LinkedIn's Looking to Make Endorsements More Relevant - Here's How
Of all the various aspects of LinkedIn, "Endorsements" may be the most maligned.
That's not because endorsements are inherently bad, but it's more because no one really knows what they actually represent.
You see, a lot of people on LinkedIn (most, I'd hazard a guess) connect with folk they don't necessarily know. LinkedIn advises against this, so it's not the fault of the network necessarily, but it does lead to people endorsing others for skills they don't necessarily know they have.
Once you're connected, of course, LinkedIn prompts you to endorse those people, and because everything's a competition, the users who want to receive endorsements back will actively endorse all their connections for everything they can in the hopes that they'll return the favor.
And while that might make your profile look better - with all those little boxes filled up to the max - it also dilutes the meaning behind them - as most people now know that a lot of endorsements aren't accurate representations of real, professional competencies, it's hard to know what they actually do mean, and how much stock you can put into them.
LinkedIn's aware of this, and as such, they're working to improve endorsements and make them a more valuable way for people to recognize the skills of others.
What, exactly, that will mean for how endorsements are represented is not 100% clear at this stage, but over on the LinkedIn Engineering blog, they've outlined (in two parts) how they've been working to build a more accurate endorsements model, and put more focus on the endorsements of value.
As per LinkedIn:
"Realizing our goal of delivering endorsements that provide even more value to our members required a blend of research, new machine learning models, and a rearchitecting of the backend infrastructure that both serves and recommends new endorsements. Our solution not only serves endorsements at a faster speed, but also allows us to deliver more insights to members based on their connections and skills."
As part of their new endorsements research LinkedIn has narrowed down their endorsements system to focus on which endorsements are more likely to be relevant and accurate.
To do this, the LinkedIn team started conducting research by surveying members when endorsements were provided.
Based on these findings, LinkedIn established a listing of 84 different candidate signals about the endorser, recipient and their relationship that could be useful in determining the accuracy of an endorsement. They then narrowed those factors down to the 12 most useful signals that highlight an endorsement as being most applicable and related to a person's actual competencies.
As noted by LinkedIn:
"At the end of the process, we developed a target Endorsement metric that can be described as the following - A highly-rated endorsement is one made by a connection that knows the person and the skill."
That kind of makes all that complex research seem a bit redundant - I mean, anyone could have told you that, right? But LinkedIn isn't building a human rating system, it's working to establish the signals that their machine-learning system can adapt and utilize to help provide a more accurate picture of what an endorsement actually means.
"For each component of the definition (knowing the skill and knowing the person), we identified thresholds for their respective top signals based on intuitive cut-offs backed by machine learning results."
In the end, through this process, LinkedIn believes they've developed a system that can rank certain, relevant endorsements higher than others - but as noted, how that will actually be displayed is not yet clear.
"With these changes in place, we're ready to change the Endorsements experience. The Endorsements dataset living in the graph enables us to provide more highly-rated endorsements and improved endorsements insights to our members. This technological advancement takes us one step closer to realizing our goal of being the largest and most trusted professional peer validation system."
So LinkedIn's engineering team is "ready to change the endorsements experience" but right now there's no change evident. What this likely means is that you can expect to see movement on this sometime in the very near future.
A change to LinkedIn endorsements would be a great step, and will help underline how LinkedIn is using their vast professional data resources to build a more accurate, relevant professional network, and one that's more representative of people's actual skills. Refining that data is important - as we've noted previously, LinkedIn is in a position to change the human resources framework, to use their dataset to help provide better employee recommendations and help guide individuals towards their best-suited career path. But they can only do this is the information they're using is accurate - this is why refining their data models, like endorsements, is crucial work.
This also enables LinkedIn to further develop their machine learning systems to get a better understanding of professional networks and graphs, which they can then use in other applications to ensure the data they provide is more accurate, and thus, more helpful to those looking to genuine insight.
And at the same time, this new research also takes more focus away from number and vanity metrics - if LinkedIn can start to single out more relevant metrics, it'll help put more focus on actual achievements and skills, as opposed to gamifying such elements through social competition.
It'll be interesting to see how LinkedIn puts this new process into practice, but given they're writing about it, you can expect to see movement sooner rather than later.
Follow Andrew Hutchinson on Twitter