‘I was taught the way of progress was neither swift nor easy’ – Marie Curie

Setbacks dominated the banking as a service headlines in 2024. Over the past 18 months, US regulators issued more BaaS related consent orders than anyone cares to count[1]. European pioneer Solaris shut down its UK arm and is, it seems, exiting the segment altogether in Europe, focusing on co-branded cards. Other launches, like NatWest Boxed, have been delayed.

The decline of some providers is contrasted with a steady rise in financial services income at non-fintech companies like Shopify and Grab. Another, ServiceTitan, reached a valuation of $9bn on its first day of trading in December 2024. The SaaS company, which offers customer management software for the trades, had converted the vast majority of its customers to using the embedded payments offering within three years from adding it.

The struggles of some BaaS incumbents (the irony of calling them that is not lost on me) creates a compelling opportunity for licensed banks in embedded finance. With a more comprehensive set of products, supported by their balance sheets and lower cost of capital, banks can step into the space vacated by platforms. Indeed, across continents, banks are busy researching the space and formalising their strategies. Some have already launched with customers including SEB (Hemköp), HSBC (SemFi) and Japan’s GMO Aozora (Habitto).

Winning in this emerging space, however, requires a fundamentally different approach from how banks operate their direct to consumer or SME business. Viewing embedded finance as simply a new distribution channel to the same products and supported by the same operating model is bound to end in frustration and missed ROI. In this nascent space, banks compete on multiple plains - one is the partner relationship level, and the other is the end customer market. Embedded finance for a bank is akin to playing three-dimensional chess on Star Trek (or in Ben Horowitz’s words on The Hard Thing about Hard Things – ‘this is not checkers…’  you can find the rest of the quote here)

A Bank’s Playbook for Embedded Finance

The opening gambit is a strategic commitment to the model at senior levels but that isn’t the subject of this piece. Success – or checkmate - is achieved by launching a superior product - a product co-created through a partnership laser focused on a specific use case. An integrated product and user experience which blurs the traditional boundaries of both bank and partner.  Integrated data sharing, a modified operating model and relevant pricing support, drives shared success.

The challenge of nearly all embedded finance projects is to change customer behaviour by enabling financial transaction outside of the traditional financial environment. End customers must be presented with a compelling offer at point of need, so that they choose it there and then, rather than explore the wider world of digital financial offers outside of the partner’s garden wall. Old habits – people contacting their banks directly – may die hard, but, with assistance, they die nevertheless.

Data Sets Embedded Finance Apart

How, then, is a superior embedded finance product different from any other accounts, payments or loans banks sell? First thing to consider are data flows.

Financial institutions have for long relied on their own data lakes to understand their customers. They spend millions trying to extract the value they believe lie deep in the lakes. They acquire the analytical tools de rigueur and look for data enriching opportunities to supplement the lake with context nutrients. Yet, most of the transaction data feeding that lake is short on context needed.  

Embedded finance partners already have plenty of context, which banks do not. More importantly they can identify strong propensity to buy the right product. Data synergies are possibly the biggest value add to the bank from embedded finance. Successful partnerships should support data enriching in both directions. Partner holds highly relevant usage data – e.g. how often an end customer SME is turning around their inventory, how often a creator is producing content and what are the inputs required – and so does the bank, primarily payment, past loan data as well as wider industry and economic trends.  The age of banks thinking that all the data they hold is proprietary data is long gone.

Banks need to also keep an open mind on where it is best to hold data and perform certain functions. For example, it might make sense to bring the affordability model into the partner domain as it may be easier to bring the model to the data, instead of duplicating it and saving it close to the model.

Consider this - Amazon has been providing loans off its balance sheet to sellers for a few years now. They have developed some risk and decision models but have not scaled that business (scale is a relevant term, Amazon is lending a few billions of dollars, but it is still a small fraction of the business for them) and the models are not fully matured. Banks looking to partner with Amazon, should incorporate, or in the future rely entirely on, Amazon models that were built on real-time, accurate data and not just historic information from accounting software provided by a borrower. These models may be augmented by macro inputs from the partner bank where relevant.

Effective management of embedded finance risk is predicated on having a similar view of performance and activities. Regulators are paying close attention to third party management risks these days following the Synapse implosion that highlighted serious inconsistencies in reconciliation (among other irregularities). More open data flows should help with the all-important reconciliation, a crucial process which ideally should happen on both bank and partner ends.

Another key data consideration is KYC. Partners often already hold much of the required information, though it may not always be current. Traditional bank KYC/KYB processes – usually designed to be completed in a single session to avoid drop offs may not be fit for purpose in embedded finance. Instead, KYC processes may need to be split into stages and fit with the desired user experience. Any superfluous data capture must be eliminated to avoid degrading of the user experience. KYC processes need to run in context of the relevant buying journey of the underlying partner service or product, rather than the financial one.

Operational Design for Complex Stakeholder Relationships

Operational design is critical for creating a seamless experience for both partners and end users. The two audiences for the bank are:

1.        The embedding partner, who needs smooth onboarding and clear processes from contract to launch.

2.        The end user, who is a customer of both the bank and the partner.

A well-designed operational model minimises costs and ensures the right teams—whether at the bank or the partner—can quickly resolve issues. Efficient and seamless experience both through the happy path as well as when things go wrong is an integral part of the product. The challenge lies in accommodating partners with different skill levels and strategies.

Banks must embrace modular operational designs that offer flexibility for the partner. A modular approach creates a menu of possible operational options from which the partner and bank can choose from. Roles and responsibilities would follow, followed by decisions on data residency and exposure to the right teams. Not all partners will use all the bank’s support infrastructure. Some would be capable of, and indeed insist on, providing holistic customer care to their customers.

Embedded Finance products are much more than just technical features delivered by APIs. Partners want a smooth lifecycle of operations, from partner onboarding to end customer dispute resolution, to ensure they can focus on offering the best service to their customers. This means banks must adapt their internal processes. Compliance teams, for instance, may need to work through third parties rather than directly with end customers. Similarly, customer service teams may need to adjust to indirect support roles. 

Good operational design not only improves end customer experiences but also reduces unit costs, making embedded finance more profitable for banks and partners alike.

The Role of Pricing

Pricing ties everything together and determines the success of an embedded finance product. Even with excellent data flows, efficient operations, and a great user experience, pricing must be competitive to drive conversions.

Since the end customer’s purchasing decisions happen at the partner’s point of need—beyond the bank’s direct control—pricing must appeal to both the partner and the end customer. Many partners aim to attract customers away from their current banks (e.g., a restaurateur choosing a Toast account over their existing one). To succeed, pricing must give partners the flexibility to remain competitive while reflecting the bank’s lower acquisition and service costs.

Collaboration with financial planning and analysis (FP&A) teams is essential to adjust pricing models for embedded finance. Asset and Liability Management (ALM) committees should debate and potentially reevaluate how BaaS deposits are treated, particularly in relation to liquidity coverage ratio (LCR) and net stable funding ratio (NSFR) requirements. These deposits are not exactly like retail/SME deposits, but then again, they are not entirely different. It all depends on the partner desired journey.

Unit costs in embedded finance can differ significantly from legacy models because some costs are shifted to the partner side. Banks must ensure that fully loaded costs reflect these differences and work with internal teams to understand the unique economics of this space. Lower acquisition and servicing costs should enable more competitive pricing compared to incumbent alternatives, driving greater adoption.

Keeping the organisations expected returns constant (because even your influence is limited), lower cost to acquire and serve should enable more competitive pricing compared to the end customer’s incumbent alternatives.

Taking the Leap

As of early 2025, embedded finance remains unproven at scale. It is still a bet. Incentive schemes in banks mean leaders often cannot resist the temptation to reject all the possible things that are "not proven yet". Alas, it is hard to win in a new game without making some bets.

Mitigating factors exist. The market’s small current size allows for testing and iteration. Most banks are likely to find willing clients that wish to pilot embedded finance propositions. These trusted existing business relationships should be resilient enough to endure the hiccups that are bound to happen. Embedded finance isn’t digital transformation, it is about building readiness for an emerging business opportunity.

In McKinsey’s Three Horizons of Growth framework, developed to help businesses manage current operations while preparing for the future, I believe embedded finance is moving from Horizon Three – creating genuinely new business models – to Horizon Two – nurturing an emerging one. Use cases exist, from mobility to international shipping to workforce management, but haven’t yet scaled. Banks must act now to develop the necessary operating models for scale before the opportunity passes them by.

An unsourced but common quote describes chess as a battle between your aversion to thinking and your aversion to losing. The same can be said about embedded finance.

Embed long and prosper.

[1] Except for Jonah Crane of Klaros Group whose coverage of the subject is a must read