The bug that isn’t a bug
A few years into running my own numbers, I chased what looked like a real drop in signups on a Tuesday. Checked the ad spend, checked the landing page, checked for an outage. Nothing had changed. What had changed was the clock. My analytics tool was bucketing days in UTC, my Stripe dashboard was bucketing them in US Eastern, and a chunk of my traffic came from customers in Singapore and Australia whose Tuesday started while my Monday was still open. The “drop” was several hours of real signups that got filed under the wrong day in one of the two tools. Nothing was broken. The two systems just didn’t agree on when a day starts.
This is one of those problems that doesn’t announce itself. It doesn’t throw an error. It just makes your daily and weekly numbers a little bit wrong, consistently, in ways that are easy to mistake for something meaningful happening in your business.
Every tool picks a timezone, whether you notice or not
Any system that reports “today,” “this week,” or “daily active users” has to draw a line somewhere and say a day ends here and the next one begins. That line is a timezone decision, made either by you or, more often, by a default setting you never looked at.
Common defaults, in my experience:
- Payment processors like Stripe often report in UTC unless you set an account timezone.
- Web analytics tools frequently default to the timezone of the property owner’s account, or to UTC on the raw export.
- Your own database timestamps are almost certainly stored in UTC, because that’s the sane engineering default.
- Spreadsheets and BI tools inherit whatever timezone was baked into the export, and that setting rarely travels with the file.
If you pull “yesterday’s revenue” from Stripe and “yesterday’s signups” from your analytics tool and they’re on different clocks, you’re not comparing yesterday to yesterday. You’re comparing two overlapping but different 24-hour windows. For a small business, that mismatch might be a few hours wide. It rarely makes a good day look bad. It is often enough to make a flat day look like it moved, or a real dip look bigger or smaller than it actually was.
Where this shows up in metrics you already track
Churn and MRR. If your billing system logs a cancellation at 11:40pm UTC, and your local reporting day runs on US Pacific time, that cancellation lands on what you’d call “yesterday” in your own head but “today” in the report. Over a month this evens out somewhat, but around month-end and quarter-end, when you’re comparing MRR snapshots taken at a specific moment, a few hours of drift can move a customer’s cancellation into the wrong billing period entirely. I’ve seen this nudge a monthly churn count up or down by one or two customers, small in absolute terms but enough to throw off a churn-rate comparison if you’re eyeballing it month over month without checking the boundary.
Cohort analysis. Cohorts are built on “customers who signed up on day X.” If your signup timestamp is stored in UTC but your cohort report groups by local calendar day, customers who signed up in the last few hours of a UTC day can get sorted into what feels like the wrong cohort relative to your local sense of when they joined. It won’t invalidate a cohort study, but it can blur the edges of week-over-week or day-over-day cohort comparisons, especially for a small list where a handful of misplaced signups is a noticeable percentage.
A/B testing. Tests are usually run for a fixed number of days, and traffic isn’t evenly spread across a 24-hour clock, it clusters around when your audience is awake. If your test’s “day” boundary falls in the middle of your busiest traffic hours because of a timezone mismatch between your testing tool and your audience’s actual clock, you can end up splitting a single behavioral peak across two reporting days. That doesn’t invalidate the test result itself, since the randomization still happened correctly at the visitor level, but it can make daily breakdowns of the test misleading if you’re trying to spot which day of the week performs best.
Funnels and attribution. Multi-step funnels (visit, signup, trial, purchase) often span more than one calendar day. If the tool tracking the top of the funnel and the tool tracking the purchase event use different timezones, a same-day conversion can appear to be a next-day conversion in one report and a same-day conversion in another. Attribution windows (like “converted within 7 days of first click”) are usually defined in exact hours, so this matters less there, but any report that buckets by calendar day is exposed to it.
Daylight saving time makes it worse, twice a year
Even if you’ve picked one consistent timezone for all your reporting, DST transitions still cause a real, if temporary, distortion. The day the clocks change, one calendar day is either 23 or 25 hours long in local time. A daily report that assumes every day is 24 hours will show an artificial dip or bump purely from the missing or extra hour, nothing to do with customer behavior. If you’re in a region that doesn’t observe DST (Singapore doesn’t), it’s easy to forget this is even happening to your US or EU customers, and to misread the resulting wobble in your numbers as something real.
What actually helps
None of this has a single fix that makes the problem disappear, but a few habits cut the confusion down a lot:
Pick one reporting timezone and use it everywhere you can. UTC is the least ambiguous choice because it doesn’t observe DST and doesn’t imply a location. If your business and most of your customers sit in one region, your local timezone can be more intuitive to read day to day, but be consistent about which one you use across every tool, not just the one that happens to be convenient.
Check the timezone setting in every tool before comparing across them. Stripe, your analytics platform, and your database each have a place this is configured. It takes a few minutes to check and it’s the single most useful thing you can do here.
Treat single-day dips or spikes with more suspicion than week-over-week trends. A timezone mismatch shows up most sharply at the daily level and washes out over a week. If a number looks off for exactly one day and nowhere else, a clock mismatch is worth ruling out before you assume something changed in your business.
Note DST transition dates on your own calendar if you serve customers who observe it. You don’t need to model the effect precisely, just recognize the date so you don’t overreact to a one-day artifact.
When in doubt, look at the raw timestamps, not the report. A daily rollup hides the exact time an event happened. Pulling a handful of raw records around a suspicious boundary is usually enough to tell you whether you’re looking at a real change or a bucketing artifact.
None of this makes your reporting timezone-proof. It just means that when a number moves, you can rule out the clock before you start rewriting your story about what happened in your business.
If you want more of this kind of plain-English walkthrough on the metrics and methods behind running a one-person business on real numbers, you can find the rest of what we’ve written here.
Get new guides and videos first — join the Telegram channel.