loader image

Why you DON’T need AI for personalisation

Why you DON’T need AI for personalisation

“Personalisation at scale.” Has been something of mission statement for a lot of marketing departments in the past few years. And while this lofty ambition often boils down to getting the right message, to the right person and the right time…too many of you reading have probably not got further than personalising the {NAME} in your emails.

What’s The Challenge?

Delivering personalised experiences is reliant on having the right data and algorithms that typically get smarter over time. The problem is that while 80% of consumers want personalised experiences (Epsilon) it’s also true that 86% of people of consumers say data privacy is a growing concern and 30% aren’t willing to share their personal data for any reason(KPMG). And so brands typically don’t have the data to deliver personal experiences in a meaningful way. So how can this be remedied?

Asking The Right Questions

All too often personalisation projects are started from a data point of view. What data do you have, where can you fill those gaps, what third party systems can you connect to…but all this does is give you quantity. It doesn’t necessarily give you that quality data that you can use to start personalising effectively.

Think carefully about the key pieces of information that are really going to make a difference to the way you would talk to a customer. What are the key points in differentiation for your brand?

For example, I was talking to an ecommerce brand that sells football shirts recently. While they saw themselves as being fashion-led, but most of their customers would purchase products based around the team they supported or their nationality. Boom! By adding in two questions into their order process, they can personalise their communications with ease.

Finding those one or two questions that really make a difference to your brand can be difficult, but can pay dividends. You may need to think about some sort of value exchange earlier in your customer journey in order to capture these data points which could be free information, a prize draw or money off etc.


With some core questions answered, you can then begin to segment your customers along those lines. Segments need to be markedly different and have a clear focus and messaging associated with them. If you were a travel company, your messaging would be very different for a business trip vs a romantic getaway. Your segments should not overlap and should be closely aligned to your offering in order to be effective. This can be done at a fairly high level so long as the relevant tagging or taxonomies have been set up. Even basic automation can ensure that if Emily clicks on an email on romantic weekend trips, her next communication will be in line with that and not just generic messaging.

It sounds so simple and straightforward, and yet think about the last time you booked a holiday, they asked you how many travellers but didn’t ask you your relation to them. Both a work trip and a romantic trip have some very clear opportunities for upselling for the travel company- hence the importance of thinking about those core questions and implementing segments as the first steps.

Getting Clever

What if we want to get smarter with our personalisation? What if we don’t want to send ads to customers who won’t spend with us? What if we only want to send emails to people at risk of churning? Here we enter the next level of complexity in which we need to start matching our first-party data on customers with various systems and channels. So how can we go about that and do we need AI?

Most companies will follow an ETL process when it comes to their data. That’s Extract, Transform and Load. They are extracting data from various different sources, transforming it by combining it with other data points, and then loading it into a data warehouse or into a visualisation tool. Reverse ETL does the opposite. It takes the data that has been transformed and then sends it back into applications such as marketing channels, CRMs etc. So for example you might model many data points to see which customers are at risk of churning and then use Reverse ETL tools such as Census and Hightouch to create a flag in salesforce so your team can take action.

But What About AI?

The idea of applying AI to personalisation is exciting, especially in the generative AI space. Perhaps one day we will get to a stage in which every visitor to every website will have a customised view, but that is still some years away! 


In order to effectively leverage AI, companies really need to have a solid bedrock in place in terms of their data. As more privacy limitations come into play, companies will be increasingly reliant on their first-party data to fuel these efforts and while the temptation may be to capture as much as possible, customers don’t like it and companies rarely leverage it well. 


So when you’re thinking about personalisation tools be wary. Nearly every solution has “Powered by AI” written on it today, but it isn’t strictly true. Many are leveraging the same advanced analytics that Reverse ETL employs but that isn’t really AI, it’s data modelling. You should also be wary of a solution is reliant on third-party data, and these tools may lose effectiveness due to increased privacy regulation.


In conclusion, why you DON’T need AI for personalisation:

  • Don’t focus on the technology, focus on which data will make a real difference.
  • All personalisation efforts are reliant on clean and modelled first-party data and without this AI will simply not provide any accurate results.
  • Reverse ETL already exists today as a great solution in this space.

1280 720 173tech