Personalisation and AI
Background
I wrote this originally as part of a course I was taking in UCD on Artificial Intelligence for Business. It tells a story of my ambition to personalise our betting product - Sports Interaction.
Introduction
Content personalisation is a strategy for software products in which web pages, and other forms of content are tailored to match the characteristics, preferences or behaviours of individual users.
For a customer-user this means that decision making is easier because the junk is cut out. Junk, meaning all of the content in the product that the person does not want to use, that is in the way of what they actually may want to do.
This case study looks at a business – Graphyte, that provides B2B solutions for software providers to personalise their products for their customers. They specifically target sports betting products which are complex, constantly changing, and invite a lot of regular customer interaction.
Graphyte has in 2023 been acquired by Optimove who make products that allow marketers to specifically target user segments for offers and information. This allows both functions to combine their expertise – user segments combined with personal behaviour and potential behaviour within that segment.
Situation
Betting products are becoming increasingly complex. Competitors are tripping over one another trying to fill feature gaps. Product managers are rewarded for driving innovative features. What often results is a pile of features laid in front of the end user who is frozen by decision overload. Personalisation is one of the key areas where a product can be competitive, and for a given individual, appear different in a sea of similar products. Graphyte is therefore driven to improving specificity to better serve their business customers. The end user is ready for a more personal experience and has diminishing patience for junk content.
Background: What has happened so far
Software personalisation has thus far gone through three main stages.
1. Customisation
2. Customer segmentation based on data
3. Tailoring based on historic behaviour.
Customisation
The user changes settings to make the product more useful for themselves. The business could track what choices the customer was making but was not pushing content specifically to this user based on these choices.
Customer segmentation
Data gathering allowed for customer segmentation. Recording who did what and when. Volume of actions taken and category of action. For instance, how much money was deposited, in what increments and when.
With this data marketers could now target customers based on segment, but not yet by individual.
Tailoring based on historic behaviour
As the data became more reliable, Graphyte was able to offer personalisation based on historic individual behaviour. This means that if a person bet on football last season they’ll do so again this season so the business shows football when the season starts and highlights big games during the season based on favourite teams.
…Graphyte is powering the on-site experience received by millions of gaming customers every day. Teams across marketing, customer experience, product, operations and engineering use Graphyte daily to deploy personalised experiences, dynamic lobbies and hyper-target marketing campaigns through CRM and third-party advertising channels.
https://www.linkedin.com/company/graphyte-ai/about/
From ‘Graphic Product Overview.pdf’ provided by Graphyte.
It also sends push notifications to customers when something they are interested in is about to start.
People, process and technology
Software personalisation will rely heavily on technological factors. Reliable data points can form a matrix to paint a picture of the user profile – behaviour, cadence of use, user location, interaction styles combined with complex correlation of behaviours.
The effectiveness of this technology for business and for end users however will rely on people and process factors.
The business mindset
The approach for a business must shift focus from looking at historic customer needs to predicting future customer needs. This way the business is behaving more like a helpful assistant and less like a pushy sales person.
The product can then evolve from something a business is presenting to the customer to a conversational tool. The product only contains what the customer needs right now and all of the other features that they don’t want, the noise, is available but demoted, out of the way. The product appears to be listening and empathetic to the user’s needs.
Customer
For the customer – the interaction with a product is increasing more like a conversation than a lecture. A person should feel like this is their product, it is serving their particular needs and paying attention. It is no longer pushing a pile of choices but asking questions and learning from those answers directly or indirectly. The products a person will choose are those that allow a higher quality of living. In the case of Graphyte / Optimove, ultimately, they aim to give a person what they want. What they want is to enjoy a sports game. Betting is a means to do that. Interacting with a betting product is a means to do that. Neither are the end goal. AI will remove or reduce these obstacles between person and enjoyment of a game.
Product designers
When the product comes to the point of hyper personalisation, and each customer is seeing a different version, it’s going to be pretty tricky to see what the actual user experience is. The AI solution will mean that designers and developers will have to re-think the architecture of the product. It can no longer be a single built unit. The approach will have to lean towards a modular, componentised assembly of parts that work together. A user may see any combination of those parts switched on or off depending on what the personalisation model is telling it to do.
Marketers – seeing value
Marketing has undergone huge changes in the last 10 years. Marketing software has allowed more focused customer segmentation, allowing incentives to be given only when they will yield the most value. Optimove, the parent of Graphyte has created some nice tools for marketers to do this. Ultimately, for betting products it amounts to giving customers bonuses and free bets in different ways in order to drive more activity in the product.
When hyper-personalisation also leads to micro-incentives that can be delivered and measured automatically, the marketer can step back from pushing bonuses and focus completely on the customer relationship.
In a competitive market the effort put into building customer relationships is what will really pay off. The bit that the machine can’t do on its own.
What the future holds for Graphyte/Optimove
Data strategy will continue to be the foundation of success. It is only with clean, reliable data that they will have the means for transformation.
Forward-thinking businesses are focusing on building clean, reliable data. There is always a balance of value versus effort of course. Those businesses lucky enough to have visionary business analysts who see, and communicate the value of this will reap the rewards later.
Automation – Eliminating the easy decisions
The automation part will happen first. The parts of customer personalisation that are based on simple rules, for instance, if a person is currently located in the vicinity of a ball park, the game is about to start, they have recently signed up for this product and bet on baseball before, let’s give them a little reminder that they can bet on the starting pitcher. Its probably what they’re talking about with their friends going to the game anyway.
Cognitive Insight – Knowing the customer
The second stage will be cognitive insight. Hyper personalisation of what’s presented using correlation of behaviours.
Secondary behaviour will come into play next – cadence of use, time on product, manner of interacting, skin temperature, tone of voice. All of this gives the product more information about the customer, allowing micro personalisation specific to this interaction. This secondary behaviour will also come strongly into play when problem patterns need to be identified.
Continuous improvement as part of the data strategy will be essential. More information will be useful to record of all different types.
Cognitive Engagement – the product relationship
The nature of a product interface is changing. Increasingly, the user can have a conversation with the product about what they’d like to do next. In the case of the betting products that Graphyte serves, it can become more like talking to a person while watching a game. ‘What do you think will happen next?’ ‘Let’s see if you’re right!’ The product is learning and getting to ‘know’ the customer better at each interaction.
Ethics
There is a very grey area between marketing and manipulation. Graphyte/Optimove will need to work with the enormous social and legal pressure to prove that it is behaving ethically. Gambling and gambling-adjacent businesses face constant scrutiny on ethics. It must deliver a valuable product that enhances the lives of people. To survive, a brand must be known to reliably deliver safety on top of that.
Graphyte, for its customers can recognise patterns and anomalies associated with problem gambling; certain cadences of use, rapid scrolling, interface mistakes, shaking the phone combined with hand temperature. The software can be built to include calming or cut-off barriers for the user.
Conclusion
Graphyte is providing a valuable service to software providers. It is on track to developing an intelligent product to help businesses deliver a tailored lifestyle application.
Its success may come down to how focused it stays on outside factors – social, ethical, legal and human factors.