Madrid restaurants without a website: the district data

Of 975 restaurants inside the 21 districts of the Madrid municipality, 275 had an empty website field. That is 28.2% of the sample, with a 95% confidence interval of 25.5% to 31.1%.

The city-wide figure is the least useful number in the study. What matters is that the district predicts the gap far better than the city does: fifty points separate Villaverde from Chamartín.

Download the district data, the category data and the review-band data.

What counts as “no website”

An empty website field is counted as missing. When the field is filled, we sort the destination into three groups:

Link destination	Listings	Share
No link	275	28.2%
Social or messaging	54	5.5%
Platform or hosted profile	5	0.5%
Own domain	641	65.7%

Add the first three and 334 restaurants (34.3%) have no site they control. The platform group came out far smaller than expected: only five listings pointed at Glovo, Just Eat, TheFork or similar. In Madrid, delivery is handled inside those apps rather than linked from the listing.

Two limits of the rule, which do not go away by restating them:

• A restaurant may have a site and never have connected it to the listing.
• A filled field does not prove the site works, is current, or belongs to the business.

The district matters more than the city

District	Listings	No link	95% CI
Villaverde	76	51.3%	40.3 – 62.2
Villa de Vallecas	60	50.0%	37.7 – 62.3
Carabanchel	52	48.1%	35.1 – 61.3
Latina	76	47.4%	36.5 – 58.4
Puente de Vallecas	33	45.5%	29.8 – 62.0
Vicálvaro	29	44.8%	28.4 – 62.5
Fuencarral-El Pardo	131	25.2%	18.5 – 33.3
Hortaleza	63	20.6%	12.5 – 32.2
Moncloa-Aravaca	92	15.2%	9.3 – 23.9
Salamanca	35	8.6%	3.0 – 22.4
Centro	75	6.7%	2.9 – 14.7
Chamberí	33	3.0%	0.5 – 15.3
Chamartín	22	0.0%	0.0 – 14.9

The CSV carries all 21 districts and flags the ones below 30 listings. Every row publishes its interval precisely because several rows are small: Usera is 14 listings, and its 42.9% could sit anywhere between 21% and 67%.

Even with that caution, the extremes do not touch. Villaverde's interval starts at 40.3% and Chamartín's ends at 14.9%. The difference is not sampling noise.

The axis of the resulting map follows the north–south income divide that the city government documents in its own open data portal. The southern and western districts hold the gap; the centre and north-east are close to saturated with owned domains. Anyone selling web design in Madrid is not looking at one market but at two.

Venue type moves the number too

Google category	Listings	No link
Bar	20	65.0%
Bar & grill	84	50.0%
Tapas bar	19	36.8%
Restaurant	456	31.8%
Peruvian restaurant	25	24.0%
Spanish restaurant	31	12.9%
Hamburger restaurant	13	7.7%
Mexican restaurant	14	7.1%
Italian restaurant	28	3.6%
Mediterranean restaurant	24	0.0%

The categories are Google's, not ours, and 105 distinct ones appear in the sample. The table holds the ten with at least 12 listings, covering 714 of the 975 records.

Restaurant is Google's catch-all and takes nearly half the sample, so its 31.8% says little. The edges are what informs: venues labelled as a bar carry double the gap of venues labelled with a specific cuisine. A listing that ever acquired a cuisine category is usually a listing somebody manages.

More reviews, smaller gap

Reviews	Listings	No link	Own domain
0	11	36.4%	54.5%
1 – 49	56	75.0%	19.6%
50 – 199	124	52.4%	37.1%
200 – 999	385	33.8%	57.4%
1000 or more	399	8.5%	89.5%

Past the first review the relationship is close to linear. A restaurant under fifty reviews has a three-in-four chance of linking no site at all; one over a thousand reviews, less than one in ten.

Why 28.2% is a floor, not a ceiling

The sample's median review count is 781, and 784 of the 975 records clear 200. This is a sample of visible restaurants.

That is not an accident. Google ranks local results by relevance and prominence, so any area-by-area extraction picks up the prominent ones first. And the table above demonstrates, from our own data, which way that bias pushes: more prominence, more website.

So 28.2% is the lowest defensible value. The true municipal figure, counting the corner bar with no reviews, is higher. The same holds for each district — and because the bias acts equally everywhere, the comparison between districts survives even though the absolute level runs short.

Turning the gap into a responsible list

An empty field is a qualification signal, not permission to make contact:

1. Pick one district, not “Madrid”.
2. Filter to the records with no web link.
3. Drop closed, duplicate and off-target listings.
4. Check the public phone number and some sign of recent activity.
5. Verify the best candidates by hand before writing.
6. Keep the source, the date and the objection status of every record.

The guide on finding businesses without websites covers the list step, and the playbook on selling web design to local businesses covers the commercial one. For international context, the 3,147-listing study compares twenty countries.

At basedonb.com you can reproduce the extraction district by district and filter to the listings with no site. A new account includes 100 free credits every month, no card; paid plans are charged on day one and carry a 7-day money-back guarantee on the value of unused credits.

Methodology

The snapshot was taken on 23 August 2026 with a single query, restaurante, against Google Maps.

The area is the Madrid municipality, not the wider region. We took the 21 OpenStreetMap administrative relations at level 9, merged their polygons, and checked the result: 603.6 km² against the official 604.3 km². Every listing was assigned to a district from its coordinates with a point-in-polygon test.

That step is not optional. The listing's city field returns Leganés and Getafe for “Madrid”, because it names the province. Attribution is only reliable by coordinates.

The extraction returned 2,150 raw listings. We dropped 1,175 for falling outside the municipality — Las Rozas, Torrelodones, Alcorcón and the rest — leaving 975. The run completed 1,505 of 3,008 grid cells, so this is a sample, not a census. Links were not opened or tested. No individual listing is published: the CSVs are aggregates.

Google's own documentation explains how the website field on a listing is edited.