Customer Data & Analytics Blog

Marketers: do you know about look-alike modeling for segmenting your contacts?

Sandeep Koul | 5 minute read

Zylotech_Marketers_do you know about look-alike modeling for segmenting your contacts_020519_Header

Through an example, this blog will describe a technique called look-alike modeling that can help marketers use data enrichment from sources they may not know about in their CRM to expand on their segmentation. Marketers can get better results when using this lesser known way of targeting lists.

Generic segmentation pitfalls

Although most marketers have lots of data sitting in their CRM, they often use only a few demographic sources. This approach shows inefficiencies. Usually based on assumptions, marketers doing this are mostly guessing at most on the attributes. For example they may assume a particular industry and title combination would perform best for their campaign. But this method is not based on data. Therefore, there is no data backed evidence about the positive relationship of the chosen attribute (industry or title) to a desired outcome (better engagement).

Another way for marketers to segment is based on mathematics and is executed by a CRM. In this approach, you’re asking your CRM to figure out which of its contacts mathematically ‘look-alike’ your best performing data. You’re trying to find out those contacts who have a better probability of behaving in a desired way (i.e. have a higher probability of engaging with your marketing campaigns). This approach is called look-alike modeling.

Look-alike approach in action: step-by-step

Usually marketers need help from data scientists to both build look-alike models and implement them within your marketing stack. But it’s possible for marketers to build look-alike models on their own just by optimizing features within their CRM. 

Step 1. To activate this approach you first start by collecting data. It’s important to include all of your data from behavioral data to demographics. If you’re wondering where to find lesser accessed data sources like behavioral data then the good news is that most marketing automation tools already Zylotech_Marketers_do you know about look-alike modeling for segmenting your contacts_020519_subcollect information on how audiences are behaving in your campaigns. You can use these data points to arrive at the second step: defining ideal contact profiles.

Step 2. To define the attributes and behaviors of ideal contact profiles, you need to know your campaign objective. Let’s start with an objective of better engagement from a relevant audience. By the end of this activity, you should arrive at the list of contacts in your CRM that have a high enough probability of engagement while being relevant.

Starting within your CRM, you want to find out the key attributes of your existing customers. Here’s an example that marketers of a B2B product may conclude after analyzing their existing customer base.

1. Most of customers are from the retail industry

2. Account revenue is more than $100 million

3. Accounts are mostly headquartered in North America

4. Majority of customers are leaders in the marketing and technology functions

5. Marketing leaders end up engaging more with your campaigns. You can arrive at this by finding out how many interactions they had with past campaigns. For example if you’re planning a marketing email campaign then you can find out past contacts who have either opened, clicked, or responded to emails.

Here is a sample 10 customers mapped to an ideal profile:

ID

Industry

Revenue

Region

Job Title

Engagement

Ideal Profile

1

Retail

$150 m

NAM

CMO

10

Yes

2

Retail

$200 m

NAM

Marketing Director

8

Yes

3

Hi-Tech

$500 m

EMEA

CFO

3

No

4

Food

$300 m

NAM

Director

4

No

5

Food

$200 m

NAM

Director

5

No

6

Hi-Tech

$150 m

EMEA

President

6

No

7

Retail

$50 m

EMEA

Director

5

No

8

Hi-Tech

$300 m

APAC

President

6

No

9

Food

$150 m

APAC

CEO

6

No

10

Retail

$250 m

NAM

CIO

7

No

Step 3: The last step is to turn on the filtration functionality in your CRM to arrive at a list of all the contacts that are marketing leaders in retail companies located in North America with $100 million or above revenue. This list can be used for your next marketing campaign.

As an alternative, marketers can also use a workflow functionality for automating this process. Workflows are a set of simple rules you can define within your marketing automation tool that trigger on specific events. Here you can trigger a workflow on new contact creation that would tag it as an ideal profile by evaluating industry, revenue, region, job title, and engagements.  

The outcome of look-alike modeling is a data driven segmentation approach resulting in a list of contacts with a higher probability of showing desired outcomes to your marketing messaging when compared to generic assumption-based segmentation. In-turn marketers receive better campaign results.

Sandeep Koul is a senior marketing operations manager at Zylotech where he brings 10+ years of technical marketing experience following his marketing studies at Indian Institute of Management, Lucknow. When he’s not working, he enjoys driving or trekking in the Himalayan region, reading books, and using his camera. Writers like George Orwell, Franz Kafka, & Fyodor Dostoevsky help him make sense of human behavior.

If you liked this post, check out our recent blog post: Personalization versus relevance for today's customer.

Topics: Customer Intelligence