Superior Customer Experience is only Possible with Good Data: Time to Establish a Dedicated Customer Data Team?

Most companies have put in place a marketing automation system and integrated it to their CRM. They’ve also designed and implemented a few automated customer journeys like newsletter subscription and lead generation. The next step, more complex data-triggered marketing programs, seems to be harder to take. The reason for this is the lack of good quality data.

Besides not being able to run always-on marketing programs, less than good data causes many other problems: Predictive analytics scores (e.g., lead scoring, propensity to buy, churn risk) cannot be calculated accurately, and many functions cannot do their work properly because they don’t know the customer well enough.

This is a shame because the other three building blocks of data-driven marketing are coming nicely together. Younger­­–more tech-savvy marketing professionals–are entering the workforce, tools are becoming more mature, and processes are improving with system thinking and built-in customer journey -tools.

Graphic 1: The four building blocks of data-driven marketing

Even with all these components in place, the marketing team still fails to implement more complex ways to communicate with leads and customers. Maybe the problem isn’t in the tools or processes or people. Maybe the problem is in the focus and specialization. Maybe it’s time to establish a dedicated Customer Data Team (CDT).

A dedicated team needs a clear mission

Establishing a dedicated CDT is a substantial investment, and this means that the team’s mission must be crystal clear from the start. Below is a set of 7 core functions of the CDT team.

1. Brainstorm use cases

Your marketing team should design the always-on (data-triggered) marketing programs, but they might need some help to do so.

The CDT can assist the marketing team by teaching them the principles of data-driven marketing and then facilitating a series of workshops to design the always-on marketing programs.

2. Model the data

There are two approaches to collecting customer data:

• extract sufficient data from your marketing, sales, support, and other relevant systems and then transform it for your purposes, or
• define the data model that is required to execute your always-on marketing programs and collect only that specific data.

Both approaches have benefits and risks.

The former will give you a lot of data that you don’t need while excluding some data that you might need. In other words, some thought and effort are needed to make sure you’re exporting the right data. With the latter approach, you might miss out on data that you need in the future for a new, as-yet-unknown, use case. You can always ask for more data later, but your business information (BI) teams might not be too happy about creating new exports, and it could take a while until you get your hands on the additional data.

I recommend starting with the latter approach – a “just-what-you-need” dataset.

3. Collect the data

You need to be able to collect all the needed customer data in your marketing dataset, which needs to be controlled by marketing and cannot be part of your organization’s common data lake or some other IT or BI-managed system. Centrally managed data depositories invariably fail to deliver what marketing needs.

Customer data can be divided into three main types: identities, attributes, and events.

• A single customer can have many identities. For example, customer’s social media userid is different from their CRM customer number.
• Attributes are more or less fixed customer characteristics like demographics (gender, age, profession, zip code) and firmographics (revenue, employee count, industry code, etc.). Attributes can be collected, for example once a day via batch file transfer.
• Everything else is events. This includes anything from website visits, ad clicks, and social media likes to purchase transactions and survey responses. Events are collected as a continuous stream using close to real-time APIs such as REST.

4. Complete the data

Regardless of which data collection strategy you choose, you need to get the job done. This means that at minimum the data fields required to run your always-on marketing campaigns need to be in the marketing dataset and the fields need to have values in them. Data fields are meaningless if they are empty.

Machine learning can help here, but only if a small number of values are missing and if these values can be predicted by a machine learning model. If many data fields are missing, they need to be populated manually.

In B2C business it is almost impossible to populate data fields manually, but in B2B salespeople should know their customers and be able to help fill in the gaps to some extent. B2B salespeople might know the right value off the top of their heads or have the data somewhere in their computer (for example, email address or title). Just remember that this is not their top priority. Sometimes the only way to make this happen is to link CRM data completeness with salespeople’s incentives.

5. Stitch together identities

Identifying the same customer in different digital services is tricky because many services use different ways of defining the customer identity: anonymous web visitors can be identified by a cookie (if they allowed it), while mobile app users can be identified by the device ID or email (or even customer ID if they logged in).

With social media things get murkier: in LinkedIn people are mostly themselves but their user ID might not match their real name and there might be many people with the same name. Twitter is even more difficult.

Facebook isn’t usually a good source of customer data because it doesn’t allow the extraction of any data beyond what is on the company's own Facebook page. But, depending on your industry, you shouldn’t discount Facebook completely: you might find your biggest fans (and biggest detractors) there!

All these different pieces need to be stitched together into one uniform customer ID.

6. Enrich the data

Enriching your data is one of the best and most reliable ways to improve its quality.

Firmographics data, such as company size, can be bought from a third party using a company’s tax ID or similar public identifier. Company names alone often aren’t enough because there can be many companies with similar names and a company can have several legal entities.

Zipcode-based demographic data providers are a new business in Europe. This is a tricky attribute to rely on because there can be all kinds of housing and thus income levels within one zip code. Nonetheless, it makes for an interesting addition.

Another, often neglected, method is to use visual analytics to increase your understanding of your customer base. This requires knowledge of data shaping and visualization methods, so your CDT should include people with the relevant competencies.

If your data includes free-form text fields (for example, Net Promoter System text comments), it might make sense to categorize these comments by topic and sentiment. Doing so turns free-form text into statistical information that you can use to create targeted messages (e.g. discount offers to price-sensitive customers).

7. Share your insights

Once you get the quality of your data to an acceptable level, other teams will be more than happy to utilize the same data. You should make the customer dataset visible to other teams via BI systems and visualization solutions such as Power BI, Tableau, or Qlik.

Which skills does my CDT need to have?

In order to bring real value to the organization, your CDT should have at least two members.

1. Customer data expert

The customer data expert’s mission is to cut through the corporate obstacles and make the other team member’s work as efficient as possible. Their mandate must be powerful enough to get any and every piece of data that is needed to execute the data-driven marketing programs.

A customer data expert’s job is also to educate, brainstorm, and facilitate customer journey design and implementation. They should have a good understanding of data modelling and how to run predictive analytics projects.

The customer data expert is also responsible for any compliance and data security issues (GDPR and privacy, industry-specific regulations, etc.).

2. Customer data scientist

A customer data scientist’s job is a mix of data shaping and advanced analytics skills.

Usually, the APIs are set up by the IT department, but a data scientist might be needed to get this job done as well, and that’s why they also need to have ETL-skills (Extract, Transform, Load).

Customer data scientist should be able to run a machine learning project using third-party APIs (Tensorflow, Sagemaker, etc.) or a tool like Rapidminer. For that they need to know Python.

These experts are hard to find because they prefer to work with like-minded people, and the other member of the team may not fit the bill in this regard. Data scientists are also, at the moment and for the foreseeable future, in high demand and this type of marketing-related work might not interest them.

One of the key things when setting up a CDT is to decide on the set of tools. This is important because the tools that you choose define the required skillset for your team. In my view the most suitable tool to accomplish the core mission of the CDT is a customer data platform (CDP). More about that in my next blog post.

If you don’t fully understand your customers, you can’t predict their behavior and target the right message at the right time via the right channel. While a dedicated CDT is a substantial investment, choosing not to invest in a CDT might turn out to be much more costly to your business in the long run.