Open Finance Gave Everyone the Same Data. Behavior Is the Only Differentiator Left.

When every licensed institution can see the same account balances, transaction histories, and income streams, the data itself stops being an advantage. What remains proprietary is how each user interacts with your product.

Published 2026-07-17 · 6 min read

Open Finance Gave Everyone the Same Data. Behavior Is the Only Differentiator Left.

The Moat That Regulation Dissolved

For most of banking history, the customer relationship was the moat. The institution holding the primary account saw the salary deposits, the recurring bills, the spending patterns, the savings behavior. Competitors saw nothing. That asymmetry was the foundation of incumbent advantage, and it was defended fiercely.

Open finance dismantled it on purpose. Brazil's implementation, phased in by the Banco Central, went further than most: not just account and transaction sharing, but payment initiation, investment data, insurance, and pension information, all under a consent framework the customer controls. The explicit policy goal was to break the data asymmetry that let incumbents hold customers through inertia.

It worked, at a scale no other market has matched. Open Finance Brasil passed its fifth anniversary in February 2026 with more than 154 million active consents and over 100 million connected customers, according to the Open Finance Brasil Association. Unique consents grew 143% between 2024 and 2025. The ecosystem now carries more than 5 billion API calls between institutions per week, and payment initiation moved R$ 15.3 billion in 2025, up from R$ 3.2 billion in 2024. For comparison, the UK — the pioneer — reached roughly 15 million open banking users by mid-2025, and its scope covers only account data and payment initiation, not the credit, investment, insurance, and pension data Brazil's model includes.

And that success produced a consequence that a lot of fintech strategy has not fully absorbed.

If a customer consents, a challenger fintech can now see essentially the same financial picture the incumbent sees. Same balances, same transaction history, same income timing, same obligations. The data that used to be proprietary is now, functionally, a shared utility.

Which raises an uncomfortable question: if everyone has the same data, what exactly is the advantage?

Same Inputs, Same Answers

Watch what happens when multiple institutions build on identical shared data.

Credit models converge. Given the same transaction history and income evidence, well-built underwriting models reach similar conclusions. There is only so much divergence available when the inputs are identical and the mathematics is well understood.

Product recommendations converge. Every platform reading the same account data identifies the same opportunities — the customer paying high interest who should consolidate, the balance sitting idle that should be invested, the recurring charge that could be cheaper elsewhere. Everyone sees it. Everyone recommends it.

Risk flags converge. The same missed payment, the same income disruption, the same leverage ratio triggers the same alert across every institution watching.

The result is a market where competitors offer strikingly similar products, identify the same opportunities at the same moment, and reach the same customers with the same messages. Differentiation collapses toward price and brand, which is exactly where margin goes to die.

Financial data describes the outcomes of a person's decisions. It records what was spent, saved, borrowed, and repaid. It is an excellent record of the past and a mediocre explanation of it. Two customers with byte-identical transaction histories can be completely different people — one deliberate and confident, the other stressed and reactive — and the shared data cannot tell them apart.

The Layer Open Finance Does Not Share

There is one category of data that open finance does not and structurally cannot commoditize: how a specific user interacts with a specific product.

When a customer opens your app, that session generates signal no competitor can see. How long they hesitate before confirming. Which screens they revisit. Whether they read the terms or skip them. How they navigate when comparing options. Whether they explore before deciding or decide before exploring. How their session rhythm changes when they are under pressure. Whether they respond to a notification in ninety seconds or ninety hours.

None of this is shareable. It is not in the open finance schema, it is not portable under consent, and it exists only in the relationship between one user and one product surface. It is generated fresh every session and it compounds.

It is also the layer that explains the financial data rather than merely restating it.

A shared transaction record shows a customer moved R$ 8,000 from a growth fund to a fixed-income product. Behavioral data from your own surface shows that before doing so they checked the portfolio screen eleven times in two days, opened the withdrawal flow twice and abandoned it, spent four minutes on the capital guarantee page, and did all of this between 11pm and 1am. The transaction is the outcome. The behavior is the reason — and the reason is what tells you whether the next conversation should be about reallocation or about reassurance.

Where the Differentiation Actually Shows Up

Underwriting on the margin. Shared data determines the broad credit picture. Behavioral signal moves the applicants that sit near the threshold. Application behavior — how someone completes a form, whether they revise income figures, whether they abandon and return, how they respond to verification friction — carries genuine incremental signal, and it is signal only you have. On a portfolio where competitors converge on identical decisions, moving the marginal cases correctly is the entire spread.

Timing. Shared data reveals that a customer has an opportunity. Behavioral state reveals whether they are receptive to hearing about it right now. Every institution watching the same open finance feed identifies the consolidation opportunity in the same week. The one that reaches the customer during a moment of financial calm rather than acute stress gets a different answer to the same offer.

Retention. Account data shows churn after it happens — balances drain, direct deposits redirect, activity stops. By then the decision was made weeks ago. Behavioral deterioration is visible far earlier: engagement depth falling, sessions shortening, feature exploration stopping, response latency stretching. That is the window where intervention still works.

Cross-model consistency. As institutions deploy AI across chat, notifications, in-app guidance, and agents, a behavioral profile attached to the person rather than to any single model keeps the experience coherent. The customer who wants terse, direct answers gets them from every surface, not just the one where they happened to establish it.

Personal Baselines, Not Population Thresholds

The technical requirement here is stricter than it looks, and it is where most attempts fail.

Behavioral signal is only useful when interpreted against the individual's own baseline. A customer who takes two minutes to review a loan agreement is not cautious — unless they normally take twenty seconds. A customer checking their balance daily is not stressed — unless they used to check weekly. Population-level thresholds generate false positives at a rate that destroys operator trust within a quarter.

That means maintaining a persistent per-user model with three distinct time horizons. Traits, which are stable dispositions measured over months. Preferences, which drift over weeks. State, which changes inside a single session and is worthless if stale. Collapsing these into one score is the most common failure mode, because it lets slow-moving traits average away the fast-moving state signal that actually drives action.

Doing this well is infrastructure work — event ingestion, per-user modeling across three horizons, and low-latency serving into a live decision path. It is not a feature a product team ships alongside a roadmap.

What It Produced in Practice

In the Fortics deployment, Fluence processed 3.4 million behavioral profiles from interaction patterns alone — timing, navigation, hesitation, channel response, session rhythm. No survey data, no demographic inference, no message content.

Churn fell 40%, because behavioral deterioration surfaced while the relationship was still recoverable. Conversion rose 2.3x, because offers landed against individual readiness instead of a fixed schedule. ML model accuracy improved 3.5x, because the models finally received per-user behavioral features instead of aggregate cohort attributes.

Notably, none of that required data any competitor could obtain. It came from interactions the platform was already generating and discarding.

The Strategic Reading

Open finance did not eliminate competitive advantage in financial services. It relocated it.

The advantage moved from having data to understanding people. From proprietary access to proprietary interpretation. Every institution competing for a Brazilian customer today can see the same financial picture. The one that also understands how that customer thinks, decides, hesitates, and responds is operating with information the others structurally cannot acquire.

Fluence delivers that layer as infrastructure. A \GET /context/{user_id}\ call returns model-ready behavioral context — traits, preferences, and current state — compressed for direct use by an LLM or a decisioning system. It reads the interaction stream a platform already produces, which is why integration typically takes under 10 hours. It analyzes behavioral patterns and never message content, making it LGPD and GDPR compliant by architecture rather than by consent flow.

Segment and Amplitude track what users do. Open finance now shows everyone what users did financially. Fluence understands how and why — and that part is still yours alone.

👉 See how behavioral intelligence becomes a moat that shared data cannot erode →