How To Communicate Your Marketing Questions for Data Scientists
When you make accurate predictions, you make smarter business decisions. This is the core value data science offers to marketers. It’s easy to get caught up in the buzzwords, but AI marketing automation and machine learning tools will only help you if a) you know exactly what you’re looking for, and b) you can communicate that clearly to a data scientist.
In this article, we’ll take a look at how data scientists think; how data science-driven predictions inform better decisions about content marketing strategies and initiatives; and how to ensure marketing and data science teams are speaking the same language.
A (very) brief introduction to data science techniques for marketing
Data scientists can combine content analytics, text mining and machine learning algorithms to draw out patterns and insights from customer behavior, social media activity, and other datasets.
Here are some specific examples:
- Collaborative filtering models help marketers to target ads and offers to the right customer segments at exactly the right time.
- Algorithmic multi-touch content-attribution models allow you to use data from every online customer touchpoint (watching your Instagram video, reading your eBook, browsing your blog, commenting on Facebook, etc) to make sense of exactly how content influences audiences and purchase decisions.
- Sentiment analysis allows you to figure out how customers feel about your content, social media output, brand announcements, or overall campaigns (and your competitors’ campaigns), collating and interrogating these datasets to establish broad patterns rather than analyzing each individual post.
The marketing-data science language barrier
One common cause of friction is that marketers often think of their jobs as exclusively creative endeavors while data scientists are more focused on hard numbers and quantifiable statistics. This alone can make it difficult to work together, as the gap between needs and results can feel too wide to cross. Instead, to make this work, you need to recognize that you are bringing different skill sets to the same questions, which require both creative thinking and analytical efforts to interpret meaningfully.
Marketers need to recognize that their intuitions need to be backed up by results, and that these insights come from data. At the same time, data scientists who work in marketing need to appreciate that the sales, conversions, and data they are working with are based on complex social, cultural, and emotional triggers. It’s not enough to stay in your lane: you need to engage with each other’s perspectives to get the complete picture.
Data scientists vs data analysis
Another fundamental problem is that marketers tend to use the terms “data scientist” and “data analyst” interchangeably. This can create some confusion about what marketers actually want from their data.
Data analysis involves reviewing your historical and current data to get a clearer idea of what’s working and whether what you’re doing is worth continuing. As a rule of thumb, if you can get these answers using an Excel spreadsheet and relatively simple graphs, charts, data visualization tools, and analytics dashboards, it’s data analysis. You don’t need a data scientist for that.
Data science, on the other hand, involves designing experiments with data in order to test possible outcomes and make predictions. These experiments require deep knowledge of statistics, calculus, and linear algebra, using programming languages like R or Python, and a sound understanding of machine learning techniques.
Of course, marketers don’t need to know exactly what’s happening under the hood. That’s the data scientist’s job. They do, however, need to appreciate that data scientists uncover and decode complex customer behaviors that are deeply embedded in datasets, in order to understand how and why audiences engage with digital content. This is not a simple data analytics problem.
It takes advanced statistical algorithms and machine learning tools to do this. It may also take significant time, resources, and careful planning. Recognizing the scale of what you are asking your data science team to do and being willing to sit down with them and explain exactly what you need is the first step towards a fruitful data science-marketing relationship.
Think in terms of models
Data scientists build models. In data science terminology, these are often input/out (I/O) models. For that reason, it’s very helpful to clarify what your inputs and outputs are, in order to help them design these models in the right way.
Inputs are the influencing factors that get into the model. In marketing, that means things like ad designs and copy, as well as audience data such as demographics, social media activity, and other data points you may collect from your marketing activities and channels.
Outputs are the things that determine the results of the model. In marketing campaigns, that might mean sales, new customer sign-ups, revenue, clicks, or conversions.
Defining machine learning tools for marketing
In fact, if you can get used to reframing more general questions as requests for specific models, you will immediately tap into a world your data science team understands, getting you off to a productive start.
Take some of these common marketing KPIs:
- ROMI (return on marketing investment). This tells you how much revenue was derived directly from marketing activities.
- CLTV (customer lifetime value) / LTV ( Lifetime Value). This indicates how much a client or customer is worth to your company, in revenue terms, over the entire course of their lifetime.
- Churn (or “rate of attrition”). This tells you the rate at which people stop doing business with you, or cancel their subscriptions within a set timeframe.
These are all expressed as mathematical calculations that can translate directly into data science techniques. As such, describing what you need in these terms will help you demonstrate precisely what it is you’re trying to achieve.
If you tell a data scientist you need a churn model, a ROMI model or an LTV model to help you predict the success of a proposed campaign, they will be on much firmer footing than if they hear “hey, can you help me figure out which demographic we should target this time?” or “can you predict where I should increase my marketing spend?”
Final thoughts: sourcing the right data
Another key issue to discuss with your data science team – and one that often gets overlooked – is “where else can we get the information we’re missing?”
As a marketer, you already know how important it is to look outside your own business for ideas and perspectives. You know you need to track what others are doing in the sector, how your competitors perform, how people talk about your brand on platforms you don’t own, and so on. Data scientists are skilled at sourcing data from a myriad of sources and systems, so have these conversations with them. Explain what kinds of datasets would enrich your understanding. See if they can source this external data for you. The more you learn to speak the same language, the more you’ll be able to work towards the same goals, together.