Back to blog

October 5, 2026

Why We Almost Sent Reminder Emails Counting From the Wrong Day

We were building an automated reminder for a client's business. Simple idea: if a new customer signs up but never finishes setting up their account, send a friendly nudge after a week, then another one if they're still stuck after two.

It sounds like a five-minute feature. It almost went out counting from the wrong starting point entirely, and the bug would have been invisible until a customer complained.

The problem we were trying to solve

Customers would sign up for the service, but some of them never finished the last step needed to actually start using it. The business wanted a gentle, automatic nudge: "hey, you signed up but haven't finished setup, here's how." A classic job for an AI agent, since it is a repeated, rule-based task that would otherwise eat someone's time every week.

The rule seemed obvious: count the days since signup, and if a customer passes 7 days with no activity, send the first reminder.

The quiet bug: the clock started on the wrong day

Here's what we missed at first. Signing up and actually activating the account are two different events, and they do not always happen on the same day. Some customers signed up right away and activated within minutes. Others signed up, got busy, and didn't come back to finish activating for two or three weeks.

If the reminder clock starts on signup day, a customer who activates late gets hit with a "you haven't finished setup" reminder the moment they finally do finish, because the system never knew they had just caught up. The reminder would arrive at exactly the wrong time, to a customer who had just done the thing the reminder was asking them to do.

We caught this before it reached a real customer, by testing the system against a case we had specifically set up to be awkward: a customer who signed up weeks before actually activating. The reminder logic flagged them as overdue. That is the kind of result that only shows up if you deliberately test the edge case, not the easy one.

The fix was to add a separate marker for the moment of actual activation, and count the reminder clock from there instead of from signup day. It sounds small. It is the difference between a system that is helpful and one that quietly annoys the very customers it is supposed to be nudging kindly.

Why this is an easy mistake to make

This kind of bug is easy to miss because the two dates, signup and activation, are often the same day for most customers, so the mistake does not show up in normal testing. It only shows up for the customers who are already behind, which is exactly the group the feature was built for. The bug hides itself in the cases that matter most.

Anytime you build a reminder, a follow-up, or an alert that triggers after a period of silence, ask the same question: silence since what, exactly? The moment you signed them up, or the moment they actually did the thing the reminder is measuring? Those are not always the same date, and picking the wrong one means your automation is quietly working against the people it is supposed to help.

The second thing we caught: not all "off" switches mean the same thing

While fixing the first issue, we found a second one sitting right next to it. The business already had a setting that let customers turn off task notifications they didn't want. The early version of the reminder system checked that same setting before sending a setup reminder.

That sounds sensible until you think about who actually uses it. A customer turns off task notifications because they don't want daily task alerts cluttering their inbox. They are not saying "please also never tell me if I forgot to finish setting up my account." Those are two different kinds of message, and treating them as the same setting meant customers who had muted one kind of email would also silently miss the other, with no way to tell that was happening.

The fix was to separate the two. One setting controls ongoing task notifications. A different, independent flag controls setup reminders, and customers who are already fully set up are simply marked as exempt from the reminder entirely, rather than relying on a setting that was never built with reminders in mind.

What this means for your business

If you are building, or have had someone build, any kind of automatic follow-up message, a reminder, a "you haven't responded" nudge, a renewal notice, two questions are worth asking before it goes live.

First: does the clock start from an event the customer actually controlled, or from an internal date that might not line up with their real timeline? Second: if a customer has an "unsubscribe" or "turn off notifications" setting somewhere, does your reminder logic quietly check that setting too, even though the customer meant something narrower when they flipped it?

Neither of these mistakes shows up by reading the code. They show up by testing the awkward cases on purpose, a customer who was slow to act, a customer who muted something unrelated, and watching what the system actually does.

How we catch this before it reaches a customer

Every automated system we build gets tested against the cases that are easy to overlook, not just the ones that work the first time. That discipline is part of what an Automation Audit looks for before anything goes live in your business: not just whether an idea sounds reasonable, but whether the timing, triggers, and settings actually behave the way a customer would expect.

If you want to see what that looks like once it's running, take a look at the agents we've already built, or get a free audit on a process in your own business before you turn a reminder loose on real customers.