How Kuda constructed its in-house core banking software

Till the latter half of 2024, core banking functions (CBAs) attracted little public consideration exterior Nigeria’s banking business. That modified when a number of main industrial banks switched or upgraded their CBAs, disrupting banking providers for hundreds of thousands of consumers.

However 5 years earlier than core banking was on Nigerians’ lips, the founding crew at Kuda, a Nigerian neobank which boasts over seven million prospects, knew that constructing its CBA may very well be the distinction between success and failure. 

“It was, in fact, one of many hardest and doubtless essentially the most daring choices I’ve ever taken in my skilled life,” Musty Mustapha, Kuda’s co-founder and chief expertise officer, stated in regards to the choice to construct an in-house CBA. “I knew I needed to stick my neck on the road with my founding companion, the buyers, the board, and all people else.”

Banking’s financial logic has not modified a lot since its early days: establishments gather deposits from prospects, assure entry to the deposits, and finance the economic system by extending credit score. What has modified is the mechanics and equipment powering it. 

At the moment, the core banking software, an always-on ledger, is the software program that information and powers each deposit, mortgage, and transaction in a financial institution. Because the engine of contemporary banking, it ensures that accounts are up to date in actual time, that financial institution operations are dependable and scalable, and that the financial institution can launch and handle merchandise. 

The way it occurred

In 2019, Kuda was nonetheless sufficiently small that its key decision-makers might match on a small desk. There have been no headline enterprise capital corporations like Goal International but. The neobank was reside and rising quick, however operating on a third-party core banking supplier. 

That is the mannequin most monetary establishments in Nigeria depend on as a result of constructing core banking infrastructure takes time, effort, and capital that may very well be higher spent elsewhere. Other than Kuda, solely a handful of Nigerian monetary establishments, like Moniepoint and Sterling Financial institution, have developed their very own core banking functions in-house.

Nevertheless, inside Kuda, the third-party CBA was beginning to crack beneath the burden of the neobank’s ambition. Extra prospects and product releases drove demand and a rising intolerance for instability, inserting strain on the CBA. 

Mustapha knew that this stress might jeopardise his younger startup’s future. A sequence of conversations together with his engineers and Kuda’s CEO, Babs Ogundeyi, led to what ultimately grew to become NERV, Kuda’s in-house core banking system, named after the human nervous system as a result of it sits beneath every thing and quietly controls how Kuda operates.

Mustapha insisted on one level above all else throughout our hour-long dialog in October 2025: when Kuda began constructing NERV, Nigeria’s banking or fintech market had not constructed an in-house core banking system like Kuda was trying to.

“On the time once we constructed NERV, there was no fintech or banking-related firm that had constructed its personal core banking system,” he stated. “NERV is the pioneer in-house-built core banking answer by a financial institution in Nigeria.”

Nevertheless, Moniepoint started its in-house core banking improvement in 2020, and a number of other Nigerian corporations like Appzone’s Qore already bought core banking merchandise to Nigerian microfinance banks. 

NERV started as most banks take into consideration CBAs: buying. Musty approached a number of native and overseas suppliers, searching for a CBA sturdy sufficient for what Kuda wished to change into and low cost sufficient for what Kuda might afford in 2019. The startup ultimately launched in August 2019 on an area core banking supplier.

Then actuality set in as Kuda grew to one million prospects shortly, and the system felt brittle beneath this weight. Mustapha described three points that saved compounding: it couldn’t scale with buyer progress, couldn’t be tailored to Kuda’s wants, and above all, couldn’t be relied upon to remain up.

“In case you are providing a monetary service, the minimal that you are able to do is to make sure that your techniques are steady to your prospects,” he stated. “Should you can’t obtain this, you’ll be able to as nicely simply pack your load and go house.”

So, Kuda did what most establishments keep away from: it saved in search of alternate options earlier than deciding there weren’t any that met its wants inside the constraints it had. At that time, Mustapha stated that the query grew to become existential moderately than technical: what sort of firm are we constructing?

His reply was easy. “Kuda is an engineering-first firm,” he stated. “That’s the type of firm we’re attempting to construct: utilizing engineering sources to resolve monetary providers issues in Nigeria.”

Mustapha added that Kuda by no means wanted to run an extended persuasion marketing campaign, regardless that constructing a core banking system might value tens of hundreds of thousands of {dollars} and is sluggish. 

In September 2019, the corporate had raised solely $1.6 million in a pre-seed spherical from angel buyers and household places of work. This lack of institutional capital meant fewer layers of approval, much less pink tape, and fewer individuals who might veto a high-conviction choice.

However even then, Kuda averted betting the enterprise on a rebuild whereas it was nonetheless studying to exist. As an alternative, NERV ran in parallel to the prevailing CBA. Kuda continued working on the third-party core whereas engineers constructed NERV within the background. That approach, if it failed, the financial institution wouldn’t die. It could hold operating, discover a new supplier, or endure the one it had.

“It wasn’t like the corporate stopped working,” Mustapha stated. “We had been nonetheless on the third-party supplier system, nonetheless serving prospects, nonetheless rising, however we’d determined and had been treating it as a facet mission.”

Kuda had one other benefit; its engineering expertise was prepared to tackle an extended, invisible construct whereas nonetheless sustaining manufacturing techniques. 

“There’s no approach I’d have even tried to go forward with it if that high quality was not there,” Mustapha stated. “I’m talking about an truly world-class crew.”

The mechanics of NERV

NERV’s architectural decisions had been much less unique than the result itself. Kayode Ilesanmi, one of many engineers who constructed it, stated the crew began with a transparent conclusion: if Kuda wished scalability and adaptability, it couldn’t construct a monolith.

“We simply needed to say, ‘You understand what, for us to have the ability to obtain that type of scalability and adaptability we wish, we’ve to go together with a microservices structure,’” he stated.

What stands out with this strategy is the ambition behind the throughput targets. Kayode described a guideline that grew to become a type of inside dare: 1,000 transactions per second, that means 1,000 prospects hitting the system concurrently ought to really feel like one.

They had been nowhere shut at first. “We began the primary time, and we couldn’t even rise up to 10,” Ilesanmi stated.

Then they did what infrastructure groups ought to do. They traced the bottleneck, inspected logs and metrics, and questioned whether or not the problem lay within the database, the cache, the appliance logic, or the compute. 

He recognized the cache connection dealing with as a weak level, studied how main implementations managed it, rewrote their strategy, and noticed the load take a look at soar to over 500, then ultimately to the 1,000 aim.

By the point they had been close to go-live, the crew says they had been hitting 5,000 transactions per second in assessments.

Migration day: when manufacturing refused to cooperate

If you wish to perceive the cultural variations between big-bank and startup migrations, begin with threat. Musty referenced Nigerian banks’ public migrations, noting that enormous banks spent hundreds of thousands of {dollars} hiring world distributors and nonetheless confronted messy outages and backlash. 

Should you apply that strain to a startup barely a yr previous, with a small crew, and no possibility to cover behind institutional distance, you get a failed migration day. 

“Sure, completely,” Mustapha informed me after I requested if it was painful. “It wasn’t all rosy.”

What Kuda did in another way, he believes, was to deal with prospects like individuals moderately than victims. Within the weeks earlier than they went reside, Kuda started telling customers they had been switching cores, when it could occur, and that the result could be fewer failures and quicker service. 

When migration day arrived, communication intensified as emails, real-time updates, and a social media thread changed into one thing uncommon in Nigerian banking: prospects defending the financial institution whereas it was down.

“If anybody was even saying, ‘Oh, what’s occurring, Kuda?’ it was prospects that may be responding to them,” Mustapha stated. Nevertheless, the plan nonetheless failed at first. 

Kuda had constructed the system in a take a look at setting, then in a staging setting designed as a near-production rehearsal house. However after they moved into manufacturing, the migration failed. 

So that they reverted to staging and pushing manufacturing information into it, as a result of staging behaved just like the setting they’d examined and load-tested repeatedly.

“The manufacturing setting was sluggish, which was actually bizarre,” Ilesanmi stated. “It was an setting that had larger specs… and the expectation was that requests… ought to be quick. However they had been sluggish.”

Once they switched again, the latency normalised, and the financial institution launched within the setting that labored.

What does proudly owning core banking change?

When fintech leaders speak about in-house core techniques, they usually lead with margins. Nevertheless, Musty stated the transfer wasn’t a cost-cutting one.

“We didn’t take this choice… as a result of I felt, ‘Oh, we’re going to cut back value,’ or ‘We’re going to make a number of cash out of it.’ No,” he stated. “The aim… was to be ready the place we will serve our prospects rightly.”

All the things else, he argued, turns into a downstream profit: higher unit economics, product velocity, and fewer vendor constraints. 

Proudly owning the core additionally modifications a financial institution’s threat profile. With a third-party core, your operational threat is now not simply your threat; you additionally inherit your supplier’s downtime, constraints, roadmap, and safety score.

“By having our personal system, we’ve type of eliminated that third-party threat,” Musty stated.

From an engineering and product standpoint, the crew described instant wins that customers might really feel. The app grew to become quicker, and transfers had been accomplished extra shortly, particularly through the COVID-19 pandemic when digital transaction quantity surged.

With NERV, Kuda scaled inside overdrafts into customer-facing credit score, constructed enterprise banking merchandise, launched time period loans, and expanded into multi-currency wallets, capabilities the crew says would have been slower, costlier, or structurally tougher on an exterior core.

From an operations standpoint, the modifications had been much more concrete. Earlier than NERV, operating end-of-day (EOD) routines (ledger balancing and banking shut processes) might trigger the system to freeze. 

With NERV, they automated EOD so it runs with out locking the financial institution’s capacity to function through the day.

“It grew to become very versatile, the place you don’t even want any individual to go there to run EOD, and every thing runs by itself,” Ilesanmi stated.