tetrel Logo tetrel Logo

Diggin’ Deeper, Vol. 3: using buried potential

2021-06-17 by Johannes Humbert (LinkedIn | Twitter)

Why constantly stake new claims when an existing one is extremely profitable and far from exhausted? With intelligent algorithms you can optimise your cross- and upselling and get even more out of existing customer relationships.

Why go far afield when good things are so close by? In this case: when existing customer relationships have much more to offer? The potential of cross- and upselling is undisputed - new opportunities for this are opening up more and more through solutions using artificial intelligence. Whereas it used to be rather difficult, time-consuming and associated with a lot of trial and error, information can now be better linked and thus potential can be exploited. Intelligent and automated.

The overriding thought here: relevance

Which product at which place at which time at which price via which channel to which customers? These are probably the central and decisive questions in the development of successful cross-selling and upselling strategies. Here it is important to act precisely - in order to achieve a maximum increase in efficiency. Because even a small mistake in the strategy can have dire consequences: New and existing customers can be scared away, sales campaigns come to nothing, budget and resources are unnecessarily burned. Simply because the relevance is not balanced. Sounds too theoretical and superficial? Then here are a few examples:

Seriously?

A new customer, great. He has decided on a product or service because of a certain offer. And just a few days later he finds out that he could have had the same offer at a lower price through another channel. Not a nice first experience. Equally unpleasant and devastating in its effect: someone has decided on a credit card - and even before he can use it for the first time, he is already asked to upgrade to a more valuable card. As the saying goes:

There is no second chance for a first impression … Or that long-standing existing customers are the last to know that there is a new offer - via non-personalised advertising. Another bad example, let’s stay with credit cards: Over months, the use of a customer’s card has continuously decreased, it tends to go towards zero. Then this is really not the best moment to offer him a new card that might have a bit more performance but is twice as expensive, is it? That would be more for heavy users, as a reward with a bonus …

Cross-selling and upselling offers are messages

Every communication with customers should always be well thought out. Because every communication is like a conversation. Now imagine you are the proud owner of a brand new mid-range car. A proud owner. And at the first inspection after three months, you are told at the dealership that this is already a nice model, but that a top-of-the-range model is much better. Thanks for the talk and no hard feelings … wrong message, wrong time, wrong place, wrong offer.

Data-Mining was just the beginning

For many years now, data mining has been an elementary component of every CRM strategy, which also includes cross-selling and upselling strategies. At least it should be the norm. As the name implies, it involves digging for information in data. How long has one been a customer? What product history is available? Number of contacts etc.? From this, possible scenarios in the Customer Lifetime Circle are then derived and assumed.

So in principle it’s nothing new to say “Optimise cross-selling and upselling with AI solutions!” In principle, yes, but the dimensions have changed. What used to be very costly and time-consuming work can be presented much more quickly and accurately on a larger scale - intelligently, automatically and, compared to data mining, multidimensionally.

Detecting gold via algorithm

One AI tool for this are multi-armed bandits: This method uses exploitation and exploration. In short, exploitation refers to the procedure of analysing and making use of existing data and their patterns - in other words, pattern recognition. For example, product recommendations are made on the basis of customers' historical purchasing decisions. Exploration is when there are no patterns (yet). Then new strategies based on assumptions are tested and analysed and tested and analysed again and … Multi-armed bandits balance the exploration-exploration trade-off. Or in other words: they ensure an optimal weighting and use of patterns from the past and new strategies. In this way, the AI model is continuously optimised and develops improved strategies.

The result in a nutshell:

Relevant offers and channels increase the individual customer value through increased turnover and profit and the sales department receives concrete recommendations for action. In other words: the cross- and upselling potential is exploited in the best possible way. It is worth digging for gold. It is more effective to dig specifically, especially where it is worthwhile. Artificial intelligence is an extremely valuable prospector in this respect.