I recently received this same question from several contacts and clients: with all these technology providers in the market and new LXP platforms being introduced, what is your favourite Learning Experience Platform?
I recognize the challenge (see also this great overview from Bersin on learning technology in general). Although I try to keep a close eye on all learning technology developments, it actually becomes harder to do so, as not only we see a lot of new platforms being introduced, but we also see many mergers in the LXP area (Leapest being acquired by Edcast, Pathgather acquired by Degreed), and on top of that the more ‘traditional’ learning management system providers like Cornerstone on Demand presenting themselves as LXP (and also their acquisition of Grovo is relevant for this article!). All in all these developments do not make it an easier job to select the right LXP.
I’ve started to write a series of blogs on how the introduction of learning experience platforms (LXP’s) is reshaping how we organize and design learning and what we need to do to get the best out of our LXP. I see 3 major area’s impacted. First is the way we design learning programs and experiences, next is the way we need to look at content, content management and curation of content, and finally the use of data to track the usage of the LXP and continuous improve not just the platform itself, but also the metadata describing the content and the catalogue(s) of content itself.
With all of this in mind, I am afraid that I have not yet seen a single LXP that meets all my requirements, and consequentially I do not have a favourite (yet!). I do however have thoughts on what my favourite LXP would look like. So here’s the list of requirements that I would be looking for on top of what every LXP platform promises: to “discover” and “share” (Bersin) relevant content/information and using Machine Learning/AI to help.
Content
My favourite LXP must have the ability to easily connect with any content repository, without (and I cannot stress this enough!) the need to physically upload content (or even a redirect package like AICC) to the LXP. It is interesting to see that many (if not most!) LXP providers also offer content. I understand why. The market for out-of-box content is bigger that it has ever been, yet I do not understand it. The reason why I would never choose an LXP that only works with content physically present in the actual platform is 2-fold.
First of all because content is everywhere. I strongly believe that we as L&D professionals should always look beyond learning content as such. The need for a colleague to learn something can be perfectly satisfied by a video, an email address to an expert or an online discussion. I believe most content (or information) in any company sits outside learning platforms. Think about document management systems like Onedrive, Sharepoint, livelinke and many others. Social tools like yammer, facebook enterprise and teams, think about all the content locked up in emails in your outlook or gmail. Within the scope of what we call learning content, most (large) enterprises do not have even have all content in a single platform. Many of them have contracts with the large content providers like LinkedIn, Coursera, Skillsoft, Plurasight and I could go on for a while.
My point is that in my ideal world, the LXP connects to all places where people generate, store and maintain content. I simply do not believe that we as L&D can convince our colleagues to completely rethink the content strategy for the entire organization and move all content creation and management to HR ownership, so let’s be smart and make sure we as L&D bring in tools that can easily connect to both content repositories we own and manage ourselves, but also the many more that exists outside learning.
The second reason why I want my perfect LXP to easily connect with content repositories is that we want our LXP’s to create personalized learning journeys for our colleagues. And in order to do so, we need machine learning/AI technology to help. For Machine Learning/AI to work, you’ll need data. A lot of data. Data is generated by people searching, accessing (registering) and consuming content. The more content you have linked to your LXP, the more data you will be generating, and the better your machine learning based suggestions will work.
Size does matter
As explained above, machine learning and AI algorithms need data to operate, learn and improve. Even the best AI algorithm cannot function properly without having access to big data. If you ask LXP providers about their main challenge, many will explain that they do not have sufficient access to relevant and quality data to train their algorithms. When this happens, they mostly refer to publicly available big data sets, but that does not provide the same quality results compared to using real learning data.
So, my favourite LXP platform has a lot of customers, and is built is such a way that the machine learning algorithms are fed from data generated by all the LXP customers. In terms of technology, this is not too complicated, in fact some providers like Cornerstone already offer this, however where this becomes interesting and complex is in being able to connect the data from different customers.
Leverage AI to manage data
Much of the promised Machine Learning based recommendations are based on 2 data sets. First, employee data is required to establish who are the people in the vicinity of an employee and can be regarded as a ‘peer’ from whom the algorithm look at training consumption: “people in a similar role as you, have completed this training, so you might want to look at it”. If you have worked in HR and especially HR systems and data, you will recognize that the quality of this HR Core data in many organizations is not what we want it to be; If you then consider how different this data can be between organizations, then realize that (as described above) you want your LXP to use data from it’s customers…it becomes dazzling to even start to think how you connect HR data between different organizations.
The 2nd set of data required in the LXP is the learning metadata. Metadata is data (or information) that describes other data. In other words, metadata described the training or other content available in your LXP. From my own experience, I know that making sure this metadata is consistent and the right level of quality is a huge challenge in itself. Again, thinking what is required to make the link between (learning) content from different organizations can make your head spin.
I’m not saying it cannot be done by hand. But if you have larger catalogues of 1.000s of objects, it will be a huge undertaking. I would like my LXP to help me. What that means is that the LXP includes intelligence to map employee as well as content metadata between customers.
You could even take this assistance one step further and use machine learning to create and improve your metadata; this machine learning-powered metadata is already more common with media content (think Netflix), and is steadily expanding to audio, text, pictures and other digital content formats. All of this with the explicit intend to make content more accessible for end users/consumers. What is an interesting concept is that using Machine Learning to manage metadata not just make the job much and much easier, it also opens up the possibility that the algorithm comes up with new metadata structures that we never would have considered ourselves.
Continuously ask for feedback
If linkedIn ‘serves’ you an advertisement, you have the ability to report the ad. LinkedIn then asks why you want to report it and you can tell you are not interested, you’re seeing too many adds, or you want to report because of something else. And as you report this, LinkedIn thanks you for the feedback. The reason LinkedIn is asking for feedback is to improve the machine learning algorithm that determines what ads to serve you. Feedback is important, as it will help the algorithm to learn to understand what works and what not.
My favourite LXP must be able to collect and process feedback in 3 ways. First it must use data generated in the LXP itself as feedback. An easy example here is data around how much time people watch (or consume) content. The time where we only track completions is behind us. If I find relevant information in a 30 minute video after 10 min, I will not bother to watch the rest. When I access a document with an estimated read time of 10 minutes, and close it seconds later, it clearly indicates it was not what I was looking for. This kind of feedback can be key to understand how people perceive the quality of content for example.
The second way to gather feedback is to ask people. Magpie from Filtered.com starts off by asking for not just your name, but also position and interests. Cornerstone allows you to select your topics of interest that then feeds one of the machine learning based learning carousels, it then also provides you the opportunity to rate the content on a scale of 1-5. The higher the rating, the more likely the content will be recommended to others. Like LinkedIn, my favourite LXP platform continuously asks for feedback, ideally in a very engaging and gamified way that for example rewards people who provide feedback…
The 3rd and most complicated way to gather feedback would be feedback on performance data. If we really want to integrate learning in the flow of work, we need to bring performance data into the LXP to allow it to identify and serve the right suggestions at the right moment in time (and preferably in the right format). Think of time required to execute a task, time management for people who struggle with their diaries, even feedback through tracking your behaviour and contribution to (online) meetings that could generate suggestions on improving your presentation or communication skills….the possibilities are endless!
Partnership, vision and roadmap
The last and most important requirement I always have with any technology provider is the ability to build a long-term partnership. I simply see no value to work with a technology provider that only asks me what I want. I want a company that understands technology, understands HR/Learning and business challenges, that has a clear vision on their role and where (and why!) they add value, a company that is honest on what is required to build success (no new technology can act as a magic wand that you wave to solve all your challenges!), and a company that has a clear roadmap containing all the right things.
Maybe even more important than the current roadmap of your LXP provider is that is has an established process to continuously assess and update the roadmap. Things are changing fast and what works today, might not work tomorrow. When you are on top of developments yourself, you would also want to influence the roadmap: add items, push items forward and make sure that what you really need is included with the right priority. A nice example of a solid roadmap is the one from Filtered/Magpie…their challenges is now to fill in the details and deliver on their promises!
(Please note that nor myself, nor SLT Consulting is affiliated with any LXP provider or has a commercial agreement in place. If you are an LXP provider and would like to comment on this article, please use the comments section, or contact me)