Introduction
Behavioural credit scoring is becoming an increasingly important part of digital lending as financial institutions look beyond traditional credit histories to evaluate applicants. Credolab, a Singapore-founded alternative credit scoring company established in 2016, has taken another step into the mainstream by joining the FICO Marketplace on 19 August 2026.
The listing places Credolab’s behavioural scoring technology within an environment used by businesses in more than 80 countries, potentially making the company’s technology easier for lenders already operating on FICO Platform to discover and test. Credolab says its technology is designed to provide an additional risk signal based on how applicants interact with their devices and digital applications rather than relying exclusively on conventional financial histories.
The development reflects a broader shift in financial technology: lenders are increasingly combining traditional bureau information, financial data and alternative signals to assess applicants who may be difficult to evaluate through conventional models alone. But behavioural data is not a replacement for affordability assessments, income verification or other established credit-risk controls.
What Is Behavioural Credit Scoring?
Behavioural credit scoring evaluates patterns in how a person interacts with a digital device or application during a credit application. Instead of focusing solely on previous borrowing and repayment records, the approach examines behavioural and device metadata generated during the application process.
According to Credolab, examples include:
- Typing cadence and correction behaviour
- Time spent reviewing terms
- Scrolling behaviour
- Whether an application is completed in one sitting
- Copy-and-paste activity
- Device storage and memory characteristics
- Battery-related patterns
- Counts of calendar entries
- Counts of photos and media files
- Counts of installed applications by category
The objective is not to identify the individual through the content of those items. Rather, the company says the data is used to identify statistically meaningful behavioural patterns that can contribute another risk signal to a lender’s existing credit model.
This positions behavioural scoring as a complementary form of alternative data rather than a standalone lending decision.
How Credolab Integrates With FICO Platform
Credolab’s addition to the FICO Marketplace is primarily an integration and distribution development.
Lenders already using FICO Platform can locate CredoScore within the Marketplace and test the service without rebuilding their broader decision-making infrastructure. Credolab says implementation generally requires embedding its software development kit (SDK) into an existing mobile application or adding JavaScript to a web-based application journey.
The resulting API provides the score and its associated features. A lender can then incorporate the score as another attribute within its existing credit strategy.
Credolab says most implementations can be completed within one sprint rather than requiring months of development.
The FICO Marketplace relationship therefore changes how easily lenders can discover and evaluate the technology, rather than fundamentally changing how the underlying score is produced.
The company still enters into contracts directly with clients and undergoes information-security and data-governance diligence. The FICO Marketplace listing does not remove the lender’s responsibility to evaluate whether the technology is appropriate for its own risk-management and compliance requirements.
What Data Does Behavioural Credit Scoring Use?
Data governance is one of the central issues surrounding alternative credit scoring.
Credolab says its technology collects metadata concerning device configuration and user behaviour during the application process. It does not, according to the company, collect the underlying personal content associated with those signals.
The company says it does not collect:
- Names
- Phone numbers
- Email addresses
- Contacts
- Message content
- Photographs
- Calendar-entry content
- Browsing history
- GPS coordinates
- Bank credentials
- Bank account data
For example, the system may count the number of photographs or calendar entries on a device without accessing the photographs themselves or reading the contents of calendar appointments.
Credolab also says that data collection is permission-based at runtime and that its SDK operates within the permissions granted to the host application on Android and iOS.
The company describes itself as a data processor rather than the controller of the customer relationship. The lender, card issuer or buy-now-pay-later provider remains responsible for its customer relationship and obtains the relevant consent within its own application journey.
Credolab also says it signs a data-processing agreement with each client, maintains ISO 27001 certification and can store metadata in-country where data-residency requirements apply.
Behavioural Data Is Different From Behavioural Biometrics
Behavioural credit scoring should not automatically be treated as behavioural biometric authentication.
Credolab says it does not authenticate an individual against a stored template of their typing or other behavioural characteristics. Instead, it uses information generated during a single application session to infer statistically relevant characteristics associated with credit risk.
That distinction matters because credit-risk scoring and identity authentication address different problems.
An identity system seeks to establish whether someone is the person they claim to be. A credit-risk model attempts to estimate the likelihood of a particular financial outcome based on an agreed definition of risk.
Can Behavioural Scoring Improve Credit-Risk Models?
Credolab argues that behavioural information can add value even when an applicant already has a substantial conventional credit file.
A recent evaluation cited by the company involved approximately 20,000 applications from a regional neobank’s unsecured loan portfolio over a six-month period. The evaluation was conducted out of sample with a credit bureau, although the parties were not identified.
The reported Gini coefficients were:
| Model | Gini coefficient |
|---|---|
| Bureau score alone | 0.26 |
| Credolab score alone | 0.37 |
| Combined model | 0.44 |
The company also reported different results according to the depth of an applicant’s credit file.
For applicants with thick files, the combined model reportedly increased from 0.36 to 0.44, an eight-point improvement.
For thin-file and new-to-credit applicants, the combined model reportedly increased from 0.35 to 0.39, a four-point improvement.
These figures come from Credolab’s cited evaluation and should therefore be understood as company-reported results rather than evidence that the same performance will occur across every lender, market or portfolio.
The underlying argument is that traditional credit data can describe historical repayment behaviour, while alternative behavioural signals may provide a different type of information.
The Limits of Behavioural Credit Scoring
Behavioural credit scoring has important limitations.
Most significantly, Credolab says its score does not measure affordability. It cannot replace income verification, debt-to-income calculations or other assessments of an applicant’s ability to repay.
The technology also depends on digital application channels. Applications completed through branches, paper processes or call centres fall outside the system’s digital collection environment.
Its usefulness may also vary by lending product. Credolab says the incremental contribution is considerably smaller in secured lending, where collateral and loan-to-value considerations play a major role.
Behavioural scoring also cannot compensate for fundamental weaknesses elsewhere in a lender’s business. A poorly priced financial product, weak collections process or unsuitable credit policy cannot necessarily be corrected by adding another data signal.
The appropriate way to understand the technology is therefore as one layer within a broader credit-risk framework.
How Lenders Can Validate Behavioural Credit Scores
Model validation is a critical part of deploying alternative credit data.
Traditional credit models can often be backtested against historical records because the relevant personal information can be matched with past outcomes. Credolab says its approach is different because the company does not hold the personal data needed to perform that type of backward-looking matching.
Instead, the company describes a forward-validation process.
Under the approach outlined by Credolab, the SDK can operate in shadow mode for as long as three months. During that period, it scores real applications in real time, but the scores are not used to make lending decisions.
The lender then compares those scores with actual outcomes as the relevant loans mature.
This approach is intended to test performance against the conditions the model will encounter in production rather than relying exclusively on an older historical population.
Once validation has been completed, Credolab says the lender can activate the score within the existing system rather than treating production deployment as a separate large-scale implementation project.
What Model-Validation Information Is Provided?
Credolab says its clients receive a detailed scorecard report covering model development and validation.
The reported documentation includes:
- The agreed target definition
- Training, testing and out-of-time samples
- Performance for each sample
- AUC
- Gini
- KS
- F1
- Score-density curves
- Bad rates by decile
- Gains analysis
- Calibration plots
- Cut-off analysis
- Confusion matrices
- Information value
- Weight of evidence by bin
- Variance inflation factor (VIF)
- Stability across five time slices
- A predefined retraining trigger
The target itself is agreed with the client before model development. Credolab describes the outcome as measuring behaviour ranging from first-payment default through to 90 days past due at nine months on book.
This level of documentation is important because alternative-data models face the same fundamental requirement as other credit models: lenders need to understand how the model performs, how stable it is and how its variables contribute to risk estimates.
Explainability and the Risk of Proxy Variables
Alternative credit scoring also raises questions about fairness and explainability.
Credolab says its approach to explainability is statistical rather than relying on narrative descriptions attached to an output after a score has been generated.
The company says every feature’s direction and magnitude of risk can be examined by bin, with monotonic ordering applied across time slices and performance measured against data that the model did not previously see.
Another important issue is whether behavioural variables could become proxies for protected characteristics.
Credolab says it does not hold race, gender, ethnicity, address, postcode or other personally identifiable attributes in its scoring environment.
The company also reports that its score typically has a correlation of less than 5% with other data sources used by lenders, including bureau, cash-flow, transactional and socio-demographic information.
That finding is presented by Credolab as evidence that behavioural scoring provides an independent signal. However, the presence or absence of particular demographic variables does not by itself eliminate the need for lenders to conduct their own fairness, compliance and model-risk assessments.
The Financial Inclusion Question
One of the strongest potential applications for alternative credit scoring is lending to people with limited conventional credit histories.
Traditional credit systems can struggle to evaluate applicants who have little or no recorded borrowing history. For these consumers, the result can be manual underwriting, a higher-priced product, limited access or rejection.
Behavioural data can provide another source of information where sufficient digital activity is available.
Credolab says the improvement reported in its neobank evaluation was smaller for thin-file and new-to-credit applicants than for thick-file customers. Yet the company argues that the practical significance can be greater because applicants with limited credit histories may otherwise have fewer pathways into formal credit.
This is one reason alternative data has become an important area of financial technology research.
Nevertheless, financial inclusion depends on more than producing another credit score. Consumers must also have access to affordable financial products, transparent terms, appropriate identity processes and responsible underwriting.
Credolab’s Scale and Global Expansion
Credolab says its technology draws on nearly 80,000 behavioural data points across Android, iOS and web.
The company reports serving nearly 200 clients across the United States, Latin America, Southeast Asia and EMEA, with offices in Singapore and Miami.
Its history also illustrates the wider internationalisation of alternative credit technology. A company founded in Singapore in 2016 is now being made available through a financial technology marketplace connected to a credit-scoring platform used across more than 80 countries.
The development reflects the growing role of specialised fintech providers within established financial infrastructure rather than a simple replacement of traditional banking systems.
What the FICO Marketplace Listing Means for Fintech Lending
The significance of the FICO Marketplace listing is partly about distribution.
Financial institutions are often reluctant to adopt unfamiliar data providers because implementation, security review, model validation and integration can all create additional operational work.
Being available through an established platform can reduce the discovery and integration barriers. It allows risk teams already working within the FICO ecosystem to identify an alternative scoring provider within an environment they already use.
But access to a marketplace does not eliminate due diligence.
Lenders still need to consider data governance, regulatory requirements, model performance, fairness, security, customer consent and whether the score provides meaningful incremental value for their particular portfolio.
For alternative credit scoring companies, established financial platforms can therefore provide a route into larger institutional markets while preserving the need for independent scrutiny.
Future Outlook for Behavioural Credit Scoring
The future of behavioural credit scoring will likely depend on how lenders balance the potential value of additional data against increasingly important requirements around privacy, explainability and responsible lending.
For applicants with limited credit histories, alternative signals may provide information that traditional credit files cannot. For customers with established files, behavioural data may offer another dimension of risk information rather than simply filling a missing-data gap.
At the same time, behavioural scoring cannot solve every credit-risk problem. It does not determine affordability, replace income verification or eliminate the need for established underwriting controls.
The expansion of Credolab through the FICO Marketplace illustrates a broader development in fintech: alternative data is moving closer to mainstream lending infrastructure. Its long-term significance will depend less on the novelty of behavioural signals and more on whether lenders can demonstrate that those signals are accurate, stable, explainable, appropriately governed and genuinely useful for responsible credit decisions.
Conclusion
Credolab’s entry into the FICO Marketplace marks another step in the integration of alternative credit data into established lending technology.
Its behavioural scoring model seeks to add information derived from device and application interactions to conventional credit-risk systems. The company says its technology can be deployed through existing digital channels and reports performance improvements when its score is combined with bureau data in a regional neobank evaluation.
The approach also has clear boundaries. It does not measure affordability, does not replace income verification and is limited to digital application environments. Data governance, validation, fairness and regulatory scrutiny remain central to its use.
For the broader fintech industry, the development points toward credit models that increasingly combine multiple forms of information rather than relying on a single traditional credit file. The key question will be whether those additional signals can expand access to credit while maintaining transparency, privacy and responsible underwriting standards.
