Improving the Marketing-Data Science Relationship

Machine Learning|8 Minute Read

Improving the Marketing-Data Science Relationship

The traditional marketing-data science dynamic is changing — and for the better.

The days of marketers relying on (or longing for) their resident data scientist to collect and organize customer and prospect data from several different sources for analysis and activation are over.

(Or, at the very least, they can be for marketers who use a customer data platform.)

Even with a CDP and concerted data strategy in place, though, there are still several instances in which marketing needs a helping hand from their data science department, including:

  • Scrubbing dated and inaccurate data from various (potentially dozens of) marketing data sources
  • Assisting with the construction and deployment of some (but not all) machine learning models
  • Interpreting complex data analytics, diagnosing issues based on the info, and prescribing remedies

It’s evident data scientists continue to play a central role in marketing teams’ success today.

What’s not evident to some of these individuals on both sides of the data-driven marketing equation, though, is how they can better align on a daily basis to improve their shared business goals.

Common issues in today’s marketing-data science dynamic

“Operational efficiency” typically refers to an organization streamlining its business operations — product development, supply chain, customer support, etc. — to ensure top-to-bottom success.

When it comes to data science and marketing, operational efficiency refers to the unification of the teams: the former supplying the latter with the necessary resources and insights to succeed.

“For me, I’m doing my job right if I can give my company and stakeholders the data they need to make decisions,” an anonymous data scientist explained to Digiday.

Success comes in many forms. In this instance, it’s defined by marketing’s constant access to pertinent, up-to-date, first-party data they can activate with ease through real-time marketing, like:

  • Real-time bidding for programmatic advertising
  • Personalized emails to top customer segments
  • Individualized emails to prospective customers
  • Dynamically populated site and app messages
  • Triggered messaging based on geolocation data

Where things tend to go wrong with this marketing-data science relationship — as is often the case with other departments in a given company — is communication (or lack thereof, technically).

Data scientists need to hear directly (and often) from marketing to fully understand:

  • What types of data they need (e.g., particular segments they may need help creating)
  • How those data sets will be used (e.g., if custom, predictive models will need to be built)
  • Where the data needs to be stored (e.g., in separate sources or a centralized database)

In other words, the onus is (mostly) on marketers to document specific data needs, then liaise with data science so they can determine how to best aid them: hands-on help or general guidance.

marketing science

Knowing data needs — and how to express them — critical

And yet, many marketers fail to communicate effectively — often because they either don’t know exactly what they want data scientists to help them with or how to articulate their ask.

If you — a current or aspiring data-driven marketer — don’t know what data you need to make marketing science-backed decisions, your friendly, neighborhood data science professional can’t help.

Here’s a common data science project marketers request of their in-house analysts today — one you’ll likely run into (if you haven’t already) given your increased focus on data-driven activities:

  • You want to take a deeper dive into your customer data to discover what’s causing potential customers to fail to purchase after a lengthy customer journey and lots of digital engagement with your website — particularly, with some of your product pages.
  • You know there are numerous prospective buyers who visit your site often, stick around for long times each session and check out many of your product offerings. But, for some reason, they just never purchase — a common pain point for ecommerce brands today.
  • So, you turn to your data scientist (or data architect, data engineer, or data administrator — whatever title your colleague in question has been assigned) to have them comb through their advanced analytics regarding your users’ site interactions.
  • And then, you wait. And wait. And wait some more. Why? Because you didn’t properly specify the exact data you needed. You simply asked for insights on all segments’ behaviors — a time-consuming task for data scientists, particularly if they don’t have a CDP.
  • If you requested insights related to specific segments (i.e., those composed of individuals who viewed at least X product pages in the last Y days, who’ve never bought before, and whose momentum scores continually increase), you would have received far more useful info.

This is a common state of affairs for B2C and B2B marketers alike: the constant desire to detect what marketing elements are driving certain customer behaviors like this and modify them accordingly.

As long as you clearly lay out your specific ask and needs to your data science expert, you can more easily and efficiently resolve issues like this sooner rather than later — not to mention prevent lost revenue due to problematic aspects of your UI/UX, such as poor-converting product pages.

marketing data science

Proactively sharing insights crucial task for data scientists

Marketers certainly need to initiate and own most conversations with data scientists, given they’re sitting at the proverbial wheel, driving their data-driven marketing programs day-to-day.

But there are certainly situations in which these data scientists need to meet marketers in the middle.

Situation #1: Your organization is stuck in dreaded data silos.

We’d like to think most marketers finally recognize the distinct value of a customer data platform today. However, we know many have yet to invest in a proper CDP and, instead, have siloed data.

This means they must instead rely mostly, if not wholly, on data science to consolidate data for them — an arduous tasks for members of this team, given it accounts for 80% of their jobs.

Sometimes, though, marketers don’t know they don’t have access to the latest customer data. They end up asking for data they assume they can access, unwittingly limiting their results.

Thus, data scientists need to regularly check in with marketing to provide fresh customer data.

And it’s not just hands-on marketers who need consistent data updates: Roughly 60% of marketing decision-makers struggle to access data (and integrate data into a single source of truth).

Many data science departments have been built for this sole purpose, and it’s easy to see why:

  • Integrating data from multiple sources (most brands tend to have several customer databases today) can help not only marketing, but also sales, customer service, and other teams.

Therefore, it makes perfect sense data scientists should proactively champion data unification — and find the most applicable, pure-play customer data platform for the job.

Situation #2: Your marketers need new, intricate data models.

Marketers who own and operate BlueConic’s CDP for their brands know they can take advantage of out-of-the-box machine learning (ML) models to better predict customers’ and prospects’ behavior.

For instance, using Jupyter notebooks in AI Workbench, our customers can deploy (or schedule) models, and train them over time, to anticipate things like customer lifetime value (CLV) for a particular segment (e.g., top buyers and the likelihood newsletter subscribers may churn.

There are certainly ML models, though, that are beyond the average marketer’s technological scope.

Enter data scientists, who can assist with these AI marketing efforts and import custom notebooks into software like ours to allow marketing to take the wheel with deployment and, in turn, set them up for long-term success with their big data.

machine learning

Situation #3: Your company simply has too much dirty data.

A Deloitte survey of consumers found the data brands collect about them was mostly inaccurate — particularly third-party data, some of which may have come from disreputable entities.

Meanwhile, a 2019 Experian study discover 95% of major businesses worldwide said poor data quality undermines their business performance and ultimately worsens their bottom lines.

While modern (see: smart) marketers now turn to first-party data to enhance their marketing efforts, there’s still the occasional need to clean up one’s database and eliminate old or imprecise data.

If you’re a data science professional, make it a focal point to conduct monthly or quarterly audits of your data sources to weed out contacts who are likely to never convert in any way, shape, or form.

Your marketing department will love you for it, as they’ll be able to focus only on the customers who matter and not waste any time, effort, or money on unworthy prospects.

New and improved processes key to better relationship

As veteran data scientist Katie Malone shared in her guest post for AdExchanger, there are several sources of tension between data science and marketing departments.

Arguably the biggest one? Simply not making the effort to understand one another’s “language.”

Malone’s solution is a sound one: Find someone to act as “a bridge across teams, representing both points of view in relevant discussions” to ensure both get what they need to succeed.

With new protocols like this in place to streamline communication from data scientists to marketers and vice versa, misunderstandings can be avoided and data can be properly leveraged.

At the end of the day, that’s what both departments desire (and, frankly, what your CMO requires).

ai for marketing

See what BlueConic can do for you.

Whether you’re looking for operational efficiencies or improved marketing effectiveness through data activation, our customer data platform can help.