Optimisation vs Personalisation

At Sky I’ve been lucky enough to have been part of a fantastic optimisation team AND played a major role in delivering our much-vaunted real-time digital personalisation program.

Stakeholders have asked what the difference between optimisation and personalisation is. If you’re not familiar with the detail, they can seem pretty similar.

And, having worked in teams that are laser-focussed on delivering one or the other, I’ve seen the understandable tension that arises when – on the surface at least – it seems like there’s a decision to be made around which one to use.

What’s the difference then?

At the most basic level: optimisation focusses on content; personalisation focusses on the individual.

Helpful, thanks… so should I use optimisation or personalisation?

You should use both.

Super helpful, I’ll get right on it…

I should probably go into a bit more detail. Here’re a couple of examples to illustrate.

Daft choice 1 - let’s sink all our effort into personalisation

A customer arrives on a clothing website and – based on everything we know about them – we infer that they want to buy some shoes.

So we personalise the website by bringing a banner for shoes to the landing page (no matter which page of the site that might be).

Awesome work everyone, job done.

Well no. Unfortunately, the checkout journey for shoes is awful. It hasn’t been looked at in years, throws up all kinds of blockers. Eventually, the customer gives up and we lose the sale. Ace.

Daft choice 2 - let’s throw all our effort into optimisation

A customer arrives on the site, and they really want to buy some new shoes. We’ve spent months fine-tuning our shoe checkout journey through optimisation – in fact, if they log in, it’s down to a single click.

Nice one guys, we’ve smashed it.

Alas not. Despite the fact this person has been an existing customer for years, and we have all the data we need to recognise them when they land on the site, we (rudely) ignore all of that and provide them a generic landing page instead… trying to sell them a fancy hat or something.

Unfortunately, they hate hats. They’ve had a bad, life-changing, hat experience, and they’ve never been the same since. Moreover, we know they hate hats. They’ve never bought one from us, and they regularly leave bad reviews about how awful hats are on an existential level.

Disgusted, they leave, never to return. Not only do we lose the sale, but we lose a loyal customer. Good job everyone.

To summarise

Optimisation and personalisation complement each other perfectly. In fact, the lack of one can undermine the effectiveness of the other.

If you don’t get the right users into the right journeys (personalisation), then you won’t get as much volume into said journeys, and it'll take longer for your optimisation efforts to reach significance.

If you personalise but serve bad content to your customers, it’ll make you think your personalisation efforts don’t work. If you’re A/B testing one model against another, bad content will make perfectly good models look bad, and invalidate the results. If you’re using machine learning to decide what model to use, the same thing will happen, just automatically and much faster.

So try to do both if you can.