called from a different number, called after that window, or called a number that isn't on FrontDesk. The "callbacks credited" figure on the card is practice-wide.
Still stuck?
Contact support with the campaign name and a screenshot of the expanded Smart Timing card. The status chip, call count and confidence line are usually enough to diagnose it.