Skip to main content

SMB Underwriting

Overview#

Heron Data’s main use case is to help lenders and fintechs underwrite SMBs using bank data. We do this by enriching each transaction with a category identifying the transaction (e.g., Revenue or Payroll), and then aggregating amounts across these labels to give insights into the company.

The key outcomes we achieve for our customers are:

  1. Supercharge underwriters: Speed up underwriting by an order of magnitude vs. having to manually parse bank data
  2. Expand addressable market: Underwrite businesses without up-to-date accounting or e-commerce data
  3. Catch fraud: Because bank data is up-to-date, we have enabled customers to avoid millions of dollars in losses to customers who were loan stacking or had other previously unidentified risk factors

We will begin by enriching the bank data, and then looking at how it can be used to analyse companies from the dashboard and via the API.

Post & enrich company bank data#

Create an end user and enrich transactions#

  1. Create an end_user: Begin by creating an end_user in Heron’s systems that corresponds to a company/applicant in your systems. You do this by sending a POST end_users request.
    1. For the end_user_id field, use a canonical reference/identifier for the company
    2. Note: If you are sending transactions for a company that you’ve already sent to Heron before, skip this step.
  2. Post Transactions. Once you have created the end_user, you can start POST’ing transactions for that end_user. Make sure that the end_user_id in the payload matches the end_user_id you created in 1). You do this by sending POST /transactions requests.
    1. As noted, please batch to 1000 transactions or fewer
    2. Note: If you receive transactions directly from Plaid, Ocrolus or in PDF, we allow you to just pass on the file without any manipulation. This replaces using the /transactions endpoint in this step. The endpoints you can use are:
      1. Plaid - Assets Report
      2. Plaid - Transactions Report
      3. Ocrolus Report
      4. PDF
  3. Process Transactions: When you are done sending us transactions for a company, please send a PUT end_users request, indicating that the end_user is ready for processing.
  4. Listen to webhook: We will notify you via a webhook when the end_user_id is processed, and available for you to retrieve. You can configure your webhook in the dashboard.
  5. Get transactions: Once you have received the webhook, you can send a GET /end_users/{end_user_id_or_heron_id}/transactions request to retrieve the enriched data.
    1. Note: We still support the /transactions endpoint that was previously used by most customers. The new endpoint that is now standard is much more performant, so we recommend using that endpoint instead. If using the old endpoint, you can use the end_user_id parameter to ensure you only pull transactions for the end_user_id that was just enriched. If you get transactions for a company that you’ve already sent to Heron before, you can use the last_updated_min filter to only get transactions where labels have changed since the last time you send and fetched transactions.

You now have enriched data for a given company. Depending on your use case, you may just want to display this enriched data back to your data science team, or use it in other internal applications.

Note on webhooks

From the Heron dashboard, underwriters are able to request a review for a company when they believe that there are errors with Heron’s classification. When the review is finalised, we will trigger a reviewed webhook similar to the example in 4). If your organisation only consumes Heron’s output via the dashboard, you don’t have to do anything else. If you pull raw data into your systems, please make sure that you repeat step 5) after receiving a webhook with the status reviewed.

Inspect company using the dashboard#

This section only describes the dashboard at a very high-level. To get more detailed information, please contact Heron Data for an onboarding session

During your integration, you can use the dashboard to make sure that all steps work as expected.

  1. Navigate to the Companies page. This will give you an overview over the companies (i.e., end_user) you have created.

Companies Page

  1. You can see the status of the company in the status column of the main table. To inspect the transactions uploaded, press transactions on the right, then view transactions.
  2. To get a sense of the possible visualisations on top of the enriched data, navigate to Summary for one of the companies. Here, you will see Revenue, P&L, risk flags and other financial metrics about the company.
  3. To verify all possible categories that your model may attach to transactions, navigate to the Categories tab. You can inspect all labels there.

Categories Page

Retrieve aggregate metrics on a company#

Instead of consuming the enriched transaction data, it may be easier to consume aggregated metrics for a given company. For example, you may want to see monthly historical revenue for a company, as estimated by Heron Data.

We allow you to do this for aggregate metrics (our Heron Scorecard) and historical time series and forecasts per category label.

Get Heron Scorecard#

We have aggregated the most common metrics calculated on top of Heron Data output in our Heron Scorecard. The Scorecard will allow you to quickly evaluate a company, by providing labels such as:

To get the Heron Scorecard, send a GET end_users/{end_user_id_or_heron_id}/scorecard request.

  • Data Quality: Measures to help you understand the quality and completeness of the bank data our analysis is based on, e.g. Data Coverage & Data Freshness. Also includes a measure to see what percentage of all transactions (as a proportion of revenue) appear to be going to an unconnected account held by the company we are analysing.
  • Risk Flags: Flags like ATM Cash Withdrawals and distinct days with insufficient funds / overdraft fees.
  • Balance: Statistics about average, max and min balances.
  • Profit & Loss: Approximations of standard cashflow-based metrics like Revenue, Net Operating Cashflow, etc.
  • Debt: Number, recency and amounts of recent debt investments and repayments

Get historical timeseries#

To get historical time series, (e.g. Revenue over time for each month), send a GET end_users/statistics request. In the parameters, you can specify:

  1. The end_user_id of the company you are looking at
  2. The time periods over which you want to aggregate, often this is per month (date_granularity)
  3. Whether you want to pivot by merchant, category or both
  4. If you want to see only certain categories, filter by category_heron_ids
Example

To get the monthly revenue for end_user_id 12345 you would send a GET request to the following:

https://app.herondata.io/api/end_users/statistics?end_user_id=12345&group_by=category&date_granularity=month&category_heron_ids={category_heron_id_for_revenue}

Get forecast#

Note: Forecasts are currently in beta

To get forecasts for a category timeseries, send a GET end_users/forecast request. You can use similar logic to the historical timeseries section above to get forecasts for the time intervals and category labels you care about.

For more information on forecasts and any feedback, please talk to us!

Edit this page