Announcing the Deduplication Model

We’ve noticed that data from aggregators including Plaid and MX return duplicate transactions across multiple accounts for a user. These erroneous transactions result in incorrect insights, features, and scores. As a result, we updated our model to remove duplicate transactions while computing insights to correct these duplicate errors.  

Why this matters

We found that 3.6% of transactions were duplicated by these data aggregators. In some cases, we saw transactions duplicated up to 5 times!

When duplication errors occur, individuals can be seriously impacted. For example, if an income transaction is duplicated multiple times, the total income that shows up for a user will be wrong by a factor of 2, 3, or more. Someone’s income could appear to be $500,000 when it’s $100,000! Deduplicating transactions means that insights, features, and scores will be more accurate as a result.

Here are some FAQs to get you started.

How are duplicate transactions from Plaid created? 

A duplicate Plaid Item is created if the end user logs into the same institution using the same credentials again using Plaid Link, and if an access token is requested for the item.

A user may do this when they:

  • forget that they’ve already connected to a bank
  • may believe they have to connect twice to add multiple accounts (like checking or savings)
  • are trying to get past an error

Therefore, a duplicate account with duplicated transactions is ingested by Plaid. 

How can I prevent duplicate transactions from being sent from my application?

You can read more about preventing duplicate transactions here

What should I do if I’m unable to resolve duplicated Plaid Items in my application?

No action is required! Our deduplication model will remove the duplicated account and its accompanying duplicate transactions. 

Will Pave resolve duplicate accounts in addition to duplicate transactions?

Yes, we’re working on further improvements to resolve two or more Account IDs from Plaid that are actually the same Account.