Solflare’s data team tracks wallet activity beyond the app
A new Solflare user does not have to be new to the crypto market. A wallet’s public transactions can reveal activity dating back to well before installation. For the Serbian-Croatian fintech operation’s data team, this opens the door to different product decisions – but also makes it essential to distinguish an address’s history from the person behind it.

Installation is not necessarily the beginning of the story. When a wallet address is used in Solflare Wallet, it may already have a transaction history on the Solana blockchain. That history exists outside the app and is public. So the data team does not have to start with a blank slate when trying to understand usage.
This is a central premise of the work described by Mihael Alapić, data lead at Solflare, to the Serbian technology site Startit. The team combines events in its own product with activity on the blockchain. Together, these data provide a broader picture than either source can on its own.
But broader does not mean complete. An address can show what has happened to assets without explaining why. App data can show how someone moves through an interface without revealing what they were trying to achieve. Solflare’s data work is interesting precisely where these two perspectives meet.
An address is not a person
Solflare Wallet is a crypto wallet for Solana. Here, the word wallet needs to be split into two things: the app the user interacts with and the address on the blockchain whose activity can be tracked. The address’s history is not limited to what happens through Solflare’s interface.
Alapić describes how previous crypto activity is used to divide users into groups. This allows the team to distinguish between being new to the product and being new to the activities the product is used for. These are two different starting points for product development.
For example, a person with previous activity may need to learn where a feature is, rather than what it does. This is a possible product question raised by the history, not proof that every active address belongs to an experienced user. Transactions are traces of actions, not a test of knowledge.
There is also an important boundary between an address and an identity. The same person can use several addresses. Linking activity to an address is therefore not the same as knowing how many people are behind a behaviour, who they are or what their intentions are.
In the approach described, a public wallet address is linked to device identifiers and an internal user ID. This makes it possible to relate blockchain activity to app usage. The link provides a basis for analysis, but does not turn the address into a reliable biography of a person.

Two data sources, different kinds of answers
App data and blockchain data answer partly different questions. Events in an app can help a product team understand which features are used and how users navigate. The blockchain adds the perspective of transactions, including activity that takes place outside the company’s own product.
Solflare’s toolkit includes PostHog for product analytics and dbt for cleaning, joining and modelling data. The latter may sound technical, but it is crucial to what a report actually says. Data from different sources need compatible definitions before they can inform decisions.
Consider the difference between an app event and a transaction. Clicking a button does not automatically mean an action has been completed on the blockchain. Conversely, a transaction on an address need not have been initiated in Solflare. An analysis that conflates these things risks attributing activity that took place elsewhere to the app, or counting an intention as a result.
The provider Alchemy describes an example in which retrieving a wallet’s history takes under a minute. This illustrates one specific part of the work: obtaining historical data. It does not, however, say how long Solflare’s entire analysis workflow takes, how the data are linked or how reliable the conclusions are.
Fast access does not resolve questions of definition either. What should count as an active user? When do several events belong to the same sequence? Which activity is relevant to the feature the team wants to improve? These are the choices that determine whether a technically accurate summary is also useful.
When revenue analysis becomes a product question
One concrete example from Alapić concerns the contribution of high-spending users to revenue. The focus shifts here from how many people use the product to how different groups contribute to the business.
The distinction matters. A widely used feature and a revenue-generating feature are not necessarily the same thing. Nor does a group that stands out in activity statistics necessarily carry the same weight in revenue. Understanding the relationship requires both clear group definitions and a considered definition of what counts as revenue.
In Alapić’s account, the analysis is linked to a VIP programme, exclusive events and the opportunity to give feedback on new features. It therefore provides a basis for how the company develops its relationship with a particular user group, rather than simply reporting transactions that have already taken place.
Here, the two data sources contribute different things. App data are needed to understand how people interact with Solflare’s features. Blockchain history provides a broader context for an address’s activity. Previous crypto activity and contributions to Solflare’s revenue, however, are not interchangeable measures: one cannot automatically be used as evidence of the other.
It is also important to distinguish between observation, interpretation and outcome. That a group contributes to revenue is an analytical observation. Deciding that it should receive a special offer is a business decision. Claiming that the offer then improves loyalty or the product is a further assertion, which needs its own follow-up. The example described provides no basis for putting a figure on such an effect.
Close dialogue with high-spending users also provides feedback from that particular group. It does not replace understanding of people who use the wallet in other ways. This is a trade-off for the product team, not something a revenue report can decide for them.

AI can write the report – but not take responsibility
Alapić also describes how AI has changed analytical work. Claude Code is among the tools used, and report writing is faster. At the same time, manual quality checks remain in place.
It is easy to see why. A well-written report may be based on the wrong selection of data. A calculation may be performed correctly but answer a different question from the one the product team asked. When data from both the app and the blockchain are used, there are several points at which an assumption can be wrong without the final result looking implausible at a glance.
The material describes human review, but not a complete quality-control process at Solflare. It is therefore not possible to attribute specific tests or approval steps to the team. The questions to check are clear, however: do the links between data sources hold up, are the groups defined correctly, and does the report really measure what the decision is about?
AI’s speed does not make these questions less important. Rather, it shifts the emphasis from producing a summary to assessing whether it stands up. A polished chart is still only a presentation of the choices made along the way.
Solflare’s example shows data work in which the product’s boundaries are not the data’s boundaries. The team can trace an address back and see activity outside its own app. But moving from public history to a well-founded product decision still requires someone to distinguish what can be observed from what can only be assumed.
Sources & further reading
- StartitSolflareov Data tim je za godinu narastao od 1 na 6 ljudi — sada traže još dvoje: 'AI je dosta promenio posao tima, ali je najvažniji deo posla — isti'
- alchemy.comSolflare - Web3 wallets - Alchemy | Alchemy
- alchemy.comHow Alchemy Powers Financial Services Onchain | Alchemy
- startit.rsSolflare je jedna od najbrže rastućih IT kompanija u Srbiji, i među najnaprednijima po uvođenju AI — podržava AI Week kao glavni partner
Share the article
Translated from Swedish with AI assistance. Read the original.



