Tapix CEO says banking features depend on transaction data quality

Digital banking features that appear similar on the screen can produce sharply different results if the transaction data beneath them is incomplete or unreliable, according to Ivan Dovica, chief executive of Tapix by Dateio. In a Finextra community opinion post, Dovica said banks risk higher support costs, weaker engagement and poor partner economics when they build products on feeds that cannot consistently identify merchants, categories or recurring payments.

Dovica said product teams can often copy the visual design of a competitor’s feature, such as an annual spending summary, carbon-impact display, transaction search tool or subscription tracker. The constraint, he argued, is whether the bank’s data layer can support the feature once customers begin using it across years of payments history.

He separated transaction data quality into three elements: richness, coverage and accuracy. Richness refers to the context attached to a payment beyond the amount, date and raw payment descriptor. Examples include identifying the merchant behind a gateway, adding store-level detail, localised names, URLs or sustainability-related metadata where a product requires it.

Coverage measures how much of the transaction base is enriched to the standard needed for a specific use case. Dovica said this is often underweighted when banks assess vendors, even though partial coverage can undermine a product after launch. A subscription tracker that finds only some recurring payments may lead customers to conclusions the data cannot support.

Accuracy concerns whether each individual datapoint is correct. Dovica said a wrong merchant name, category or subscription flag can damage trust because customers tend to remember the visible error rather than average it against a broader feed.

Costs, engagement and commercial use cases

Dovica argued that banks should begin with the metric they want to improve rather than selecting a feature first and discovering later that the data cannot support it. For cost reduction, he identified support contact volume, disputes, chargebacks and fraud losses as relevant measures. Unrecognisable transactions can generate operational work, while clearer merchant feeds, location details, logos, search tools and merchant-level history can help customers resolve questions themselves, he said.

For engagement, Dovica pointed to adoption, session frequency and churn. Personal finance tools such as budgeting, category summaries and trend views depend on customers accepting the classifications as consistent. Recurring payment detection, he said, requires merchant recognition, frequency analysis and amount stability to identify subscriptions, duplicates and price increases.

For loyalty and partner revenue, the key constraint is merchant attribution, according to Dovica. Targeting, redemption and partner reporting rely on correctly assigning transactions to merchants. If attribution is unreliable, he said, settlement with partners becomes difficult to verify and the customer experience deteriorates.

Sustainability features impose another data requirement: consistency over time. Dovica said per-transaction carbon estimates depend on merchant category, location and related metadata, while Scope 3 reporting for business customers rests on the same foundation. Changing categorisation between periods can make the resulting figures hard to defend under scrutiny.

Dovica concluded that digital feature parity is becoming cheaper at the presentation layer and more costly below it. In his view, banks planning product roadmaps should assess which features their current data layer can support without eroding customer trust, raising service costs or weakening downstream analytics.

This story draws on original reporting from Finextra Research.