Check the numbers.

A summary is a good start. It shouldn’t be where you stop.

I’m Canadian. I spend most of my time between Latin America and the United States, and the work I do is in eCommerce, payments, and health companies. All three run on numbers that move every day. Orders, refunds, approvals, appointments, invoices. Most of the time someone else has already added them up and put them in a chart. That is helpful. It is also where a lot of small mistakes hide.

So I try to look at the underlying rows before I lean on a summary. Not every time, and not every row. Just enough to know the total came from where I think it did.

This isn’t about distrust. The people who build reports are usually careful. The problem is that every summary makes choices. Which dates count. Which currency. Whether a refund lands on the day it was asked for or the day it went through. Whether a test order got filtered out. Each choice is reasonable on its own. Put a few of them together and two honest reports can disagree, and neither one is wrong.

Time zones make this worse. A day that ends at midnight in one city is still going in another. If one system closes the day in Bogotá time and another closes it in UTC, the totals will never quite match, and someone will lose an afternoon trying to work out why. Asking “which clock is this report on?” has saved me more time than almost any other question.

The habit itself is simple. When a number matters, I open the source and pick a few lines at random. Does this order look like a real order? Is this amount in the currency I assumed? Does the date line up with what the summary says? If three or four rows make sense, I trust the total a lot more. If one doesn’t, I’ve found the thing worth asking about before it turned into a decision.

The other half is saying what I checked. “I looked at ten rows from last week and they match” is more useful than “looks fine.” It tells the next person what was covered and what wasn’t. It also makes it easy for someone to say, “You checked last week, but the problem is this week.”

I don’t always do this. When I’m tired or moving fast, I take the chart at face value, and sometimes that works out. When it doesn’t, the fix usually takes longer than the check would have. That’s the trade I keep relearning.

There’s a kindness in it too. When I ask someone about a number, I want to show up with a specific line and a specific question, not a vague sense that something is off. That respects their work. It turns a hard conversation into a short one.

None of this needs special software. A spreadsheet, a filter, and ten quiet minutes are usually enough. The point isn’t to audit everything. It’s to stay close enough to the real numbers that the summary means what you think it means.