Web Analytics

Creating Personas with Adobe Analytics

This article is an echo of Matt Isherwood article on Medium : https://medium.com/@ishmatt/your-websites-persona-how-to-quickly-find-out-with-google-analytics-data-20312ad9a360

I wrote a lot of technical articles on this blog and I cannot claim that Web Analytics is for everyone if I keep posting codes.
It is not necessary to know coding for doing web analytics, not at all ! Sure it helps to do even more stuffs but I think it was time for me to start again doing some articles for business practitioner. This will be an easy one, thanks to Matt that chew most of the work.

Use Adobe Analytics segmentation for creating Personas

What are personas ?

As Matt explained on his article, you usually create Personas to define the high level customer of your company. If you are an online retailer it is the basic unique visitor type that you can encounter on your website. The process of creating persona can be quite complex and as good as web analyst can be, it is usually not good enough to define those personas. Web Analytics is good to find pattern and valid statistically some hypothesis or ideas you have about your visitors but something that web analytics is less good at is answering the “why” question.

Of course you can know that 30% of your users is bouncing back from your homepage but you don’t know why. To know why, you have to do tests and get experience. This is how experience help but can also be a bias as you think you know the why… but do you really ?

For creating personas, you need to understand the intrinsic reasons of the visitors’ behavior. This is usually done with UX persons, in a room where the user explain the reasoning behind each of his actions or motivations. Doing this kind of studies is actually harder than what most people think. You need to avoid bias and be very scientist in the approach. Not everyone can do that as they are usually more eager to have their theory confirmed than to wait patiently for the outcome of their tests. So bias interaction can be easily introduced into the process.

But if web analytics is not able to give you the intrinsic reasons of your users it can help you giving you a really good idea of who they are.

Where to start building your personas ?

When I taught to my colleagues about Analytics, I (tried to) always end up my training by Segmentation. One of the sentence I am always referring to is : “Without segmentation, your data means nothing”

It is harsh but it is quite true.
Think about it…

If you know that 35% of your users is converting. Wouldn’t be interesting to know if they are coming from your marketing campaigns ?
I mean, you probably spend some money on them, it would be the minimum that there are some results out of it.

And does your mobile website reach the same level of your desktop website ? Or are you missing conversion of hundreds / thousands of users without even knowing about it ?

For most web analyst, these are trivial questions that you (always) consider when building an analysis, but for most of the business manager (which are not always web analyst) they are very less obvious. Especially as getting a number (ie 35%) is like getting a result, even if this number means nothing itself. It takes one to know one.

Recommended segmentation for personas

The segments that I would recommend when you start you personas hunt are the following :

  • Geographical Segments : because we may be in the internet where everything is borderless, people from different region of the world / country may have different behavior. Like the guy in the south looking for sunscreen in September, where the guy in the north is already looking for a bobble hat. 😉
  • Buyers vs Non Buyer segment : This one is an easy one, I hope every one had it. By buyer, it could be a lead generation. whatever end-goal your website has. Divide the website between who does it, who doesn’t.
  • Mobile vs Desktop (vs Tablet) : Determine your personas by the type of device the user is using. A different device mean a different experience. You are not browsing the same way on a computer than when you are on your mobile.
  • Going a bit further, you can go to Operating System, because the Apple users are probably going to act differently than the one using Linux.
  • You can look at your marketing campaign and see if people that arrive from a specific channel has a different behavior than the average user.

So this already give you a nice idea of who your customer are. It is already quite a lot if you really consider all of the possibility above.
You can look at those different possibilities :

  1. User from Desktop Apple that accomplish your goal
  2. User from Mobile Apple that accomplish your goal
  3. User from Desktop not Apple that accomplish your goal
  4. User from Mobile not Apple that accomplish your goal
  5. User split by geographical regions and determine their conversion rate
  6. User split by Marketing Channel and determine their conversion rate.

You can see that it starts to get a lot of data to process, and obviously not all of them will give you relevant information.
Unfortunately there are no easy way out and I would recommend you to :

  1. Identify your biggest global segment (Mobile / Desktop / SEA / Germany / France / US / Converters / Non converters)
  2. Select maximum 5 of them. You can group them if you need to.
  3. And compare the data between one another.

If you have identify your 5 segments, you will have enough combination possible to keep you busy.
It also depends if your segment are antagonist (as you cannot be on Mobile & Desktop at the same time).

Going Further with Adobe Analytics

On the good post from Matt, this is mostly what you can do with Google Analytics, but with Adobe Analytics, you can (a bit) futher.

Here are a bunch of idea that you could apply :

  • One of the best feature of Adobe Analytics is the classification, what you can do is to import customer data (not identifiable information though 😉 ) and try to look for new pattern from what you know.
  • What you can also do from classification is basically import your BI segmentation in Adobe Analytics. True it is not finding anymore, just copying. 😉
  • You can use the segment comparison (Venn visualization) to discover if the segments that you have created are actually matching the same users.
    Venn Diagram
    On the example above it seems that the 2 population are quite distinct.
  • When you encounter an anomaly, if that makes sense to you, run an analysis.
    This use the power of the Adobe Sensei (Machine learning tool) to check between dimension and try to find the correlation.
    This can give you insight into specific pattern of your users. Why lots of users didn’t behave the same way than the average user ?
    I ran one when I detected an increase on my website and here are the type of result you can expect. As you can see, the bigger the correlation, more probable this is the reason.
    Maybe I should do a persona for users actively looking for automatizing the Adobe Analytics implementation 😉
    Contribution analysis Adobe Analytics

My recommended approach for creating Personas

For Personas, as I explained above, this is often the job of UX specialist with lots of qualitative method (opposed of quantitative approach such as Adobe Analytics). However, as I demonstrated, Adobe Analytics can really help you covering most of the ground work. With this approach I would also recommend to have a brainstorm internally of how you expect the user to interact with your website.

To give a bit of advice from experience, I would give you type of Persona that I have seen the most :

  • Explorer : User that browse to discover the website. (S)He may be quite new to your website, generally content or information of the product drive her to your website. (S)He rarely starts his/her journey from the Homepage.
  • Finder : User that browse to find a specific information. (S)He may be using your website as external source of information and not to its primary purpose. (ie : buying offline but checking competitor online)
  • Discount searcher : user that looks for promotion.
  • Buyer / Consumer: user that is here to buy but browse at the same time. (High pv/v on their buying visit). There may be potential here for cross selling.
  • Fast Buyer : user that already been on a precedent phase (Exporer / Finder) and knows already what he wants.
  • Re-Orderer : User that come naturally to re-buy always the same type of product. (work also as re-consumer : For news site : user always check the same category and look at 4 articles).
  • Abandonner : The user that starts a process but don’t complete it.
  • Contributer : The user that is seeking to interact with the website. (Twitter / contact form / etc…)

I think that there are 2 schools about the personas. Depending on the context, if you sell very expensive product, users can be one persona at a time of his visit, but don’t imagine that it is not possible for user to switch persona during a single visit.

I would say that this is the goal of the Conversion Optimizer to make sure that some type of persona are changing during their visit. You want people to become more and more buyer.
You want users to become more and more consumer. Even for non e-commerce website, having a single type of user is not the best.

Of course, you HAVE to adapt those personas to your website content.
Per example : There are Football fans vs Tennis fans for a sport website. Mutli-sport fans have to be consider as well.

Personas is a kind of endless topic but doing the research on them for your website make you understand what is your audience and where opportunities lie. At the same time, you better understand if you are doing the right thing for the ultimate business goal you have been assigned.

You spend your time doing content when actually you already have a lot of new users and high pv/v… if you are not a content website, you may want to change your strategy. 😉

I hope this was helpful for you.

As usual, don’t hesitate to post a comment if you want to add something.

Leave a Reply

Your email address will not be published. Required fields are marked *