As digital marketers, we work constantly with numbers. There are hundreds of metrics from multiple sources with their own biases and nuances for interpretation.

While marketing analysts can be great at collecting data, bringing it together and making it visible, a data scientist with strong knowledge of digital marketing can add that invaluable layer of context to everything.

As marketers, probably the most common role we see the data scientist in is “Marketing Mix Modelling” or “attribution”, with the focus being on understanding the relative value of marketing channels and campaigns. But the marketing scientist can do so much more for you than just suggest a better spend mix between marketing channels. In fact, the process of modelling which channel is better itself illuminates vital knowledge for success growing a business.

Adding context such as pricing, promotional strategies, product stock holdings, seasonality and customer lifetime value leads not only to a better understanding of what is working in digital campaigns, it helps with understanding how much stock is needed to fulfil demand in peak season. The marketing scientists’ work informs product and pricing strategy and encourages teams to work together better.

Data scientists can add so much value to any business’s marketing, in so many ways. But we would like to focus on a key selection of valuable topic areas that are especially important to us as growth marketers:

What really drove a change in performance?
What is the incremental value of a campaign?
What should our overall target be?
How much can we spend and hit our target?

What really drove a change in performance?

It’s the weekly trade meeting. Marketing reports a strong weekend with top line revenue up 20% week on week, and markedly strong results reported in the marketing platforms’ data. Conversion rate is up on the site. Someone chimes in a thought: “we had a sunny weekend and it was payday recently”. Marketing says that over the course of this Month, they have been keeping budgets stable despite the expected drop off in sales. The plan has been to keep building awareness knowing that pay day would come, so they had hoped this would drive a bigger result at the end of the Month.

A data scientist versed in marketing mix modelling, however, will be able to say: we had a great weekend not just because it was sunny and we had payday – we expected a 10% week on week uplift only, so we did a lot better than expected given these conditions. Previously we analysed the effect of stock holdings of key products on performance, and in this case stock holding has been comparable, so we can rule that out as a reason as well.

What is the incremental value of a campaign?

Let’s set the scene: you invest in a new digital marketing channel. Your daily budget in other channels has remained the same, whilst all additional spend has gone into a pilot campaign for this new channel. The last click results on Google Analytics, and the in-platform results show great cost per order driven, better than the blended overall target for the business. 2 weeks pass, and the blended cost per customer acquired for the business has worsened. There has been no change in acquisition, despite the increase in spend. A deep dive into the channel shows the test campaigns have largely been served to individuals who have already given their information or visited the website. Any system using click-based attribution is highly unlikely to spot this issue and recommend to turn off / try again with this new channel.

The data scientist is there to help you understand how much time and budget you need to give to a significant change to be able to feel its effects on solid, top line business numbers. Given the variance in daily sales overall, how many days of data do we need to collect? Given our beliefs about how long it takes for new impressions on this channel to convert into customers, how does this affect how long our experiment needs to last? When evaluating how effective these changes were, the understanding of the context is needed in order to get valid answers as we cannot time all our efforts to not coincide with important changes, like running a sale.

What should our overall target be?

Say we have discovered that when we run a 20% discount, we can scale up spend 50%, and revenue also is expected to grow 50%. The AOV is lower, so our ACOS (or new customer ROAS) remains the same, but our CAC is going to look better. So which should we focus on more? If we do not expect customers to ever buy from us again, then ACOS makes the most sense. But if the first order is likely to only make up a small portion of the customers’ value over the next 6 Months, then we are better off focusing more on CAC.

So, if instead repeat purchases make up a significant proportion of our revenue, then we really need to understand what affects the lifetime value of the customers we acquire. After all, customers who are willing to pay less for your product now may well also be less likely to keep coming back for more.

The marketing scientist can forecast LTV for you, so you can set the right targets throughout the year, and estimate how much budget you will need to maximise acquisition for the given target!

In summary:

There are so many ways a data scientist specialising in marketing can add value to a company’s growth efforts. While some might feel too small to warrant doing a lot of work on media mix modelling to understand which channel performs best, everyone can benefit from a strong grasp of the effects of large budget decisions, sales, seasonality, product and website CRO on the scalability of the business. Your marketing scientist could be saving you valuable time by making you a tool in a spreadsheet or app to help you see how various tweaks would affect your rest of year forecast – what if LTV gets a little better? What if black friday isn’t as big as last year? Can we get an optimistic and a pessimistic view? In a fast-paced environment, a data scientist can be your ally in achieving insight close to the speed of thought.