Statistical study · open data · pre-registered protocol
The digital presence of Taranto's tourist-facing businesses
A snapshot on the eve of the XX Mediterranean Games: how many of the businesses a visitor ought to be able to find — restaurants, places to stay, shops — actually exist on the web, and in what state.
In brief
63.1%
95% CI: 61.1–65.1
have no traceable web presence at all
no site, and no social page recorded in the open sources
24.0%
95% CI: 22.2–25.8
have a website of their own
2.8% are on social media only, 2.7% have a parked site
20.3%
95% CI: 17.7–23.2
of the declared addresses are dead
one site in five, among those declaring a site at all
33.8%
of the sites offer content in English
38.7% in accommodation — two months before the Games
Performance on mobile devices is critical: a median Lighthouse score of 0.59 and a median largest-contentful-paint of 9.5 seconds. 23.3% of live sites show at least one sign of abandonment.
Download the report
The document is deposited open access with a permanent DOI. This page hosts a copy; the citable, versioned record is the deposited one.
A methodological result: no single source is enough
The census merges three open sources — Overture Maps, OpenStreetMap and the Puglia Region's DMS registers. They almost never overlap: only 24 businesses out of 3,021 appear in all three, and each source on its own brings hundreds of businesses the others ignore.
And where two sources describe the same business, in 49.5% of cases one declares a website and the other does not. Half the time, on a binary field.
Any picture of the «local web» built on a single source measures the source at least as much as the territory — with no internal signal that anything is amiss. It is the part of this work that holds beyond Taranto.
Chains and independents
Businesses belonging to chains have a web presence in 71.5% of cases, against 26.2% for independents (odds ratio 7.1; 95% CI: 4.9–10.1). But the reading inverts once you look at what that presence is: 84% of chain sites are the brand's national site, not a page for the Taranto outlet. Taranto's local web is almost entirely the work of independents — and three in four do not have one.
The association is observational and implies no causal link. The sector perimeter excludes by construction some fields with a high density of chains: it is a documented choice in the protocol, and it affects this comparison.
Differences by sector
| Sector | Web presence | 95% CI |
|---|---|---|
| Retail (tourist-facing) | 43.7% | 40.1–47.4 |
| Food and restaurants | 25.1% | 22.1–28.5 |
| Accommodation | 20.9% | 18.3–23.8 |
The territorial differences between Città Vecchia (35.4%), Borgo (30.5%) and the rest of the municipality (38.8%) exist but are weak. Taranto's digital divide runs by sector, not by district.
How it was done
Research questions, inclusion criteria, thresholds and analysis plan were frozen on 7 July 2026, before a single data point was collected; every subsequent change is dated and justified in a changelog, and the instrument defects found during collection are declared in a deviations register rather than quietly corrected. The frame comes from the deduplicated union of three open sources. Collection is passive: three requests per site, a user-agent identified with a link to the privacy notice, robots.txt respected; performance comes from the PageSpeed Insights API. Every estimate carries its confidence interval; missing data is declared and never imputed. The full protocol is deposited alongside the report: every number traces back to the frozen rule that produced it.
Limitations
- The frame covers businesses traceable in the open sources: «no web presence» means «no site declared in the sources», not necessarily «no site in existence».
- The completeness of the frame is not identifiable under this design, and its precision was not manually validated: the verification sample the protocol provided for was dropped, by a documented decision.
- One city, one collection round, no comparison with other territories measured with the same instrument.
- The language detector and the abandonment signals are unvalidated heuristics, limited to the homepage: the resulting rates carry an unquantified error.
- Observational associations: the design supports no causal claim.
Conflict of interest and purpose lock
Whoever ran this study runs a web agency, and therefore measured the market they work in. That is why the protocol was frozen before collection, why the limitations are published alongside the results, and why the report is descriptive and contains no recommendations. The study is self-funded and was commissioned by nobody.
The data is not used to contact the businesses surveyed. It is a binding clause of the protocol and a commitment made in the privacy notice published before collection: no business is named or identifiable in any output, all published data is aggregated with a minimum threshold of three units per cell, and the identifier archive was deleted on 17 July 2026 once extraction was complete.
How to cite
Moerynck, N. (2026). The Digital Presence of Tourist-Facing Businesses in Taranto: a Pre-Registered Open-Data Census Ahead of the XX Mediterranean Games (v1.0) [Report]. Zenodo. https://doi.org/10.5281/zenodo.21834633
Processed from ISTAT data (CC BY 4.0), © OpenStreetMap contributors (ODbL), Regione Puglia — DMS Puglia (CC BY 4.0), Overture Maps Foundation (per-record licences). No Google data was used as a directory source; the PageSpeed Insights API is used purely as a measuring instrument.
← Back to the site