Ecommerce Personalisation: Beyond 'Hello [First Name]'

Posted by Sean on January 22, 2019

Every January my inbox fills up with vendor emails promising “one-to-one personalisation at scale”. Most of them open with my first name, which I suppose is meant to prove the point. Putting someone’s name in a subject line is mail merge, and we’ve had that since the nineties.

I’ve been thinking about this properly because personalisation keeps coming up as a line in ecommerce plans, mine included, and the word covers so many different things that it’s close to meaningless. So here’s an attempt to break it down into what I think is worth doing for a smaller luxury retailer, and what I’d leave to the businesses with a data science team and millions of sessions a month.

Start with the data you actually have

The honest starting point is that a small brand doesn’t know very much about most of its visitors. The majority arrive once, look at a few pages and leave. You know where they came from, what they looked at, roughly where they are in the world and what device they’re on. That’s about it.

The customers you know well are the ones who have bought, and in a business like tailoring or leather goods they’re a small group who buy again and again. That changes where the effort should go. I’d put the basics for people who have already bought from you well ahead of a clever on-site engine that reshuffles the homepage for anonymous visitors. The engine is a lot of work, and for most visitors it has very little to go on.

The things I think are worth doing

Location, done properly. This is the least glamorous and probably the most useful. Show the right currency, the right delivery cost and the right delivery time for where the visitor is. Don’t make an American customer do the sums on a sterling price and then find out about duties at the door. Nobody would call it personalisation in a pitch deck, but it removes a real reason not to buy.

Size and fit memory. If someone has bought a shirt in a particular collar size and fit, the site should know that next time. This is where luxury clothing has an advantage over mass fashion; the repeat customer often buys the same size for years. Pre-selecting it is a small courtesy and saves a lot of friction.

Recently viewed, and back in stock. People browse expensive things over several visits. Making it easy to find the item they were looking at last week, and telling them when the cloth or colour they wanted comes back, is simple and genuinely helpful.

Behavioural email triggers, with restraint. A browse or basket reminder is useful once. Three of them, escalating to a discount code, teaches customers to abandon baskets on purpose. I’d rather send one well written reminder with some extra information about the product than a sequence of nags.

Recommendations that make sense. “Customers also bought” works when the catalogue is big enough to produce sensible pairings. With a few hundred products it often recommends something odd. Hand-picked pairings, a tie that goes with a shirt or a wallet that matches a bag, will usually beat the algorithm at this size. It’s tedious to set up, but it’s a one-off job per product.

The things I’d hold off on

Dynamic homepages that change per visitor. Personalised search results. “AI-powered” anything that needs a long integration and a monthly fee bigger than the uplift it’s likely to deliver. Some of these tools are clever, and I can see the case for a big retailer with the traffic to train them. For a small brand they mostly create a site that’s harder to test, harder to explain and harder to cache, which matters for speed.

There’s a privacy angle too. Since GDPR, I’m much more careful about anything that depends on tracking people closely across sessions before they’ve given you any reason to think they’ve agreed to it. I wrote about the first week of GDPR last year, and the conversations since haven’t made me more relaxed about it.

The shop still does it better

The best personalisation I’ve seen in this sector doesn’t involve software at all. It’s a shop assistant who remembers a regular, emails when something they’d like comes in, and knows not to bother them with things they won’t. Online, the closest equivalent is a CRM that the shop team actually uses, with notes that someone reads before sending anything.

So if I were setting a small brand’s priorities for this year, the list would be fairly modest. Get location and currency right, remember sizes, tidy up the triggered emails, and give the shop staff better tools for what they already do well. I’d look at anything more ambitious after that, probably not before next year.

← Previous Draft Draft Index Next Draft →