Evolution has begun in digital ecosystems

Evolution has begun in digital ecosystems

[ad_1]

Most of the large Russian digital ecosystems by the end of 2023 had significantly updated themselves, increasing the number of services, according to a Spektr study. Basically, operators, regardless of their core business, invested in entertainment and media services. Experts believe that in the future, ecosystems whose owners will integrate their own neural network models will benefit – so far only Yandex and Sberbank have them.

Kommersant got acquainted with a study by the analytical company Spektr on the state of the largest Russian digital ecosystems (Sbera, Yandex, VK, MTS, Tinkoff) by the beginning of 2024. According to analysts, operators significantly updated the composition of ecosystems in 2022–2023: in total, they closed 37 services (15% of the original number) and opened 67 new ones (an increase of 28%). VK has closed the most ecosystem services in two years (15), but is also the leader in opening new ones (20). In relative terms, the greatest increase in the filling of the ecosystem was shown by MTS (an increase of 43% compared to the beginning of 2022). Only Sber saw a decrease in the total number of services, but only by 2%.

If at the beginning of the history of the formation of Russian digital ecosystems their owners sought to strengthen their presence in the maximum number of verticals (for example, in the delivery and job search sectors), then after 2022 they began to “exit problematic projects – for example, with weak positions or excessive costs,” they note in Spektr. Among the failures, analysts name VK Combo, Citymobil (owned by a joint venture between Sberbank and VK), and in the case of Yandex, the previously owned Novosti and Zen (now owned by VK).

Ecosystem operators are primarily strengthening the areas of media and entertainment. Only in them all market participants increased the number of services and seriously updated existing ones.

MTS confirmed the growing popularity of entertainment services (Kion, MTS Music, Strings), noting also the anti-spam and anti-fraud product “Defender”. “Each of these products individually is growing by more than 30–50% annually, both in money and in users,” Olga Ziborova, the company’s vice president for ecosystem development and marketing, explained to Kommersant.

Yandex notes an increase in the audience of entertainment services and the time they use in the Plus subscription: “On average, subscribers spend 21 days a month in the services.”

VK only clarified that they are developing “products and services that the audience needs.” Sberbank and Tinkoff Bank did not respond to Kommersant’s requests.

Among the technologies around which ecosystems will develop in the future, Spektr primarily names neural network language models (LLM and GPT).

We are talking about their end-to-end use in different services. Similarly, YandexGPT 2 has been introduced into Yandex services, analysts explain, but “so far, the low quality of the model itself remains an obstacle to their popularization.”

The end-to-end implementation of neural network technologies requires more serious costs than, for example, simply combining different services into one subscription, notes Larisa Malkova, managing director of the Data and Applied AI practice at Axenix: “But the effect will be much larger in terms of efficiency, relevance and advancement of these services.” According to her, all large ecosystems will inevitably do this, but “not everyone will have enough resources to develop basic language models (Yandex and Sber currently have them.”— “Kommersant”) — in fact, this is the creation of our own infrastructure.”

At the same time, the head of the Content Review agency, Sergei Polovnikov, emphasizes that “in the moment” it is easier and more efficient for ecosystem operators to invest in specific “everyday use services.” Moreover, first of all, the same content ones, the expert clarifies, – the development, for example, of e-commerce requires serious expenses on infrastructure and logistics.

Yuri Litvinenko

[ad_2]

Source link