Skip to content
Market

Digital lending in Colombia: fintechs, delinquency and decision engines

What the credit market in Colombia looks like: fintechs vs. banks, delinquency, vintages and why decision technology separates who originates well.

Updated July 2026 · 9 min read

In short

In Colombia fintechs are growing (originations +11% while banks fall 20%), but almost entirely on one product, the short-term personal loan, and with riskier customers. Delinquency is mixed, though fintech vintages from 2023 on improve by originating with better tools. The edge is not growing more, but originating better. Lulo Bank is the case. Data from TransUnion.

Credit in Colombia is in the middle of a reshuffle. Fintechs are gaining ground on traditional banks, but they are doing it on a single product and with customers who are increasingly stretched. The numbers below, most of them from TransUnion's Colombia credit report, help show where the market is growing, where it is deteriorating, and why decision technology ended up being what separates lenders who originate well from those who do not.

Fintechs gain ground, banks pull back

Most consumers still go to traditional banks, between 74% and 79%. Fintechs hold steady at around 16%. To read the trend cleanly, TransUnion pulls Nubank out of the fintech group and counts it as financial sector, because its size distorts everything else.

The interesting part is the flow, not the snapshot. While the financial and cooperative sector originated 20% less new credit over the year, fintechs grew 11%, filling the space banks left behind.

A market tied to short-term personal loans

Fintech growth rests almost entirely on one product: the short-term personal loan, which already accounts for 94% of their new originations. Fintech share of that product jumped from 12% in 2021 to nearly 30% today.

And they started reaching for bigger amounts. In short-term loans, the 500,000 to 1.5 million peso band went from 14% to 21%. Traditional banks are moving the other way, lowering amounts to compete for those users.

The profile: the riskiest segments get included

Fintechs mostly serve the hardest segments: under 30, lower income, people new to credit with less than two years of history, and below-prime profiles with a score under 600. In fintech personal loans, the subprime segment went from 34% in 2021 to 43% in 2024.

On top of that, nearly 6 in 10 fintech users also hold credit with a bank, and many are stretched: they put up to 45% of monthly income toward installments. The system as a whole is tense. Over the past year, 38% of consumers saw their score worsen and only 27% improved it.

Delinquency and vintages: the signal is in how it was originated

The deterioration indicators are mixed. In banking, the hardest-hit product is microcredit, with delinquency up almost 300 basis points over the year. In fintechs, delinquency on short-term personal loans doubled.

There is an important nuance in roll rates, which measure the share of accounts moving from current into delinquency. Fintech short-term loans roll faster into early delinquency, yet in payroll loans fintechs come out better than banks on advanced delinquency, 20% against 46%.

The most telling picture is in the vintages. Traditional bank originations from 2023 on are deteriorating more than earlier years; fintech ones, by contrast, are improving. The report gives two reasons: they did not overgrow between 2021 and 2022, and they are originating with better processes and better tools.

Technology is what separates the vintages

Here is the point that matters for a lender. Keeping rigid legacy systems delays product launches and makes changing a risk policy take months, with high costs and little room to react.

Lulo Bank is a good example of the other path. Moving to a modern decision engine with uFlow, it left behind a model with over 300 variables and a 700-branch tree. Policy changes dropped from months to weeks or hours, processing grew almost 300%, and approval rose by up to 12 points.

For very large institutions, where touching the core is almost impossible, the market recommends "speedboats": setting up units or small parallel companies with new, agile technology instead of fighting the old structure. And the underlying trend is to drop the rigid portfolio and build personalized approval flows, with alternative data, that adapt to each applicant.

What to watch if you lend in Colombia

The lesson from the vintages is clear: the edge is not in growing faster, but in originating better. And that translates into concrete things. Orchestrate data to see the whole customer, test every new policy on 1 or 2% of the portfolio before releasing it, and move limits with payment evidence instead of a fixed portfolio.

Above all, govern risk with traceability. In a market that is tightening, being able to explain and reverse a decision is worth as much as making it fast.

Frequently asked questions

Common questions

What share of credit in Colombia do fintechs hold?+

Around 16% of total consumers, with Nubank set aside because its size distorts the measurement. But in short-term personal loans their share jumped from 12% in 2021 to nearly 30% today, and that product is 94% of their new originations.

Why are fintech vintages improving since 2023?+

Per TransUnion's Colombia credit report, for two reasons: they did not overgrow between 2021 and 2022, and they are originating with better processes and technology. Traditional bank vintages, by contrast, are deteriorating.

What did Lulo Bank achieve with a decision engine?+

It left a model with over 300 variables and a 700-branch tree. With uFlow it cut policy change time from months to weeks or hours, grew processing almost 300%, and raised approval by up to 12 points.

What is a "speedboat" strategy?+

For large, complex institutions, instead of fighting the old core, you set up a unit or parallel company with new, agile technology. That lets it innovate fast without clashing with the traditional structure.

Would this work for your decisioning process?

Transform your credit assessment process with the decision engine.