@mattab opened this Issue on October 7th 2018 Member

What is personalization?

Definition from wikipedia

Personalization, broadly known as customization, consists of tailoring a service or a product to accommodate specific individuals, sometimes tied to groups or segments of individuals. A wide variety of organizations use personalization to improve customer satisfaction, digital sales conversion, marketing results, branding, and improved website metrics as well as for advertising. Personalization is a key element in social media and recommender systems.
Technically, web personalization can be achieved by associating a visitor segment with a predefined action. Customizing the user experience based on behavioural, contextual and technical data is proven to have a positive impact on conversion rate optimization efforts. Associated actions can range from changing the content of a webpage, presenting a modal display, presenting interstitials, triggering a personalized email or even automating a phone call to the user.

Why personalization matters?

Personalisation can help companies improve their websites, increase conversion rates, increase average order value, customer lifetime value, improve marketing ROI, and deliver a better user experience.

There are a few tools that offer personalization but none that are open source and give you full control over your data. And a good personalization tool will need to collect heaps of data about users (so we can create these "visitor segments"), and have complex interface to let users define the "predefined actions".

And with Matomo Analytics and Tag Manager we're in a unique position to offer personalization as a service since we already have a whole platform dedicated to measuring users and creating segments (Matomo Analytics), and a product to let people define complex actions and triggers (Matomo Tag Manager).

How would personalization work in Matomo?

Using Matomo stack personalization could work in the following way:

  • User create visitor segments using Matomo segment editor: https://matomo.org/docs/segmentation/. With Matomo segments are very powerful and you can match on 85+ dimensions and metrics like Location, Behavior on pages/clicks/downloads/videos/forms interactions, User journey, Conversions, Purchases, Time of visit, and much more. See the list here: https://developer.matomo.org/api-reference/reporting-api-segmentation#list-of-segments
  • Matomo would then compile the list of segments matched by all known visitors, so that we have a list { visitorid -> [visitorid1 segments matched ], visitorid2 -> [visitorid2 segments matched], ... } `. If we do it right, it could even be almost real-time as the list could be lazy updated every hour or more often.
  • User creates a tag, using a personalisation trigger or so, that lets user select one of the visitor segments (eg. "User has visited the website more than 10 times so far"), and then define an action (popover, message, offering special offer, etc.).
  • When a visitor browses the site, Matomo Tag Manager executes and when there's a personalization trigger, the tag manager will look up the current visitorid and checks if it matches the segment. If it does, then the tag executes and the personalised content is displayed.

This personalization would need to be an extremely fast service, as easy to use as possible, work cross-devices on the web (ie. this will be the case when using User ID instead of Visitor ID), it would eventually be possible to combine it with A/B Testing and run tests to ensure that personalised content performs better.

@tsteur commented on October 8th 2018 Member

One file per visitor in one directory could "kill" / slow down your filesystem and expose potentially lots of information as eg a 16 string unique id might not be too unique. It would also not work for first time visitors etc. Plus it doesn't really work for things like country. There are likely some better ways to do this that I had in mind a while ago where couple of them are performed in JS directly and couple bit differently.

@shynes commented on July 19th 2019

Hello, is there any new code published to make this easier? @tsteur , I think he presumed it would be in JS anyway, I don't think he was suggesting one File be created , unless you are saying the logs are the files.

In terms of this, "list of segments matched by all known visitors, " I think he one file will be created.

@tsteur commented on July 20th 2019 Member

@shynes there isn't any new code published to make this easier. Having this in one file be problematic potentially as well since you would have a lot of information about many visitors in one file publicly available and big loading time etc. There are heaps of problems / challenges with the suggested solution and it wouldn't really work unfortunately.

@shynes commented on July 20th 2019

@tsteur, thank you so much for the response. What did you have in mind as a better way?

Maybe not storing them in a file, but maybe doing a query against the segment database on maybe some low overhead environmental variables of the visitor? Once they login, we can store many variables in the database.

@tsteur commented on July 21st 2019 Member

@shynes you would need a new endpoint for personalisation and start step by step in what can be targeted. Eg Location, IP, etc. Some other things could be possibly resolved in JS directly, eg if personalising only based on user agent, local time, ... without needing to fetch any data server side. Ideally it wouldn't hit the DB as it may be too slow for personalisation unless maybe https://github.com/matomo-org/matomo/issues/14119 is implemented for faster visitor lookup. It's a complex feature so hard to summarise in a few sentences :)

Powered by GitHub Issue Mirror