Customer Data & Analytics Blog

5 Things Marketers Should Know When Pulling Together Customer Data

Jeff Whitney | 3 minute read


Marketing is becoming increasingly innovative, not just in its campaigns, but also with its use of customer data. Today enterprises have, on average, over 50 forms for unique customer data, therefore it’s imperative for marketers to look beyond IT and understand the issues and trade-offs associated with pulling together customer data.   

1. Finding Quality Data – The foundation of Quality Customer Marketing

Smart companies know that retaining and monetizing their customer base is crucial to their success.  And they know that the only way to do this effectively is to understand each customer on an individual level, including their psychographics, demographics, past behavior and current intent.

Understanding individual customer behavior isn’t easy.  It’s hard enough to understand the behavior of your friends and family and it’s doubly hard to understand the behavior of thousands, or millions, of fast-moving customers.   To do this, you need all the help you can get. Help in this case means pulling together the appropriate data on each customer, and doing it in a timely manner. Finding those attribute and fields is very crucial and cannot be left for IT to decide. 

2.  Pulling together quality customer data fields

Pulling together customer data isn’t easy:

  • Data resides in dozens of data sources that exist both inside and outside your company’s walls (e.g. marketing cloud, social, email, online, mobile, billing, CRM…)
  • It exists in both structured (e.g. billing system) and semi-structured forms (e.g. emails, clickstream, social media posts)
  • It lacks a consistent, unique identifier (e.g. customer #, social security number)
  • Digital data is very messy, yet very rich, and tieing it together with your known customer is crucial 

3.  There are two methods for connecting customer data – Deterministic and Probabilistic

Even if you are going to have a team of data engineers and data scientists pull your customer data together, it’s important to have a basic understanding of the pros and cons of each approach, as well as an understanding of how they work. Let’s take a look: 

Deterministic Matching

Deterministic Matching looks for an exact match between two pieces of data. This could be a customer number, social security number, or driver’s license.   But, collecting this information often isn’t practical – and even if you try, customers frequently baulk at sharing such sensitive information.

As a fallback, the majority of deterministic implementations rely on matching several data elements – such as name, address, phone number, email address, date of birth, and gender.  Each is matched separately and the results are tallied to create an overall match score.    

Probabilistic Matching

Probabilistic Matching uses a statistical approach to calculate the probability that two customer records represent the same customer.   Unlike deterministic matching, it uses a broad range of data elements.

 Pros & Cons of Each

Deterministic can be easier to set up. But with limited field matching, it can result in more false negatives (failing to match records from the same customer).  It can also be prohibitive when looking for exact matches over a large sample of data within a realistic time period.

Probabilistic matching is generally more difficult to set up, but is likely to match more customer records. The downside here is that you may end up with a handful of false positives, as well.

To leverage their different strengths, it is sometimes advisable to use both methods. 

4. Catch 22 – Too Little Data, Too Late

Regardless of the matching approach, marketing must rely on a team of data engineers and data scientists, who must have a strong working knowledge of the matching techniques and the company’s customer data.

Unfortunately, even with a quality team, applying these methods takes time. In fact, study after study shows that on average it takes multiple weeks, and sometimes months, to curate just 10-15% of all relevant data from each target customer.  Marketing is left with too little data on each customer that is delivered too late to support effective customer marketing initiatives. 

5. Artificial Intelligence and Machine Learning to the Rescue

The good news is that artificial intelligence (AI), and machine learning are a perfect match to combat the challenge of connecting customer data.  AI and machine learning enable machines to apply both deterministic and probabilistic methods to customer data more effectively and at a much faster rate than humans ever could.  So instead of 10-15% of data in weeks or months, you can leverage almost all relevant customer data in near real time.

In essence, AI and machine learning systems, like Zylotech, resolve the current mismatch between limited human capacity and the tidal wave of available data.  They allow marketers to make use of the relevant customer data they need, exactly when they need it. Machine Learning can be coupled with human guidance to guide future processes (i.e Matt Hurley or M Hurley or Mathew Hurley can be all trained as Mathew Hurley). 

We’d love to hear about your challenges connecting customer data or discuss how Zylotech can help you better understand your customers.

Topics: Customer Intelligence