Legend of the Phoenix: Personalization

Personalization is a weird thing. Like the legend of the Phoenix, the bird that is cyclically reborn, personalization is one of these marketing challenges that as soon as you think you have addressed it, it almost immediately comes back. You think you get ahead of the market, only to find out that you’ve fallen more behind.

The reason for that is that most solutions think of personalization like the industry did in the 80s.

The traditional model for personalization is the so-called “market basket analysis”, where algorithms analyze what customers buy together (in the same “basket” – which can be anything from a web checkout to a real basket) and then analyze that, for example, “people who buy diapers also buy beer.”

In the last 10 years a number of vendors came out promising an easier way to do market basket personalization. For instance, they would analyze your web data and analyze what products are bought or browsed together. Then they could generate web or email recommendations based on what you’ve been purchasing. Even though such approaches sound like the Holy Grail, they’re far from it – and results often disappoint.

There are 3 main problems with this traditional personalization approach:

  1. It is channel-specific
  2. It can’t use cross-channel data
  3. It relies on explicit customer signals

The first and the second point are almost obvious in retrospect. In today’s world of converging channels, where customer-centricity and a single marketing voice are strategic priorities for forward-thinking companies, terms like “web personalization” or “email personalization” are outdated. A true personalization solution that can scale needs to leverage data from “any” channel and help advance the personalization of marketing & products across any channel. However, very few products can truly leverage all data – connect to Data Warehouses and Hadoop systems, ingest billions of events without expensive and (often) delayed IT projects.

The third point is more fundamental – the “people that bought this also bought that” type of recommendation – popularized by Amazon in the early 2000s, is already outdated. You must have experienced this yourself! Every customer is different so just because some random customer bought two products together doesn’t mean that  *you* also want to purchase this product or these two products. With “market basket” algorithms being part of the past, the future lies in behavioral micro-segmentation – being able to create segments that contain only the people that truly care about a specific product. What would make a certain customer beyond in such a segment? Anything! It could be what they browsed. It could be a question they asked customer service three months ago. Perhaps they bought a similar product offline last year and it’s time for an update. Or maybe, just maybe, the one or two emails that they truly engaged with had to do with products in this category.

We can even take this a step further. Why should recommendations produce the products that customers are most likely to buy? That’s like optimizing for the next purchase. But what would recommendations look like if we were optimizing for lifetime value (LTV) instead? Maybe instead of showing the customer something that they already know they want to buy, we want to show them a product that they like but don’t know yet. In other words, we may want to point customers towards  “gateway” products, that if they buy we know they will rise to a whole new level of engagement. Instead of achieving just a few product sales, we gain customers that buy a much larger range of products, are much more engaged and purchase a lot more frequently.

The three key points to remember are:

  1. Most personalization vendors do single-channel, market-basket personalization
  2. To get to the next level of personalization, you have to first connect your customer data and then make sure to leverage all of that with channel-agnostic personalization capabilities
  3. Optimizing for purchases but also customer LTV generates a personalization mix that boost your numbers and is much more strategic for your business 

Getting personalization right can be a game changer for your business. ActionIQ focuses on the future, not the past. Let us know if you would like to discuss together what the future of personalization looks like.

 

Tasso Argyros

Tasso Argyros

Tasso Argyros is the founder and CEO at ActionIQ. He started ActionIQ to combine his passions for solving real-world business data problems and developing innovative technology. Originally from Greece, he dropped out of the PhD program at Stanford to start one of the first companies in the Big Data infrastructure space, Aster Data.
Tasso Argyros

Latest posts by Tasso Argyros (see all)