The Demyst Team

The Demyst Team

Predicting Customer Trajectory with External Data

As unemployment soars to historic highs, risk professionals are seeking to predict the financial trajectory of their customers in order to mitigate losses and inform go-to-market strategies.

Most financial organizations have established task forces to mine their historical 1st party data (losses, CRM, transaction data), and perhaps credit scores from a preferred bureau to answer key questions, such as:

  • What is the impact (+/-) on industries, by location, over 3, 6, 12, 24 months?
  • Which employers will need to lay off staff or default, and which will emerge stronger from the crisis?
  • Which customers are likely to be impacted by layoffs, and which will emerge stronger from the crisis?

However, leading risk professionals are recognizing that current disruptions require looking beyond internal data to understand and respond to emerging customer trajectory. Historic models are no longer predictive, and traditional bureau data is too stale for risk professionals to adequately implement near and longer-term responses.

We know that it typically takes a large bank 8–14 months and $2–3M to find, onboard and operationalize a new external dataset, resulting in an over-reliance on internal data and/or on 1–2 legacy vendors (typically bureaus). Below we highlight a 4 step, 6-week approach, that any banking or lending organization can take to score their portfolio and set themselves up to emerge stronger from the current scenario.

  1. Prioritize business objectives ;

Now that the dust is settling from organizational and business continuity shocks brought on the initial shock of COVID-19, organizations are evaluating how they could and should respond. The ability to predict customer trajectory is critical for driving marketing, acquisition and risk strategies as below :


  • Establish guard rails for new applicants by
    a) limiting exposure to at-risk segments
    b) consider employer risk in customer lending (recent or expected layoffs)
    c) implementing stricter fraud controls
  • Score and monitor existing customer portfolio in order to
    a) mitigate near-term exposure to high-risk/low growth segments
    b) update loss provisions and capital allocations
  • Identify and engage low-risk/high growth segments likely to survive or emerge stronger from the current crisis


  • Enable fully digital and streamlined origination workflows
  • Replace periodic reviews with continuous and/or event-based monitoring
  • Rebuild risk strategies to shift portfolio towards favorable risk profiles
  • Deploy capital towards market growth channels

2. Discover the best third-party data

Below is an overview of the key categories of (non-FCRA) information that are used in predicting employer and customer risk and trajectory.

Employer Trajectory

  • Location/industry risk : Consumer Spend, Footfall (vehicle and individual), UCC Filings, COVID Exposure score
  • Employer Risk : Credit Scores (multiple bureaus), Delinquent payments, Business spend, Company Size
  • Layoff Risk : Employee changes, recent layoffs and announcements

Customer Trajectory

  • Income/assets : W-2 Income, Debt, Discretionary Income, Property ownership, value, mortgage status, Vehicle ownership
  • Social Capital : Employer, Job Title & Description, Employment History, Education & Skills, Location History, Associations
  • Credit : Non-FCRA credit scores, Aggregated credit files and trends
  • Family Exposure : Immediate family member employed by small business

Please contact us for free sample data and/or vendor recommendations across the above data categories.

3. Configure the data and scores to meet your needs

Once a shortlist of signals have been selected, we want to translate this data into insights i.e. a series of flags and/or configured scores (employer risk score, individual risk score, customer trajectory score) that can be applied to any customer on demand.

To do this, third-party data is appended onto customer outcome data i.e. take a set of customers that have grown, become delinquent or not changed in the past 3–6 months, enrich with these profiles with monthly or weekly snapshots of the key external data signals, and then analyze which attributes are predictive of movements.

Given the appended third party is already structured and joined to customer first party data, this step generally takes 1–2 weeks, and can be iterated between Demyst and client risk teams.

Optionally, Demyst has curated the top-performing attributes from 9 different data vendors for predicting customer trajectory. The raw data and clear box scores can be appended to customer files, and then easily optimized by in-house analysts to meet specific cost/fill/lift/risk requirements — saving 90% of the upfront analytical time and effort.


Please contact us for free samples of this solution.

4. Deploy and optimize over time

Once you’ve selected the best combination of attributes that meet your business and budget requirements, you’ll receive a single order form to license all of the selected third party data. No need for multiple MSAs and security reviews. Demyst will also provide all the legal, compliance, security and commercial information for internal reviews and approvals.

The bespoke data solution (scores, as well as underlying raw data) can be configured and deployed into your preferred architecture, data exchanges or workflow tools to minimize internal IT effort. Example implementation:


We are in a period of unprecedented change and uncertainty. Risk professionals not only need to leverage more up-to-date signals from verified third-party data sources to streamline and de-risk decision making, they also need to ensure that they have the flexibility to quickly respond to changes in the market and their portfolio. Once deployed, our clients are able to constantly evaluate new data and re-tune the scores based on customer outcomes, with minimal friction.

Don't settle for half the story

Demyst gives you access to all of the data you need. Evaluate thousands of data attributes from hundreds of possible data connectors all pulled into your own custom-built APIs for instant data deployment.