Welcome to our blog series on decoding identity resolution. This is a nine part blog that offers an attempt at a friendly, comprehensive view of how to think about the concept of identity resolution as well as how to interpret the way it is represented in marketing and sales materials by different companies across the tech landscape. The other articles in the series can be found here:
Part Eight: The Amperity Perspective
Introduction
As this is a blog from Amperity, I want to set aside a moment to be clear about how Amperity solves for identity resolution differently. While most companies in the data management or CDP space treat identity resolution as an afterthought, Amperity built our entire platform around solving what seemed to be an unsolvable problem.
Below are some main areas that Amperity differentiates from the market around identity resolution.
Identity resolution, not “match and merge”
A common term in the industry for identity resolution is “match and merge.” The vast majority of software doing any sort of identity resolution combines identity resolution and what to do with the resulting identity into a single process. We believe that is incredibly limiting.
Amperity generates an identity graph that shows raw data and all of the links our algorithm establishes. This is a flexible data asset that can be used for everything from marketing, to compliance (CCPA and GDPR), to enabling data science teams to build custom algorithms.
The “merge” logic is performed separately, which enables a radical level of customization and transparency.
Identity resolution overview
First of all, if this series has been compelling to you and you want to learn more about the Amperity platform, please reach out to us and we’ll be happy to give you a deep dive into how our identity resolution process works. But since you’re here, what follows is a high-level overview of what Amperity does:
We ingest raw data at any scale — no expensive ETL development or data prep necessary.
We apply a layer of semantic tags to the raw data, instructing the platform how to think about your data. In other words, we tailor the platform to your data, compared to the market norm of expecting you to do the opposite.
Our AmpID process uses semantic tags to create a standardized table with all records in it.
We break the massive single table into thousands of “blocks” so that we can distribute the processing over hundreds of computers to provide maximum speed.
We intelligently choose from up to 45 different machine learning algorithms based on the available semantic tags and run them to compare blocks of data and generate billions of “links”.
Each link connects two of the original records and provides a clear weight to indicate our confidence in the link.
We break the links into “clusters” and apply a threshold to ensure that every match meets a minimum level of confidence.
We assign the clusters an “Amperity ID” and propagate that ID back to all of the original raw data tables.
We create a growing library of “unified” tables from the results, each crafted toward providing a new layer of transparency or simplifying common use cases.
The result of all of this is a graph and a UI full of tools to take advantage of it. We will use industry-proven models to merge the graph into the best possible profile, and layer attribute models on the newly standardized data for a growing library of common breakdowns marketers want.
Radical transparency
The majority of solutions for identity resolution you will encounter in the market do not provide a clear data lineage to understand how matches were arrived at. You put a bunch of data in and get less data out, and you trust that it was merged well.
Amperity believes that isn’t good enough for enterprise companies. The AmpID graph makes it easy to trace all data back to where it was acquired to better understand how different datasets impact the results. For example, if you are using a third-party data asset, you can clearly see which parts of your Customer 360 profile came from the third-party versus directly from your customers.
Multiple patents
“Show me your patents” should become a more common part of any sales motion or RFP. If a company can’t show you patents they probably haven’t solved a unique problem. Amperity approaches identity resolution as an academic problem — we have earned multiple patents on our approach to identity resolution, published scientific papers on the topic, and also earned patents on how we handle change management for a live system.
See more here: https://amperity.com/patents
Comprehensive Identity
In this series, we have talked about rules-based identity, digital identity, and advanced data science identity. The good news is that Amperity can handle it all.
While our default is to use our advanced ML models for identity resolution, we support deterministic models. We also support a layered use of both when use cases call for it.
There’s no restriction to the type of data that you can bring into Amperity. We take raw clickstream data from Adobe Analytics or Google Analytics and make digital identity just another part of the customer profile.
We also support using third-party licenses in a regulatory-compliant way to enhance your data or activate it in ad-tech focused channels.
There are many use cases for customer data, but we believe that they all depend on getting the data right first and foremost, and building from there. Frankly, if something calls itself a CDP, it needs to be focused on customer data, not just any data, and it needs to be a platform for building on.
Next Up
All good things must come to an end, including this blog series on identity resolution. In the final installment we will wrap this series up with some closing thoughts. Click here to advance to part nine!