In part 1 of our personalization strategy series we talked about getting started and whether personalization is right for you. Last week, in part 2, we talked about the importance of prioritization and testing. If you have not read part 1 and 2 yet and you are just getting started with personalization strategy, go back and start reading
Finished? Good. Because this week we talk about the big stuff.
We are going to talk about rules-based personalization, predictive personalization and how predictive personalization works.
Are you ready? Let’s go!
What Is Rules-Based Personalization?
Rules-based personalization is exactly what it says it is.
It is a personalization method that uses simple rules that can be adjusted quickly, with a number of attributes, that divides your audience into small segments. These segments can then be targeted by you to sell your products or services.
If you are familiar with computer programming, rules-based personalization works a bit like an IF/THEN loop. Your rules are the “IF”. So for example you may have a segment that is “20-23 year olds in Bangkok that have purchased Nike trainers”.
Your rule for this segment may be if those 20-23 year olds arrive on your site, your “THEN” part of the loop could present them with other Nike products, including trainers.
This is a rather simplistic way of explaining rules-based personalization but at its core, rules-based personalization involves you creating rules based on your segments and serving content to those segments based on the rules you have defined.
Rules-based personalization is what most businesses who have reached a “halfway” stage on the personalization maturity curve are comfortable with. It’s far less complex than predictive personalization and, with the right team, you can sense and respond to changing customer needs fast.
What Is Behavioral Personalization?
If you have come this far then maybe you have experimented with rules-based personalization and you have well-organized and actionable data. But you are not quite ready to move into predictive personalization just yet. It is at this point you might start testing out recommendations based on user behavior.
In part 1 we discussed implicit and explicit personalization and the difference between them. Behavioral personalization falls very much on the “implicit” side. Here you would look at behavioral data and then make recommendations to the user based on the behavior that you have tracked on, for example, your website or app.
This behavior will likely also form part of the rules for your rules-based personalization strategy. But the benefit here is that with the right data, insights and user tracking technology in place you can make personalized recommendations to BRAND NEW users that you do not know anything about. You would simply look at a user’s behavior and then make recommendations to them based on that behavior.
While that may sound more simple than most things we have discussed in this series, it really is not. You can only get to the point of making behavioral recommendations by breaking down silos in your organization, solid data collection, testing and turning your data into insights.
Personalization is a process, you have to take the first steps before getting to this point in the journey.
The Omnichannel Stage
Before we head to the big, complicated “predictive personalization” stuff, we need to talk a bit about omnichannel optimization. This is different but related to O2O, or offline to online (and vice versa) strategy. Because we like to confuse ourselves, O2O is sometimes referred to as omnichannel marketing/messaging.
In this context we are talking about your organization.
This is because if you really want to present true personalized experiences to all of your customers, organizational changes need to take place.
By this point in the journey you may have some good results with which you can make decisions and even double your efforts. But if your organization still has different sales, marketing and customer facing teams working in a legacy way, you will fail.
Every channel needs to speak to each other and act as a single unit. With a focus on customer experience. Your marketing communications channel should be sending and receiving data but so should your customer support channel and your sales channel.
If you need help aligning your teams to become more customer centric, talk to MAQE via firstname.lastname@example.org. We are on this journey too, so we can give you some advice!
What Is Predictive Personalization?
Predictive content personalization, which is sometimes called machine-learning personalization, is an advanced and AI-powered way to display personalized content to each user at scale.
It is different to rules-based personalization as it does not use segments. User identification takes place at a granular level. After this, the algorithm will create a personalized experience for that user. It is however, dependent on the user data you actually hold as it focuses on intent. Often predictive personalization uses a mix of signals to create predictive recommendations and experiences for users.
How Does Predictive Personalization Work?
This is a VERY complex topic. We have written an article on how personalized product engines work that delves into some of the details.
But predictive personalization makes use of algorithms. Rather than getting too complex, we are going to identify the sorts of algorithms that are used in simple terms:
Basic: A basic algorithm is able to serve experiences without using personally identifiable information. For example on an e-commerce marketplace you can show new products, promotions or what other users are browsing in the store right now.
Advanced: Advanced algorithms customize content to each user by using personally identifiable information. They also use previous or demonstrated behaviors.
So if you have demonstrated behavior data, an algorithm can assign individual users to a cohort of users with similar demonstrated behavior. The algorithm will then dynamically predict other things that each user will like.
This can have a big impact on conversion rates. An algorithm can create decision trees for each cohort, and an individual within that cohort, that will have the best chance of leading to a conversion.
This is the kind of system that Netflix and Spotify use. Netflix published a detailed Medium post about the process which is an informative look at how they put personalization in place on their platform.
Predictive personalization at scale relies on data science and machine learning technology. This is beyond the reach of many companies but by working with an external partner, like MAQE for example or Recombee, you can start to make use of this cutting edge technology to increase return on investment and user engagement.
But before delving into predictive personalization and the exciting technology it uses, you need to have some “must haves” in place.
The Five Must-Haves Of An Effective Personalization Strategy
We have gone into a lot of detail in this series so far, which we will update as technology and trends change (part 4 is coming soon…). But over the course of the series several key fundamentals arise that are essential for anyone who wants to build an effective personalization strategy.
Break Down Those Silos
We have mentioned that aligning sales, marketing and customer support are essential if you want your personalization strategy to succeed. We would go one step further. Consider creating a customer experience team that has design, UX, marketing, sales and customer support all actively involved at every step. This will help you generate a “single customer view” in your organization and put you closer to the customer.
This is a big organizational change, but a valuable one.
Collect That Data
You cannot embark on a personalization strategy without knowing your customers, what they like and what they do on your site. You need to know your customer at every touchpoint. Be sure to talk to them and not just use analytics as they do not always tell the whole story.
Turn That Data Into Insights
Data, without anything else, will not get you anywhere. Organizing that data and then analyzing it is the only way to make sense of it all.
Test, Test and Test Again
Use the insights you gain to run experiments. Then run them again. Personalization relies on experimentation. Do not be afraid to “fail” or disprove your assumptions. That is all part of the process.
Come back soon for part 4 where we will be talking all about data. We will look at Data management platforms, first-party data and how it all fits together with predictive personalization technology.
Talk to MAQE
If you need help with personalization, talk to MAQE. We can help you in every phase of your journey. We can even use machine learning to help you personalize at scale. Get in touch with us via email@example.com to discuss your needs.