--- title: "Setting up product analytics at TVBET" description: "How I set up product analytics at TVBET: get the events tracked, get people reading the numbers, then build product on them." url: https://alisher.ataev.net/notes/tvbet-product-analytics author: Alisher Ataev date: 2026-11-04 updated: 2026-11-04 use_case: "product analytics in iGaming" company: TVBET role: Product Analytics Lead period: Mar 2023 to Feb 2024 keywords: product analytics, LTV, retention, cohort analysis, iGaming, hypothesis testing, recommendation engine, fraud detection --- # Setting up product analytics at TVBET ## 1. Two or three dashboards March 2023. I joined TVBET as a consultant, through my own consultancy, Rubberduck Consulting. Fully remote, with the team in Poland and Malta. TVBET makes live-casino games with a lot of traffic. General goal was simple enough: add product analytics and set it up inside the company. So my first order of business was to see what analytics were actually there. Not what people told me was there. What I could actually open. It didn't take long to go through all of it. Two or three Power BI dashboards. Deposits, winnings, bets and new users, broken down by country and traffic source. And that's about it. Good for finance, sure. But it told you nothing about what a player does once they are inside the product. So I went looking for the events. Game opens, sessions, clicks, anything. None. The web analytics were simply not on the platform. Nobody actually bothered to do it. So no product analytics, just dashboards about money. Before any of the analytics work, somebody had to put the tracking in. That somebody was the web team. ## 2. Six story pints "Our sprints are all planned out for the next three months. We have no space." That was the web department. I tried to negotiate, I pleaded a bit as well. Same answer. Then I found out the lead of the web analytics team was based 60 kilometres from me. 60 km. Close enough for a beer. So I invited him for a beer. No agenda. (Well. Some agenda.) We got to know each other first. At the end of the evening, a few beers deep, we agreed on six story points in every sprint for implementing web analytics. Every sprint, not a one-off favour. We're still friends to this day, by the way. The tracking went in and the events started coming. Then I found the second problem, and that one you can't fix with a beer. ## 3. Nobody liked numbers People were scared of the numbers. When I talked to the leads of the departments and products, it was clear they didn't really know how to work with analytics. They were intimidated by it. So tracking on its own was not going to change much. I went at it from a few sides. First one was 1:1 sessions with the leads. Mentoring and coaching on how to work with analytics and how to work with a hypothesis, with the goal of having it in their day-to-day. We sat down together and built their own dashboards, on their own KPIs, goals and targets. Then a weekly company-wide report. Sent to everybody and reviewed, showing how the company is doing. Obviously with some obfuscation, some numbers are not for everyone. But the point was that everybody sees the same data, every week. It lowers the psychological barrier. And then a couple of all-hands lectures from me, about how analytics was used in other products and other cases. My favourite one is the airline that stopped giving cocktail napkins and saved so much on fuel. Point being, numbers are significant, numbers need context, and small changes can have a big effect given the scale and time. I wanted to interest people and inspire them a bit. I think I did. The whole thing took a couple of months, and then we got to see if any of it was worth the effort. ## 4. What it moved Here are the results. Player LTV went up by about 20%. Average player lifetime got longer by 3 months. Upsell and cross-sell performance went up by about 14%, through experimentation, behavioural segmentation and personalisation. How it got there is honestly quite boring. Something shows up in the data, it goes on the roadmap, it gets shipped, we measure it. Then again. Then again. Boring, but it works. The fun part came after. ## 5. The fun part Once the analytics were running, we started gathering data on player behaviour. That allowed us to do new things. First was a recommendation engine. We looked at player cohorts and tracked their behaviour, with a simple hypothesis. Players with this particular behaviour, who play these particular games, are also interested in those games. So you put those games in their suggestions. They find the next game they like, they stay on the platform longer and keep playing, and their lifetime goes up. So it was a player retention mechanic, based on behavioural patterns and player preference. It needed some fine-tuning. Then it worked. The second iteration came out of the same understanding of behaviour. By then we had seen enough fraudulent cases to notice certain patterns in them. So we built an internal anti-fraud system with models that flag suspicious behaviour when it matches those patterns. It let us monitor fraud and reduce the chance of it. ## 6. The order holds Now, I understand this is not impressive today. Not rocket science. But you need to keep in mind this was done before AI was where it is now. What I still use from it is the order. Get the data flowing first, because without events there is nothing to look at. Then get the people reading it, and that is a people problem, not a tooling one. Only then build on top of it: recommendations, anti-fraud, whatever your product needs. I build AllCars.cy now, with AI coding agents doing a lot of the building.