Are CMOs Wrong About Data Science Skills Gaps?

Chart Source: The CMO Club, Feb. 2020

In this February 2020 survey, The CMO Club asked it members, “What are the most critical skill gaps in your organization? Select up to 3.” 78% chose Data Science, the top answer. The reason is clear. Data science knowledge is seen as indispensable to predictive marketing—the key to improving customer journeys, increasing satisfaction, and building sustainable, profitable revenue streams. For example, optimized cross selling, content personalization, message timing, and ad effectiveness measurement are perceived at the top as the province of data scientists.

For some sophisticated predictive problems this may be true, but for the majority of ordinary marketing predictions it is old news. The advent of practical AI and machine learning has made predictive tasks feasible for most organizations without data scientists.

So why has adoption of AI and machine learning for marketing been slow, especially in the mid- to small-size companies? Several reasons are often cited*:

  1. 57% of analysts say allocating budget was a challenge
  2. 48% of analysts say getting upper management on board was a challenge
  3. 43% of analysts say training and recruiting staff with the right skills was a challenge
  4. 39% of analysts say knowing where to start was a challenge

*Source “AI for Retail Marketers USPS Research,” USPS & SIS International, June 2019.

Why do these roadblocks still exist when there are a number of cost efficient, versatile and ease to use solutions available to immediately breakdown these impediments? Let’s address each separately:

57% of respondents said allocating budget was a challenge

Budget allocation is a perennial challenge, and always a poor excuse for inaction. After all, if a new technique delivers better results with demonstrable ROI, who would object? Reasonable business people don’t.  The real budget objections are perceptions that:

  • High cost consultancies or big brand/high cost software will be required, putting return on investment out of the question.
  • Data science teams will need to be built, with scarce talent and high salaries.
  • Expensive projects have failed to deliver to date.

All three of those points had been obstacles, but aren’t any longer. Low-cost tools are available that not only offer the flexibility to answer a large range of marketing questions, but also can be implemented—without data scientists or programmers—in hours vs. weeks. The versatility of these tools allow organizations to equip multiple analysts to make action-oriented predictions easily.

It is time to revisit budgets. Efficiency can be gained nearly instantly within the resources of even small companies.

48% of respondents said getting upper management on board was a challenge

This one is easy to understand. Management teams have been promised results that failed to materialized after large projects and massive expenditures. Why will it be different this time?

Fortunately, this obstacle is simple to overcome. Automated machine learning makes attacking a proof-of-concept straightforward and fast. Nothing speaks truth like results. Read How to Prove AI Value to Skeptics in our AI Tips archive for more.

43% of respondents said training and recruiting staff with the right skills was a challenge

This one is also based on old information. Solutions built to provide predictive marketing power to analysts are specifically designed with them in mind. That means there is no recruiting. Existing teams can now act on forward-looking data without writing a line of code. It is as simple as using a spreadsheet. Training expenses are minimal for applications that are self-service for anyone with basic analytical skills.

39% of respondents said knowing where to start was a challenge

Starting points can be difficult to find when processes are complicated. The first step is always “learn the process.” For approaches that require data science knowledge and programming skill, there is no simple path for analysts who do not have legions of specialists available. Once the process becomes as straightforward as uploading a few data sets and inspecting results minutes later, experimentation becomes an automatic starting point.

Knowledge gained in the first few runs reveals whether data is sufficiently broad and deep, if the predictive question is appropriate, and how the results can be translated to actions. Incremental tries with tweaked data and assumptions are fast, which helps analysts converge on a solution quickly. Low overhead means refinements are low effort and cost.

The take-away is that CMO’s do not have to fear lack of data science and analytical talent. AI and machine learning tools are available now to begin answering critically valuable questions with existing teams. Misconceptions may seem stubbornly persistent, but C-level execs did not reach their positions because they are resistant to change. They just need good information on alternatives. Give it to them.

Squark is available free to produce your own evidence.

What is Predictive Marketing?

Do You Even Need Predictive Marketing?

It comes down to insights versus actions.

Predictive Marketing is the umbrella term encompassing processes that rely on AI and machine learning to drive actions with accurate predictions. This is in contrast to traditional marketing, which has been built on descriptive and diagnostic processes.

Traditional Analysis
Business questions are formulated and translated into requirements that include data collection, tooling, and governance. KPI’s are created and these leading and lagging indicators are put into reports and visualizations for statistical analysis and “data storytelling.” The results are insights—impressions of “what’s going on.”

There are two main limitations of this approach. First, statistical methods rely on guesses of what factors to consider. It is difficult or impossible to analyze every variable with respect to every other. Second, reports and visuals that demonstrate trends do not identify individual records that exhibit them. More steps need to take place to actually apply the insights.

Predictive Analysis
Business questions are answered by the data itself. Machine learning is used to understand relationships within sets of known outcomes so that the patterns can be applied to never-before-seen data. Typical questions are “yes-no?” (binary), “which of three or more?” (multivariate), or “how much?” (forecast regression). No presumption of the causal factors is required. Simply by identifying which features (columns) in the known outcome data represent the answers, machine learning can fill the blanks for the new data. The result of AI-driven predictive analytics are data sets with record-by-record scores of which individuals are likely to exhibit the behaviors.

With predictive AI, knowledge of statistics, data science, and programming is unnecessary. Most importantly, marketers can instantly switch from insights to actions. For use cases such as lead scoring, conversion optimization, attribution, content targeting, cross-sell uplift, lifetime value forecasting, and churn reduction, actions always beat insights. Think of the value of simply examining data within your existing martech stack to produce ranked to-do lists for your most important marketing programs.

Sound like magic?
Predictive marketing is built upon sound AI science, with proven reliability and accuracy. In fact, the explanations of why machine learning algorithms made their predictions are tremendously valuable. By listing which variables are most predictive, AI reveals the most important elements for focus in marketing outreach—even ones you never suspected might be operative.

A famous movie director once opined that if film began as color, no one would have thought of inventing black-and-white. Try predictive marketing with Squark and see if your vision of the future is suddenly more colorful.