• personality-based-marketing

The A-Z of a Personality-Based Marketing Campaign

Personality-based marketing can seem daunting. For many, it is a brand new way of understanding customers and creating content. So it’s not always easy to know whether it’s right for you, to see how it might fit into your organisation and workflow, and to understand the value it can bring.

With this article, I hope to clear things up. We recently ran a campaign with a leading bank, where we increased Conversion Rates by 59% through matching content to customer personality. Using this bank –  referred to throughout as ‘DataBank’ due to NDA restrictions –  as an example, I will go through the process of running a personality-based marketing campaign step-by-step; from the data and analysis required to understand personality through to personalising content and measuring the results. As we go, I’ll be looking at how AI makes this all possible.

Step 1: Understanding customer personality using AI

The data

Our personalities are reflected in the decisions that we make every day.  So the first requirement for understanding a customer’s personality is data that shows this decision making. With our retail banking clients, this typically means information about where customers choose to spend their money. For DataBank, we used one years’ worth of debit card and credit card spending data for 500k customers.

At DataSine, we’ve built models that can handle all sorts of behavioural and transactional datasets, one of which is how a customer has engaged with previous email campaigns. Here we take emails that have been sent to customers, rate the elements contained within them according to how appealing they would be for different personalities, and then look at which element each customer clicked. For instance, if an image is deemed to have high appeal to Extraverts and customer x clicks on that image, then customer x is likely to be an Extravert. Take multiple data points like this and you can build an accurate picture of a person’s personality.

Other datasets

At DataSine, we’ve built models that can handle all sorts of behavioural and transactional datasets, one of which is how a customer has engaged with previous email campaigns. Here we take emails that have been sent to customers, rate the elements contained within them according to how appealing they would be for different personalities, and then look at which element each customer clicked. For instance, if an image is deemed to have high appeal to Extraverts and customer x clicks on that image, then customer x is likely to be an Extravert. Take multiple data points like this and you can build an accurate picture of a person’s personality.

Data security

Of course, privacy and ethics must be front and centre when it comes to customer data.  Using third-party data, such as Facebook likes, can be tricky due to the ethical and regulatory issues that can come with it. So at DataSine, we stick to first-party data like in the examples given above. We also avoid storing any personal data. If we can access the required data through a secure API, for instance by plugging into HubSpot or Mailchimp, then we assign each customer a unique, but anonymous, identifier. This removes the need for us to store email addresses or names. If that’s not possible, as is often the case when accessing transactional data, then we perform the analysis onsite at our client’s offices.

The analysis

Our ability to predict personality and content appeal is down to the advanced machine learning techniques pioneered by James, our Chief Scientist, and his team. AI allows us to uncover complex patterns in how transactional and behavioural data is linked to specific personalities and how customers engage with content.

To give a very simple example – as there are many small nuances that our algorithms are picking up on – people that spend more money on restaurants and bars are more likely to be extraverts while people that spend more money in bookshops are more likely to be introverts. When it comes to content appeal, images that contain people will be more appealing to extraverts while images that have more solitary themes (e.g. a beautiful and remote cove by the sea) will appeal more to introverts. This lines up with current thinking in psychology and neuroscience suggesting that extraverts need more social stimulation than introverts.

Personality segmentation

Once we have the personality profiles of every customer, we take their strongest traits and segment the customer base accordingly. In the case of DataBank, we focused on six personality segments:

  • Extraverts
  • Introverts
  • Agreeable
  • Competitive
  • Open-minded
  • Traditional
Segmentation: personality versus demographics

Off-the-shelf segmentation solutions divide your client base into groups with similar demographics and behaviour. This is a great starting point for creating better content and workflows, but actually understanding these groups is left up to you. For example, figuring out what motivates urban-dwelling professional women in their 30s to purchase your products and services requires careful research, which doesn’t scale well as you add new groups or new products.

Luckily for us, psychologists have already spent decades building a system for grouping similar people together. Rather than generating arbitrary behavioural segments and studying them one-by-one, we use machine learning to predict each customer’s personality according to the Big Five dimensions. Using our extensive datasets, we can then attach detailed preferences to each customer, such as how and when they like to be communicated with. This reduces any research and analysis workload from months to minutes and removes the guesswork from strategic planning in a data-driven way.

The Big Five Personality Model

Step 2: Personalising your content

A/B and multivariate testing are great optimisation tools, but on their own, can’t offer much insight that’s repeatable for future campaigns. AI combined with insights from psychology can.  You can know in advance how appealing different images and words will be, understand why something works well, and tailor content to the individual, rather than to what works best for the majority – all automatically.  I want to stress that this does not mean we are building some sort of automated persuasion machine. This is about replacing arbitrary segmentation and generic communications with a personalised customer experience, that is tailor-made to an individual level, fit for the twenty-first century.

Here is how the process worked with DataBank.

Many types of content can be personalised to customer personality – landing pages, text messages, call scripts – but in this instance, we were personalising emails. Our platform can give recommendations for how to match a whole range of elements within the content to personality – words, images, colours, layout, and themes – but here we focused on images and words.

The process

  • Import email: emails can be both created directly in our platform and imported as an HTML file or via our integrations with CRM and Marketing Automation Platforms such as HubSpot and MailChimp. For this campaign, the import via HTML function was used.
  • Create variations: As we were using six different personality segments, we needed to create six new variations of this email. This is done automatically through the platform by selecting the different personality segments we want to personalise for.
  • Personalise images: Our platform analyses the header image of the original email and tags it, in this case,  with things like  ‘travel’, ‘airplane’, and ‘airport’. Then by clicking on it, we are immediately given recommendations for images that would be more appealing to the chosen personality segment. By default, these recommendations come from UnSplash (a repository of high quality, free-to-use images) but they could come from any image databank provided.
  • Personalise words: Our platform gives recommendations for increasing appeal at both the level of the individual words (e.g. changing ‘wonderful’ to ‘enticing’) and the sentence (e.g. focusing on the scarcity of the product rather than the quality)

Resulting email variations

Step 3: Analyse the results

So does this work? In short, yes. Incredibly well actually. In the case of DataBank the main KPIs were Click Rates and Conversion Rates (i.e. the percentage of customers that completed the credit card sign up form after clicking through to the landing page). Using personalised messaging tailored to the aforementioned personality segments, we were able to increase Click Through Rates (CTR)  by 59% and Conversion Rates by 22%. This was not an isolated event either.  To give just two other examples, Hello bank! Belgium increased customer engagement by 80% and a leading French bank increased sales by 71% using our approach.

These results are not so surprising when you consider that currently, the typical marketing email conveys the same message to very different people. As if you would meet up with your friends, all with different personalities, and talk to everyone in the same way. Doesn’t sound right, does it?

When running projects with our Enterprise clients, we always set up a rigorous experimental design that controls for every factor and variable so that we provide proof of our method in a scientifically valid way.

With DataBank, there was roughly 30k customers in each of the six personality segments. We split each segment into two – one half received the personalised emails and one half received the control email.  This told us how sending personalised emails to the correct personality segment compared to sending the original generic version. But we also needed to test whether the personalised emails were just generally “better” or “worse” than the original email (i.e. personality was irrelevant). So we added in an extra 150k DataBank customers (on top of the 500k) as a separate control group with a random mix of personalities. This group was sent a random mix of the control email and the six variations.

This meant that there were 18 groups of customers in total. Here’s a breakdown of the full results:

  • Proof that our personalised emails work: when sending personalised DataSine emails to the correct personality segment we achieved a 17% increase in CTR and an 81% increase in Conversion compared to sending the non-personalised control email to the same personality segments. Both results were statistically significant at CI-99, p.008 and CI-99, p.001.
  • Proof that our personality segmentation is accurate: when sending personalised DataSine emails to the correct personality segment we achieved a 9% increase in CTR and a 154% increase in Conversion compared to sending personalised DataSine emails to the control group with a random mix of personalities. The conversion rate was statistically significant at CI-99, p.001.
  • Proof that personality matters: When sending both the personalised emails and the non-personalised control emails to random people there was no significant difference between the two in CTR or Conversion.

Find out more about our ‘Enterprise’ offering here. The ‘Premium’ version of our platform is currently available for free; you can sign up to be a beta user here, and we’ll be giving out access over the coming weeks. This version of the platform integrates with HubSpot and MailChimp and provides details on the personality of your customers and customer base, reports on the appeal of your past email campaigns and potential for improvement, and inline recommendations of images that can deliver greater appeal.

2018-09-11T11:41:54+00:00