City data, country data, and which one is lying

Audience geography looks like one dataset and is several. Understanding which one you are reading decides whether a decision based on it holds up.

Two granularities, two purposes

City data typically arrives as a list of top cities with a current and sometimes a peak listener figure. It exists for a reason: streaming platforms have a location signal per listener, and city is the finest granularity anyone publishes.

Use it for live decisions. Which rooms, which regions, where the audience is dense enough to fill a venue.

Country data arrives per platform, often through a different endpoint and sometimes with a different method behind it.

Use it for market decisions. Which territories to service, where to prioritise press and radio, which language and timezone assumptions to make, which rights and distribution questions to ask.

The mistake in both directions. Using city data for a market decision over-weights whichever city happens to be measured well. Using country data for routing tells you Germany is strong without telling you whether that means Berlin, Cologne or a diffuse audience with no venue anywhere.

They will not reconcile, and that is fine

Sum a project's top cities and compare against its country totals and the numbers will not match. Three legitimate reasons.

Top cities is a top N, so everything outside it is missing. City attribution fails more often than country attribution, because it needs a finer location signal. And the two figures can come from different endpoints with different windows.

Treat them as two views rather than one dataset, and do not build anything that requires them to agree.

Platforms disagree, informatively

The same project can look concentrated in one country on one platform and spread across several on another.

That is not an error to average away. Platforms have different user bases by region and by demographic, and a divergence tells you something real: which audience is on which platform, and therefore which platform's tools to use to reach a given market.

The one thing to avoid is blending them into a single audience-by-country figure. A follower on one platform, a monthly listener on another and a subscriber on a third are different units, and a weighted average of them means nothing.

Refreshing it daily is enough

A practical engineering note that saves money and latency.

Upstream geography refreshes at most daily, so a stored daily snapshot is exactly as current as a live call. Once a snapshot exists there is no reason to hit a provider on page load: read the latest stored row, and only fall back to a live call when no snapshot exists yet, which happens for a newly connected project before its first sync.

We did that and the page got faster with no loss of accuracy, because there was no accuracy to lose. The general rule is to match your refresh rate to the source's real update rate rather than to how fresh you would like the data to feel.

Reading it well

Current against peak, per market. The single most useful comparison. A market at a third of a peak set two years ago is history, not opportunity.

Share, not count. A market growing as a share of your audience is more informative than one growing in absolute terms while your total grows faster.

Over-index, not rank. Divide your share of a market by that market's expected share. The ranking that produces is the one worth acting on, as covered in finding core markets from your own data.

Wait after a placement. During and immediately after a large placement, geography reflects the playlist's audience. Re-read three weeks later.

What it cannot tell you

It cannot tell you about an individual, and it should not. Aggregate counts by city and country answer every question a manager has about where an audience is, without anyone holding a record of where a person was.

It also cannot tell you why a market is strong. Data locates, people explain. The answer is usually a support slot, a station, a scene or a sync, and finding it takes a phone call rather than another query.