The OECD’s calculation of the market shares is based on a dataset consisting of the number of visits to each website, limited to the relevant geographic region and adjusted using the website’s bounce rate to capture only engaged visits, on a monthly basis over the time period July 2022 to January 2026. The construction of the dataset is explained below, with the methodology varying slightly for Poland as compared to Lithuania and Latvia.
Poland
The OECD extracted from SimilarWeb the following data for all the firms specified in the relevant scenarios in Table A.A.1, for each month in the time period July 2022 to January 2026:
Monthly visits (Poland) – the number of visits to each website within a month from users based in Poland, across desktop and mobile devices. A visit is counted when a user accesses one or more pages during a session, with a session ending once user is inactive for more than 30 minutes (SimilarWeb, 2026[1]).
Bounce rate (Poland) – the proportion of visits to the website from users based in Poland ending after a single page view, without any further interactions (SimilarWeb, 2026[2]).
Then, the bounce rate was used to remove all the visits which ended after a single page view from the total count of monthly visits, leaving only the website’s “engaged visits” as part of the dataset (see Table A A.2. for a more detailed description of the formula used). This adjustment is important as the original count of monthly visits meant that every session was treated equally, including single‑page bounces that often reflect accidental landings, mis-clicks or extremely low-intent visits. In contrast, engaged visits reflect only sessions that continue beyond the user’s landing page, providing a more accurate picture of user engagement and a closer approximation to the number of transactions taking place.
Lithuania and Latvia
As described previously, there is no country filter available for Lithuania and Latvia, as was the case Poland above. As such, the OECD extracted from SimilarWeb the following data for all the firms specified in the relevant scenarios in Table A.A.1, for each month in the time period July 2022 to January 2026:
Monthly visits (worldwide) – total visits to each website within a month, across desktop and mobile devices. A visit is counted when a user accesses one or more pages during a session, with a session ending once user is inactive for more than 30 minutes (SimilarWeb, 2026[1]).
Country traffic share (Lithuania and Latvia) – the proportion of a website’s total worldwide traffic that originates from a specific country.
Bounce rate (worldwide or Poland) – the proportion of visits from either worldwide users or users based in Poland (explained further below) to the website ending after a single page view, without any further interactions (SimilarWeb, 2026[2]).
Next, the country traffic share was used to derive the relevant traffic figures for each country by taking monthly visits (worldwide) and multiplying this by the country traffic share for the same period, to produce an estimate of total visits from that country (i.e. worldwide visits × country’s percentage share = country‑specific visits) (see Table A A.2). This weighting exercise ensures that the OECD’s analysis reflects traffic originating within the relevant national borders.
Then, as explained for Poland above, the bounce rate was used to remove all the visits which ended after a single page view from the total count of monthly visits, leaving only the website’s engaged visits as part of the dataset. For platforms with a lt or.lv domain (which generally had around 90% of their total traffic originating from Lithuania or Latvia, respectively), the worldwide bounce rate was used, while for global platforms (which generally had <0.20% of their total traffic originating from Lithuania or Latvia), the Polish bounce rate was used as the OECD considered it a more appropriate approximation of the behaviour of consumers from a similar regional area (also in light of the common characteristics discussed in Chapter 5).3
