Personalisation: Man vs. Machine

Video

Efe Yuceman, Consultant at emarsys, discusses whether machine, or human-controlled personalisation is more effective for retailers.

This clip is from our webinar, "Win Back Revenue: Recover Revenue with Successful Abandoned Basket Campaigns", in association with emarsys. You can watch the full on-demand recording of the webinar here.

[Morag:] Talking about the personalisation of content and how deeply you can go into these things - we're seeing it across marketing, personalisation's always been a hot topic. We've got all the data - we can identify this customer down to an individual level as we've got their email - but to what degree do we really need to develop that personalised content? From a creative point of view, that is a budget muncher, surely?

[Efe:] Absolutely. One-to-one targeting is the dream, but does that mean if I've got 100,000 contacts in my database, do I need 100,000 different messages, banners, calls to action? It's just not feasible. So here, technology comes to our aid. The way that we think about it is that there are two approaches - the first is looking at one-to-many targeting and personalisation, if I can group my customers into value brackets, i.e. my top spenders, mid-spenders, low spenders and my life cycle brackets, first-time shoppers, active shoppers and those that are defecting, then I can target them and set up content for those groups, and that's a relatively discrete task, so we're talking 5 or 6 banners here.

On the other hand, what I want to do is to incorporate product-based personalisation into the email as well, so that when I am building an email, I don't have to think: out of my 10,000/20,000 SKUs, which one should I show my customer? Let the machine decide that, and that's the kind of mixture between man vs. machine and how to use both. In some respects, the machine is very smart. It can say that a purple jacket is the next product that this customer will buy based on the statistics and what we know about our database, but what it won't be able to do is match that against spending levels, what sort of demographics they have, and so on. It misses out some of the more human factors, but in terms of being able to put together a personalised message quickly, it's going to help me.

Then, I need to employ some strategic thinking around: "well, I've identified that out of my 200,000 contacts, 50,000 of them are high value - how do I want to treat these guys?" That's still a human behaviour, but I don't think you get into a hyper-personalised problem there, you just need to focus on value segmentation and life-cycle personalisation.