How complete is OpenStreetMap coverage for a district

You scanned a district and got two hundred and forty businesses. On its own that number says nothing. If there are three hundred businesses in the district you have a good list; if there are eight hundred, you do not know what your list represents.

Why the gap matters

Coverage is not trivia. It feeds three decisions directly.

It sets how much time a district is worth: where coverage is low, the same effort reaches fewer businesses. It sets which channel to use: the businesses missing from the map are generally the ones with the weakest online presence, and email reaches them badly. And it tells you something about competition, because nobody has written to a business that isn’t on the map either.

Sampling

You will not find a full count anywhere, but a sample gives you a good enough estimate.

Pick a street, preferably one of the district’s main commercial ones, and a stretch of about five hundred metres. Count the shops on that stretch one by one; street imagery is enough, walking it is better. Then count how many records your list holds for the same stretch.

The ratio between the two is your coverage for that street. Repeat on three separate streets: one central, one on a side road, one in a neighbourhood away from the centre. If the three numbers are close, you have a reasonable estimate for the district. If they scatter, your coverage varies by area, which is useful information in itself.

Where the gaps are systematic

The missing records are not distributed randomly and the pattern is predictable.

New businesses are missing. Getting onto the map generally requires someone to go there and add it. Places less than six months old are usually absent.

Upstairs businesses are missing. Accountants, solicitors, consultancies, small clinics. They are not visible from the street, so they do not make it onto the map either. Coverage in those sectors is markedly lower than for ground-floor shops.

Small settlements are missing. Contributor density falls off with city size. Coverage is high in the central districts of large cities and drops away quickly as you move out.

Closed businesses are still there. That is an error in the other direction: records on the map for places that no longer exist. They make your coverage look better than it is, because they inflate the count.

Tag completeness is a separate measure

Coverage tells you whether a business is on the map. Whether the record is usable depends on which of its fields are filled.

Count separately how many of your records carry a phone number, how many a website, how many opening hours. In local business data, name and location are present on nearly every record, a phone number on a large share, and a website on far fewer.

The two measures only mean something together. High coverage with low tag completeness means a list where you have to find the contact yourself. Low coverage with high completeness means a small list you can work directly.

Measure once and write it down

The estimate takes half an hour for a district, and unless you write the result down you will redo it every time.

Three lines are enough to keep: the district, the estimated coverage, the date of the measurement. Come back a year later and the ratio has probably moved, and the direction carries information. Rising coverage means there are contributors interested in the area; flat coverage means your list will be the same size on the next pass.