Page 144 - NEWNORMAL_ANIMATED

NEW NORMAL

have a perfect product, but to have the best 6XFFHVV LV QRW ɮQDO
product. They needed to know, how many wi- failure is not fatal:
nes do they need to be relevant for most users.
The number was originally estimated by Heini it is the courage to continue
at 20,000 wines, but it just kept growing. To- that counts
day they have information on 200,000 produ- Winston Churchill
cers, 10 million wines and 1,3 billion photos of
different wine labels in the app.
“But the hardest thing about that is that the wines we recommend have been reviewed
we’re now a 200-person company that sells by friends, their feedback will be highlighted in
wine in 17 countries,” he continued. “Which the Vivino Market feed. The more a user scans
means we’re relatively thin in all these mar- and rates wines, the more Vivino Market learns
kets. So one of the big things here, is actually about their preferences and the stronger the re-
to go much deeper in each market and say, commendations will be.”
okay, we now know it works here, let’s put
more resources in every single market.”
What´s behing the screen? How does it run?

Now Vivino has raised $155 million - a sum
over twice as large as all of its previous fun-
ding to date. Spurred by rapid growth that has
seen its user base grow from 29 million in 2018
to 50 million currently, Vivino wants to use the
large cash injection to significantly boost its
core tech and personalized recommendation
engine, while also expanding its presence in
key growth markets globally.
The potential benefits, then, seem sub-
stantial, for the users who ultimately engage
„Vivino’s machine learning algorithms look at with the Vivino Market and find themselves
the scanning and rating behavior of each indivi- exposed to specific wines that should appeal
dual user to determine wines they are most like- to them, to the retail partners who gain new
ly to be interested in (scans) and love (ratings). potential customers, to Vivino itself, which col-
We also filter out wines they are likely to not like lects a flat commission based on orders that
based on lower ratings they have given,” Zacha- have been shipped. It’s a fascinating conflu-
riassen elaborated. „We then show each indi- ence of social media, machine learning, and
vidual user similar wines that have been highly wine, and it could be the first step in a whole
rated by our 23 million-strong community. If new way to purchase it. ƅ
144