Approach to Understanding User Better: Introduction

Corry Lamria
3 min readMar 14, 2023

There are five chapters related to entity resolution topics. Chapter 1 is background on business implications associated with resolving entities.

First, let’s address the most common questions decision-makers often ask during business review sessions. We lost count of how often questions here were asked, on how to optimize marketing costs, why questions behind monthly active users drop, and what actions should take to improve engagement numbers and CLTV. The problems we encountered here are far more common than most companies care to admit and are often difficult to spot.

We will likely start an analysis using internal information and conduct benchmarking for validation. In the absence of poor data quality, we struggle to resolve multiple labels for one user. It is even more difficult with fragmented data silos in the database or inconsistent data due to multiple systems. Moreover, as the company values data-driven decision-making, we are more likely to add assumptions to close the gap from missing internal data about customers’ needs and preferences across their journeys.

Why should we care about the unified view and fill in the missing information?

Pushing our business into the digital world will add more complex problems, mainly when the user gives limited information to our company. They can create a fresh and new user account, whatever the intention is. We do not know what we do not know. Let’s look at three significant cases that often experience challenges and problems:

Multiple Identities Lead to Ineffective Acquisition

Distinct data silos that failed to be predicted into a unified view of the entity led to ineffective acquisitions. Suppose we are marketing analysts and want to create a campaign that attracts users to use specific promos. User A has three accounts, and what usually happens is that the three accounts are aimed at the same promo. The actual reach is only for one person, making the impression useless because other accounts may need to be more active.

Interact Often with Similar Fraudsters

Failing to match fuzzy records in the internal database led to higher false positives from the fraud detection system. From the point of view of risk analysts, fraudsters usually work with other fraudulent to find ways to abuse the system for their interests. The false positive from fraud rule-based in the company led to high fraud budget. Apart from working together, fraud people generally create multiple accounts to abuse the promos given. If this happens, the company will suffer financial losses.

Dealing with Partial information

Recommendation engines using distinct data silos lead to lower predictability and accuracy. Imagine we are a data scientist who wants to create a recommendation engine that we can later use for cross-selling or increasing sales of certain items. But it turns out that user A has several accounts, each with a different behavior. Impartial information undoubtedly is a problem for our recommendation engine because these accounts provide partial information. The model only generalizes the problem due to the uncompleted user journey perspective.

What if there is an ideal condition for resolving user entities, providing a single view from the plethora of data, determining users’ patterns across their journey, and accessing their preferences? Identifying links between seemingly unrelated entities through data analytics would lead to better customer engagement, happier customers, and ultimately higher returns for your business.

Oppna will help your business to get a better view of your users just in one hit!