In the previous posts, we got introduced with Data Science and some of the most famous use cases from the industry leaders. In our future posts, we will talk about the applications in specific industries. Today on the menu – marketing.
Marketing is one of the most fertile fields for Data Science. There are many different techniques and directions which marketing teams are striving to realize, but the two most frequently discussed are – up sell and cross sell. For those of you who are less familiar with the terms, up sell stands for the action of convincing a customer to buy (or upgrade to) a more expensive product/service, and thus – spend more money, while cross sell stands for making a customer buy one additional diverse product/service besides the one already chosen and thus – spend more money. I will leave the further explanation here, for the reference.
But is there a missing link that could enhance this process of “convincing” a customer to make a purchase, and increase upselling and cross selling? If you ask me – that is Data Science. To be honest – we are light years away from the period when there was no big deal if we lose a customer. Nowadays – things are way different. The situation on the market is as follows – numerous and strong competitors, mass production, high customer acquisition costs, high churn rates and – poor loyal relationships. And the evergreen claim – it is less expensive to keep the current customer, than to acquire a new one.
The attractiveness of Data Science to the marketing teams comes from several factors, all sinking into one funnel: tailor-made marketing. Life would be much easier if you could know what the customer would like to buy, right? But then again, it’s not only about the preferences, but the ability to afford it (from the financial aspect). Besides, maybe he cannot afford it currently, which does not mean he could not afford it by the end of the month, when he gets his salary. On the other hand, maybe he’s waiting for the special discounts and actions… How would you know? The list goes on and on. What to do with all the pressure? Well, look into the data, there lies the answer.
The main benefits of applying advanced analytics and machine learning in marketing are:
- personalization – understand customers preferences and needs, in order to determine the content, frequency and channel to target them and nurture their customer journey
- prioritization – identify high-value customers and improve decision making process including time, resource and budget allocation
- performance – monitor and evaluate transitions and progress over the time, in order to quantify how promoters and detractors are growing or shrinking through the time
- predictiveness – the possibility to project and influence the future customer behaviour, based on their historical behavioural patterns, preferences and habits
It is clear that there is no all-mighty algorithm, or tool that can give you the answer to all these questions, in such a way that you could use that information to solve all the problems. Here are few ideas on how marketing could leverage the underlying campaigning processes.
Recommender systems are maybe the most powerful, yet certainly the most attractive, asset that can be embedded into campaigning tools. The main purpose of using recommenders is to analyze customers preferences and similarities, in order to evaluate which products are highly likeable to be bought by the customer in the future. This makes recommenders a good call for cross-selling campaigns. You can extract the rating from recommender engines describing the affinity level of the customer towards each product, and use that information to create the “next best offer”.
Segmentation (or in machine learning terms – clustering) is used for grouping customers with similar characteristics and purchasing habits into small segments. There are plenty of segmentation types that can be developed – activity segmentation, behavioural segmentation, brand awareness segmentation, etc. The type of the segmentation depends on the information we would like to obtain and the way in which we would like to use it. Dividing customers into small segments makes the process of prioritization and focused targeting a lot easier (for example, one would define a different campaign for the segment of “overachievers” and the one with “dormant” customers).
Propensity to purchase
Analyzing the probability that the customer will make a purchase in the future can further direct the actions taken towards. In a nutshell, based on the historical data of activity level, spending habits, periodicity, preferences and final outcome (made purchase, didn’t make purchase) – an algorithm can be trained in order to retrieved the probability or a flag indicating the propensity to purchase in the future, given current state describing customers features.
The dark horse of tailor-made marketing certainly is – the survival analysis. The main goal of this analysis is to estimate the time to a given event (e.g. purchase), and to quantitatively explain how this time depends on various properties of the treatment (campaign), customers and other variables. Sometimes, having estimated the propensity to make a purchase isn’t enough. If we want to create and send a promotion – we need to know when is the right time to do that. Why send the promotion, if a customer will come and make a purchase either way? On the other hand – if you wait for too long, the customer will give up. This time estimation can help target customers in a timely manner.
Customer lifetime value
Estimating customer’s lifetime (monetary) value is performed by taking into account all previously mentioned features – spending habits, periodicity, activity level, spending trend, demographic features, etc. The information of the expected revenue from the given customer in some future period can help in identifying high-value customers and the potential that can be (out) reached if the customer is well nurtured and targeted. CLV can be a valuable insight for maintaining the cash flow and strategic planning regarding marketing and sales campaigns.
Many, many more…
Integrating all these use cases and joining their outputs could help in wrapping up the answers to these questions:
- what to target the customer with?
- how important is this customer for our business?
- is this customer going to make a purchase in the future time period?
- when is it expected for this customer to make a purchase?
- how much money is the customer likely to spend in some future period of time?
- how sensitive is this customer to the specific campaign/discount value?
Having answers to all these questions means being able to define and direct future marketing activities in order to differentiate from others and nurture loyal relationships with customers by placing them products, content and promotions they could really be interested in. The main prerequisite is to find the answers though clearly defined use cases, and well engineered and developed machine learning algorithms.
There are plenty of other use cases that can be defined and developed in order to leverage marketing processes focused on upselling and cross selling. Hopefully this post will wake up some new ideas in which data science could be applied in marketing.