TL;DR: A recent study by Evergage Inc. revealed that only 30% of the surveyed companies are satisfied with the personalization efforts of their marketing teams. This shows that marketers need to step up their game and with this article we help them do so. We look at different approaches to personalization, explaining the two main trends: rules-based and automated with the pros and cons of each. Additionally, we show you how to apply both and which one might be the right choice for you.
Ways of Personalization
Everyone is talking about personalization but… Often when you hear the word, it is used as this incredibly broad term, encompassing everything! Like marketing magic, that increases everything from conversion rates and leads to sales. Well, that is of course not how things really work.
It is a hard-to-define concept as it means something different to everyone but to give a broad definition: personalization is showing targeted content to different segments of your audience to enhance their experience and increase your conversion rate. The ways of doing so can be explained through a spectrum of different methods, with rules-based and automated personalization as the two ends of this spectrum.
This spectrum also contains methods such as contextual segments, behavioral targeting, integrated 1st and 3rd party audiences, frequency and recency, 1–1 content attributes, and recommended content. But to be honest with you, while these may be presented as different personalization methods at times, they actually all fall under rules-based and automated (algorithm-based) personalization, which are the focus of this article.
It is important to keep in mind that the goal of both methods is the same, and the difference lies in the way of getting there. Now, let’s take a look at each of these approaches, define what they are, analyze their pros and cons, and help you figure out which might be the better option for you!
Rules-based Personalization
Rules-based personalization is when you manually create certain rules to target different segments of your audience, deciding on which content to show to whom and when. These rules encompass things such as location, weather, time, device type to referral source, abandoned cart, whether the visitor is a new or recurring user etc. The possibilities are nearly endless.
This is a manual process where you decide on the audience segments and create different versions of your page for each segment. On top, you set up rules to define which page variation will be triggered. It can be considered similar to A/B testing in the sense that it is a process of trial and error as you test different options and keep on tinkering and improving.
Pros of Rules-based Personalization
- More hands on: You are the one in control and you get to create audience-specific variations of your page.
- Can be done more easily, in-house: With the right tool, rules-based personalization can be very easy, as it does not require knowledge or implementation of complex algorithms.
- Capitalizing on existing marketing strategy: Almost all businesses already have some data and prior knowledge of their customer base and different segments, which means you are already halfway!
- Can reach big, important segments: You can apply a single rule and you already split your big audience into more manageable, smaller segments, and provided better targeting. Yes, it’s that simple.
Cons of Rules-based Personalization
- Very manual: While the hands-on approach gives you control, it does require more effort to set up each page variation and rule.
- Time consuming: This also means spending more time of course. Although the time it takes you to create variations will get less and less over time.
- You really have to know your audience: The first step to creating variations is knowing your audience. If your assumptions are not correct, the variations will not live up to their potential. But this is why you must keep on testing and improving.
- Variations are not super specific: It’d be a herculean task to create a personalized page for each visitor, consequently you’ll end up with a few, rather broad variations.
Final Notes on Rules-based Personalization
With rules-based personalization there is certainly a learning curve. As you start discovering new possibilities, you come up with new ideas and learn what really affects conversion. This will take some time and effort at first, but once you are on a roll it gets much easier and much faster.
Additionally, you could and you should keep on trying various options, this is the only way to know what works best in your specific case. Gotta keep on testing on your way to good, better, and best!
Automated Personalization
Automated personalization is personalization through machine learning algorithms, where each variation is automatically crafted for the individual visitor of the website. Decisions are made based on the existing metadata, as well as demographics and past behavior. The quality of these algorithms heavily relies on the amount of data you can provide. The more individual use cases you feed it with, the better they get at crafting an individual experience.
While the advancements in AI and machine learning will bring along new possibilities, today’s understanding of automated personalization is quite limited. It mostly refers to recommendation engines, used particularly in e-commerce.
Pros of Automated Personalization
- Targeting is easier: An algorithm can analyze data in more depth and identify more segments.
- More specific variations: More segments = many, smaller variations specific to each segment.
- Less workload in the long-term: Once everything is set up and runs smoothly, automated personalization saves you a lot of time.
- Requires less maintenance: It’s automated and constantly updating itself, duh…
Cons of Automated Personalization
- Harder to implement: It requires a lot of technical knowledge and an understanding of the algorithms. You will need a data-scientist to analyze your data and come up with possible personalization scenarios.
- Less control: You need to trust the algorithm and the results it will produce. Since it leads to numerous variations, performance for each is harder to monitor. It’s also important to keep in mind that the result might be different from what you had predicted.
- Harder to analyze results: Automated personalization means constant changes to what each visitor sees which makes analyzing the results produced harder.
- Costly: It is more suitable for very big companies with big pockets and big amounts of data, such as Amazon and Netflix.
- We are not there yet: There is much more to be discovered, built and improved in the area of automated personalization. The possibilities are endless but so far we remain mostly limited to recommendation algorithms.
Final Notes on Automated Personalization
When it comes to automated personalization there are three big names that need to be dropped: Amazon Netflix and YouTube, who are doing this at a very large scale with high levels of efficiency. However when looking at an Industry Report published by Amazon, one can see that even they acknowledge the hardships of automated personalization in the current day. Various methods that are either computationally expensive, or have scalability issues, possible fixes with the downside of lower quality recommendations and so on.
So for the time being automated personalization is not exactly within reach for many. That is with the exception of (big enough) e-commerce sites, where product recommendations have become the new standard.
How can I apply rules-based and automated personalization?
Use case 1
Say you are a travel agency wanting to implement personalization on your website. What can you do?
Rules-based example:
- Location + Weather: Living in Holland, my favorite example is this. If a travel website said to me “Leave the rainy weather in Amsterdam, for the sunny beaches of Spain.”, I would be pretty tempted. I am already very tempted. Let me check some flights…
Automated example:
- Recommendations: If you have a regular customer who likes trips to a specific destination, you (i.e. your personalization algorithm) can also start showing them trips to similar destinations, that other visitors with a comparable purchasing history have liked.
Use case 2
The second example will be from the industry where personalization is implemented the most: Online Retail! Say you have a website selling clothes from different brands; for men, women, and kids. You could definitely use some personalization here, so you don’t end up showing summer dresses to a man shopping for sports gear.
Rules-based example:
- Logged in customer: If a customer is logged in chances are you already have some information about their preferences (brand, size, color, product type,…). So you can curate their experience around this.
Automated example:
- Recommendations: Based on a customer’s purchasing history or products viewed, the personalization algorithm calculates an affinity score for each product in your inventory. The items with the highest affinity score will be displayed (i.e. recommended) first. Did the visitor look at a pair of black sneakers? Let’s show him more sneakers like that. Also, let’s use dynamic re-marketing to follow him around online, showing ads for the products he viewed.
- Another use case is social proof through “customers who bought this also bought…” page sections. If you want to step it up further, teach your algorithm to add a “you can combine this item with…” section for cross-selling other items.
Which personalization method is right for me?
Unfortunately, the answer is that there is no single answer. What you choose, what is best, depends heavily on your goals, resources (time, knowledge, money), and target audience.
However, there are a few indicators that can help you find an approach that is right for you.
First of all, ask yourself what you want to achieve with personalization. What are your goals? Then go through this checklist:
Personalization Checklist:
- Are you new to the personalization game?
- How well do you know your audience? Can they be divided into clear segments?
- Do you have an idea of what you want to personalize?
- Do you want to test ideas?
- Do you like to be more hands on; and in control for monitoring and analyzing purposes?
- So you have money to spend on personalization?
- Are you prepared to put in the time and effort?
- Do you have a small or mid-range customer base?
If your answers to these questions were mostly ‘yes’, you should opt for rules-based personalization.
Rules-based personalization is easier to start with as it’s easy to implement, it’s less costly and requires less data. Starting with this method, you can gain an idea of the possible benefits of personalization. This will also be a period of trial and error as you test different variations, while gaining a more indepth understanding of both — your visitors and customers as well as the value of personalization.
Later (if you have enough data and big enough pockets) you could implement automation on top of your existing personalization efforts. But while automated personalization has its benefits, it is probably not a good idea to make that your first step towards personalization. It requires heavy investment (time, effort, money, knowledge) right from the start. But don’t feel discouraged. Thanks to machine learning, automated personalization will become less costly and more accessible in the next years. Get a head start by learning the basics through rules-based personalization.
In short: With personalization, it’s better to test the waters first and dip your toe before jumping straight in.
What about Unless?
Starting with rules-based personalization we are discovering more possibilities each day and regularly adding new rules to our tool. At Unless, we are also working on small automations but since we know about the importance of staying in control (something that current AI-based solutions often lack), everything can be adjusted by you.
With our automation efforts we mostly focus on personalization based on sentiment, language, and content, not so much on testing design. Leave usa comment if you have any questions.
A special thanks to Ecesu Erol for her input on this article.
