How often marketing data should refresh.

Clients ask for more frequent data when a report has lost their trust. The reasoning is that fresher numbers are more correct, so hourly must be better than daily.

It is worth separating the two. Freshness is how recently the data arrived. Correctness is whether it is right. Increasing the first does nothing for the second, and in marketing it often makes it worse.

Four reasons, and the first one is the one that matters.

• The platforms have not finished counting

  Conversions arrive late and invalid clicks get refunded, so a platform revises figures for days that have already closed. An hourly figure for today is a partial count of an incomplete day, and it will change.

• It produces decisions on noise

  A report that updates hourly gets checked hourly, and an hour of ad performance is mostly variance. Teams respond to it anyway, which is worse than not looking.

• Rate limits

  Platform APIs cap how often you can ask. Pulling more often means pulling less each time or getting throttled, and a throttled sync is a sync that quietly returns partial data.

• Cost

  Every refresh is queries against your warehouse. Twenty-four times the syncs is roughly twenty-four times that cost, for a number that is less settled than yesterday's.

Some cases are real, and they have a common shape: a decision that must be made today and cannot wait until tomorrow's numbers.

A launch or a sale where you are watching whether something is broken. A budget cap that could be hit before the day ends. An outage where you need to know whether spend is still being delivered. All three are operational monitoring rather than reporting, and they are better served by an alert on a threshold than by a dashboard somebody has to keep looking at.

Show the client when the data was last updated, on the report itself. Most of the anxiety about freshness is not knowing, and a timestamp answers it.

Then add a freshness check that alerts you when a sync is late, so a stale report is something you are told about rather than something somebody discovers. That combination solves the actual problem, which was never the interval. It was not being able to tell a quiet day from a broken pipeline.

Syncs run daily for every client, which is the cadence that matches how the platforms themselves settle their figures. A sync adds to the tables rather than replacing them, so when a platform restates a closed day both the original number and the correction stay on record.

Freshness and threshold checks run against those tables and raise an alert when one fails, which covers the operational cases without turning every report into something that has to be watched.