Making manual review faster

Making manual review faster

Ayush Rodrigues

Oct 21, 2024

People, including business owners, are very impatient. If there’s a product I want to use, I really don’t want to wait 2 days to be onboarded.

While a large majority of applications can be automated, manual review will (and should) remain part of any healthy fintech’s risk processes.

Here’s what we’ve learned about managing it effectively. Let me know what you think!

Sources of Data

The more data you have, the more likely you can make an automated and fast decision. Here’s a list of data sources that are frequently not used but have big impacts:

  • Social Media Accounts: Twitter and LinkedIn are often accessible, especially with a business email. Facebook and Instagram are harder to access but can also be useful. Often this helps validate that a person is not someone who’s partially matched a sanctions screening hit.

  • Business Registers: official sources of company information can be pulled directly. This means that you don’t need to ask the applicant for it (who frequently can provide wrong information).

  • Google Maps: satellite and street view imagery can be used to verify physical locations of businesses and residential addresses

  • Customer Reviews: customer feedback and ratings from G2 and Trustpilot are useful in assessing business reputation and operational history

  • Website Profilers: analysis of company websites can be done via companies like Webshield and Truebiz to give a legitimacy score. Make sure the industry matches what you were expecting.

  • Sanctions, PEP, and Adverse Media: hard to deal with as there are often partial/fuzzy matches that are false positives. New AI agents (Y Combinator has funded several) may be useful in reducing noise.

Often there is a cost to this data, so it’s useful to think in which higher-risk scenarios can more data be used.

It's worthwhile investing in a case management system that can centralize this data, so an analyst to quickly make a decision. There’s nothing worse than having 10 tabs open to research one case!

Business vs. Consumer Onboarding

Your thresholds for success are different depending on who you’re onboarding.

Business Onboarding

Even large tech companies like Gusto and Mercury typically have a manual review rate of around 50% for business customers.

Business onboarding often requires more thorough checks due to complex ownership structures and higher potential risks.

It’s typically manual to start, with companies implementing some automated approval flow once they reach 10–30 cases per week.

Consumer Onboarding

Consumer applications typically have a much lower manual review rate, usually around 2–5%.

These rely primarily on KYC and sanctions checks, with less data available compared to business onboarding.

Due to high volume and expected low-risk profiles, consumer apps often require automatic approval processes from the start.

What the best companies do

The part you’ve been waiting for. Here’s how we help companies that use Recase.

Start manual and automate over time

When launching a new product, entering a new market, or initiating a new marketing campaign, start with a higher rate of manual reviews. This approach allows you to:

  1. Gain confidence in what your typical customers look like

  2. Identify patterns to refine your automated processes

As you gain experience and data, gradually increase your automation rate. Repeat this process when you’re launching in a new market, a new product (did I hear you say instant ACH?) or even a new marketing campaign. Different onboarding groups will have very different risk profiles.

Collect ALL information upfront where possible

  • Use dynamic forms that adjust based on user inputs to collect as much relevant information as possible during the initial application.

  • Implement a stepped approach that increases friction only when necessary, based on risk factors or data gaps.

  • Use AI/OCR technology to validate documents have the right information before reaching the analyst.

Follow-up only if absolutely necessary

  • At companies like Stripe, analysts are trained to exhaust all available resources before reaching out to customers for additional information.

  • Customer follow-ups can delay the process by days and may lead to application abandonment.

  • Only contact the customer if necessary and when the analyst is 100% certain that the additional information is critical.


Having access to more of the data sources we described above can often help with this.

Final thoughts

There’s no silver bullet. Onboarding conversion is important, so you’ll have to start somewhere and refine it over time.

Often the real fun starts once they’re onboarded. Good luck!