If you’re in sourcing, recruiting or HR, you no doubt search LinkedIn from time to time or perhaps even every day.
So why not gain a better understanding of how LinkedIn search actually works?
And what better way to learn how LinkedIn search works than from the Heads of Search Relevance and Query Understanding at LinkedIn?
Yes, you read that correctly – LinkedIn has folks specifically dedicated to understanding your queries in an effort to return the right set of results, emphasizing query rewriting, elaboration, and refinement.
Here is the LinkedIn Search Slideshare deck where you can learn about, among other things, LinkedIn’s contextual word sense and how LinkedIn deals with “keyword stuffers” and spammers.
If you’re interested in learning how you can more effectively source talent from LinkedIn and you aren’t one of the 5,400+ people who has already registered for my LinkedIn sourcing webinar (I know, right?!?!), you can click the image below to sign up to attend on Tuesday, June 4th at 1:00 PM ET.
The content I’ve prepared covers quite a broad range, from the very basics of Boolean search logic to some of the most sophisticated ways you can source potential candidates on LinkedIn, whether you’re using LinkedIn for free or you have one of the premium Talent Solutions, up to and including LinkedIn Recruiter.
The interest in leveraging big data, analytics and Moneyball in HR and recruiting is gaining significant steam.
Ever since my first article on the subject back in 2011, I’ve set up Google Alerts and Hootsuite streams set up to catch any mention of big data, analytics and/or Moneyball in conjunction with HR, sourcing or recruiting, and the volume of activity is bordering on surprisingly massive and overwhelming, and I’m not the only person to notice this.
Yes, it does seem like everyone is talking about big data in HR.
However, as is often the case with relatively new and nebulous concepts, there is quite a bit of confusion surrounding big data and Moneyball and how they can be applied to HR and recruiting, as evidenced by the obviously incorrect usage of the terms in many cases. It’s also nearly impossible to stay on top of all of the content being generated on the subject (although I am trying my best!).
This is precisely why I’m going to take the opportunity to clear up any confusion by concisely explaining the concepts of big data, analytics, and Moneyball as it relates to HR and recruiting, as well as illustrate some obviously incorrect references to these concepts in recent articles, including those from the Wall Street Journal, Forbes, The Economist, The New York Times, and more.
I’ll tackle analytics first, big data second, and then Moneyball in HR/recruiting, leveraging Slideshare presentations and YouTube videos from experts for support. Read More
I’ve compiled a list of the top 100 most connected people on LinkedIn, who represent the top .000044% of LinkedIn’s reported 225 million+ registered users.
Here are a few interesting facts:
There are only 5 women in the top 100, but 2 are in the top 10, and a total of 4 are in the top 20
The U.S. accounts for more than half of the top 100 (55)
The other 45 members represent the following 15 countries: U.K. (11), India (6), Netherlands (5), Canada (5), Brazil (5), Australia (3), U.A.E. (2), Turkey (1), Spain (1), South Africa (1), Israel (1), Singapore (1), Venezuela (1), Monacao (1), and France (1)
As might be suspected, a solid percentage of the top 100 are in staffing/recruiting/HR (28%). Other highly represented industries include I.T., Management Consulting, Financial Services, and Marketing and Advertising.
Beyond those top 5, the industry diversity in the top 100 is quite broad, including Logistics and Supply Chain, Telecommunications, Wine and Spirits, Construction, Transportation, Internet, Online Media, Think Tanks, Venture Capital, Utilities, Aviation and Aerospace, Computer and Network Security, Research, Translation and Localization, Photography, Mining and Metals, Real Estate, Security and Investigations, Public Policy, Accounting, Pharmaceuticals, and Non-Profit.
You might be surprised to know that it takes at least 36,000 1st degree connections to crack the top 10.
If you try to send invitations to connect with these folks, don’t be upset if they aren’t accepted – some of them simply can’t be.
Once you hit the 30,000 1st degree connection mark, LinkedIn won’t allow you to accept invitations to connect – so any connections added past that point must be invitations sent out to others to accept.
And so, without further ado, here is the list of the top 100 most connected people on LinkedIn: Read More
Granted, the figures listed for your second and third degree network are only estimates - here’s proof. More on that in a bit.
Unfortunately, if your network is quite large, the link might not actually work for you. Sadly, it doesn’t work for me – I get an unexpected error every time.
Just curious – would the error not ever be “unexpected?”
Of course, because LinkedIn no longer supports your network statistics, they can pull the plug on the link I’m sharing with you at any time.
Another Method for Determining Your Network Statistics: LinkedIn Search
If you’re like me and the LinkedIn Network Statistics link doesn’t work for you, there is another way to determine your 2nd degree network.
Simply enter a special, non-searchable character (I use an asterisk) in the first name field, select “Anywhere” as the location and hit search.
You can do this with a free LinkedIn account as well as with LinkedIn Recruiter, although with Recruiter, you will get evidence that you are effectively searching the entire LinkedIn network, as evidenced by the 200M+ results. LinkedIn Recruiter:
If you are searching with a free account, you will get a much smaller number – not sure why:
Regardless of the type of LinkedIn account you’re using, go down and look at the relationship facet, and you can see the number of your “2nd Connections” and even your Group connections, although sadly, these are not Group-ONLY connections.
Unlike the figures listed in your network statistics, which are only estimates (read further for proof), I believe the number of second degree connections that are shown in the search results above are actual numbers. Unfortunately, your 3rd degree connections are lumped in with “Everyone Else,” so we can’t use this method to divine the size of our 3rd degree networks.
LinkedIn Network Statistics Don’t Add Up
If you can’t use the LinkedIn Network Statistics link due to network size (or if LinkedIn kills the link), you can still use your first and second degree connection numbers from the LinkedIn search method shown above to roughly calculate your third degree network, or at least what LinkedIn might estimate to be the size of your 3rd degree network.
For example, LinkedIn claims my total network size is 30,995,402 professionals.
If I take 30,995,402 and subtract my 10,842,992 2nd degree connections as well as my ~30,000 1st degree connections, I get about 20,122,410.
Here’s the issue – that seems artificially low, doesn’t it?
That leads me to believe that the figure displayed as the number of professionals that LinkedIn claims my 1st degree network links me to (30,995,402+) is only an estimate, and I happen to think it’s much lower than the real number.
Even with significant network connection overlap, I don’t see how 10,800,000 2nd degree connections can yield a total network of only 31,000,000 people.
Am I missing something here?
Any math/statistics gurus out there care to weigh in?
I really would like to hear from anyone who has insight into what a more realistic 2nd degree to 3rd degree ratio would be.
Bonus LinkedIn Network Content
While some of you no doubt know about LinkedIn’s InMaps, I am quite sure many people don’t. As such, I thought I would share that you can get a very cool visualization of your LinkedIn network by going here: http://inmaps.linkedinlabs.com/
Unless, of course, you have a large LinkedIn network, in which case, you will only see this:
In this article I will detail two of Monster’s undocumented search operators, how to compress your Boolean search strings by more than 30%, and remind you of Monster’s documented but seldom used NEAR search operator.
AND = & + OR = |
Although I can’t seem to find any documentation of it, Monster’s search functionality does support the & for the Boolean AND search operator as well as | for OR Boolean search operator – which can save on character space for longer queries.
While most people don’t run searches that will test Monster’s main search field limit of 500 total characters (including spaces), there are those sourcers and recruiters who extensively leverage conceptual search, employing comprehensive OR statements for each concept in their Boolean search string, which can easily exceed 500 characters, especially when searching for a number of target companies.
In cases such as these, it can be helpful to use the ampersand (&) for AND and the pipe symbol (|) for OR, effectively cutting the number of characters used for AND’s and OR’s by 60% (5 total characters down to 2).
For example, compare these two searches which return the exact same results:
iOS AND (ObjectiveC OR “Objective-C”) AND (cocoa OR xcode) AND (iPhone* OR iPad*) AND (“apple store” OR iTunes OR “app store”) AND (SQL* OR xib)
Now we’re down to 101 characters, which is nearly 30% more efficient than our original 144 character search.
How’s that for Boolean search efficiency?
If you’re wondering how I figured this stuff out, it’s actually quite simple – curiosity and experimentation.
I challenge you to be curious and to experiment – from time to time, simply ask, “I wonder what would happen if…..?” and give something a try.
Hopefully all of what I’ve shared with you today has made you curious about your other sources and how you might be able to experiment and tweak your searches for other sites to make discoveries and yield additional benefits.
If you you do – please let me know!
Monster’s NEAR Operator: Documented but Seldom Used
Monster’s NEAR operator is an example of fixed proximity search, which can be used to return results with words, phrases or OR statements within 10 words of other words/phrases, or OR statements, which can enable semantic search at the sentence level.
Would you be interested in learning more about sentence level semantic search using the NEAR operator?
Additionally, I thought I might as well put all of my best work all in one blog post as well – over 110 of my articles in one place for easy referencing!
My blog is a pursuit of passion and not of profit – if you’ve ever found anything I’ve written helpful to you, all I ask is that you tweet this out, share it on LinkedIn, like it on Facebook, or give this a +1 on Google.
Many thanks for your readership and support – please pay it forward to someone who can benefit.
It was just the other day that I was hacking around on Google and Bing trying to find people on Twitter based on the text in their bio’s (yes, I am familiar with Follwerwonk – you’ll see why I prefer Google/Bing in a moment) and while I was getting some results, I wasn’t getting as many as I thought I should, nor were the results as “clean” as I would like.
That led me to a few minutes of tinkering with Bing and Google and I made a few discoveries with some simple pattern recognition that I would like to share that will help you quickly find your target talent pool on Twitter.
I use two main examples – mechanical engineers in South Africa and software engineers in Chicago – you can of course fork my Boolean strings to suit your specific sourcing needs replacing my titles and locations with yours.
How to X-Ray Search Twitter with Bing
While I do search for what people tweet about, I prefer to search for information contained in Twitter bio’s/profile summaries where people often identify themselves by what they do for a living (e.g., software engineering, accounting, etc.).
Furthermore, I prefer to search for bio data using Google and Bing, as there is no service/app I am aware of that indexes as many Twitter profiles as the 2 search engine titans. When I was using Bing to search for Twitter profiles the other day, I was looking for patterns in the results that were consistent across my desired results (actual Twitter profiles) and not my undesired results (Tweets and jobs/job posting-only accounts),
I noticed that Twitter profiles all mentioned “followers,” “tweets” and “following.”
As you can see below, only 1 job posting account was able to sneak in – the rest are profiles of people.
Simply overlooking the job spewing Twitter profiles is easy – I often advise people that an acceptable percentage of false positives is fine with any search. Trying to “over cleanse” results can have undesired consequences, such as eliminating valid results.
Always remember – every search you run both includes AND excludes qualified people/desired results. Think before you tweak!
So how many results would Follerwonk return in a Twitter bio search for mechanical engineers in South Africa? 51 vs 88 for Bing.
As you can see from just the first page of results, Google turns up more job posting accounts than Bing, which returned only 1 job posting account with the exact same search. Google only returned 4 real people in the results.
I find it interesting to see the differences between Google and Bing, especially when it comes to such a simple search!
I’ve been trying to tell people for years that Bing is a bit “cleaner” than Google with regard to searching sites like LinkedIn and Twitter. The results above offer further evidence to support my claim.
Anyhow, I looked at the results and noticed a pattern in the false positives (job spewing/non-people Twitter accounts) – most mentioned “status” or “statuses,” so I decided to exclude those terms from the URL’s.
As you can see, Twitter search services like Followerwonk do a good job of making it easy to search for and find people on Twitter, but they don’t index as many Twitter profiles as the major search engines such as Google or Bing.
As such, if you’re only using Followerwonk or similar sites to find people on Twitter, you’re only finding some people – and certainly not all of the people that are actually on Twitter.
Also, when it comes to any information retrieval exercise, a little bit of pattern recognition goes a long way.
Hopefully I’ve provided you with at least a couple of new ways to search Twitter via Google and Bing to find people with specific skills/titles in your target locations while reducing false positive results. Grab these bits of Boolean and add your location and title/skills:
Also, we’re lucky that this person took the time to explain what a “code sensei” is – if they didn’t make mention of “software engineer,” no one could find this person by searching for that title:
Imagine how many people describe themselves and their locations with non-standard terminology and you have a glimpse into the hidden talent pool waiting for you to explore on Twitter, Google Plus and other social networking sites.
However, if you’d like to know what a LinkedIn power user and sourcing/information retrieval geek thinks about LinkedIn’s new search functionality, you’ve come to the right place.
I’ve had access to LinkedIn’s new search interface and functionality for a week now, and I wanted to share with you my first impressions, discoveries, disappointments, concerns, and suggestions for LinkedIn.
LinkedIn Search: New, but Improved?
LinkedIn’s Smart Query Intent Algorithm
Before I had access to the new LinkedIn search, I was excited when I first read about the concept of a “smarter query intent algorithm.” LinkedIn claims that the more you search for content on LinkedIn, the more the query intent algorithm “learns and understands your intent over time to provide the most relevant results.”
Of course, I’ve only had access to LinkedIn’s new search for about a week now, so I can’t tell how “smart” is has become based on the queries I’ve been feeding it. However, the issue I have with any query intent algorithm that claims to be able to provide me with more relevant results is that only the user can determine if results are “relevant” or not.
According to Merriam Webster, relevance is defined as “the ability (as of an information retrieval system) to retrieve material that satisfies the needs of the user.”
As such, by definition, only the user can truly determine or judge relevance. A search engine cannot ever truly “know” the needs of the user.
While I appreciate and applaud the intent behind an “intelligent query algorithm,” which isn’t dissimilar to what many have been trying to do for years when it comes to search, the best way to implement such a system is to incorporate a feedback loop for the user to tell the algorithm which results the user truly finds relevant, rather than relying on supervised or unsupervised machine learning or some other method based on which profiles are clicked vs. which ones are not, and/or perhaps time spent reviewing specific profiles.
I’d love to know exactly how LinkedIn’s smarter query intent algorithm works (I’d love to make it smarter!), but something tells me that’s not something they would disclose.
I’m not a fan of black box search algorithms – I like to know exactly why I get the results I do.
LinkedIn’s Suggested Searches
I was also excited when I read about suggested searches, because my mind immediately raced to thoughts of LinkedIn being able to suggest better queries or perhaps searches other people had run for similar terms/people.
However, what LinkedIn is really referring to with regard to “suggested searches” is related to new unified search functionality in that if you type in a term or a title into the main search box on LinkedIn, you will see a list of options you can choose from, such as searching for related jobs, people, connections, groups, and skills.
I’m not saying this isn’t cool functionality, it’s just that I have high expectations when someone makes a claim of “suggested searches.”
Customized LinkedIn Results
According to LinkedIn Product Manager Johnathan Podemsky, ”No two professionals are alike on LinkedIn. This means even if you search for the same thing as someone else, your results will be customized to you,” ”LinkedIn’s search efforts are founded on the ability to take into account who you are, who you know, and what your network is doing to help you find what you’re looking for.”
This makes total sense based on the LinkedIn’s underlying fundamental concepts, but from a recruiting perspective – what if the best candidates aren’t within the network of the person conducting the search?
While Stephanie Mlot from PCMag claims LinkedIn’s changes put “…LinkedIn on a more level playing field with Facebook, which introduced Graph Search earlier this year as a way for users to sift through the network’s 1 trillion connections for more details about their friends,” I don’t agree. One major distinction is that a user can search for and find anyone using Graph Search – regardless of whether or not they are connected to them in any way.
Of course, LinkedIn does offer a solution for people who want the ability to search for anyone regardless of network connection – it’s called LinkedIn Recruiter.
However, if you’re searching LinkedIn for free, you’ll notice you no longer have the ability to sort all of the results of a search, which leads me to what’s missing from LinkedIn’s new search interface and functionality.
While LinkedIn Signal isn’t new – what IS new is that you no longer have to go to “News” on the top nav bar and click “Signal” – you can now simply click “Updates” on LinkedIn’s new search interface to instantly be taken to a Signal search for the keywords you’ve already entered.
Signal is one of LinkedIn’s most powerful and underutilized features. With the new and more prominent placement, I hope Signal will get the use and appreciation that it deserves.
What’s Missing from LinkedIn’s New Search
Curious to know what’s NOT included in LinkedIn’s new search interface and functionality?
A number of things.
First and foremost, you can no longer sort your search results.
I always searched by keyword relevance when searching LinkedIn, because even with a large network, I am not so ignorant as to believe that the best people for any given position I may be sourcing and recruiting for are always going to be within my 1st or 2nd degree network, let alone my 3rd degree connections or within my LinkedIn network at all. If the best match to a search happens to be in my 3rd degree network, I’d like to see them come up on page 1 of the results.
Say goodbye to this if you’re using a free LinkedIn account:
LinkedIn’s sort by “relevance” option was a mix of network connection and keyword relevance. Based on my searches using LinkedIn’s new search interface, it seems that search results are sorted based on some combination of keyword relevance and relationship, as 1st and 2nd degree connections are returned early in search results and 3rd degree and group only search results come much later in ranking.
While you can still search specific layers of your LinkedIn network, there is no way to search for Group-only connections that are not also connected to you in the 1st, 2nd, or 3rd degree.
With a free account, the only way you can try and achieve anything close to searching solely by keyword relevance is an X-Ray search. Thankfully, you can still sort your results by keyword relevance within LinkedIn Recruiter.
The Ability to Run SUPER LONG Boolean Search Strings
I am sad to report that LinkedIn’s once-epic ability to run Boolean search strings of over 3,000 characters has come to an end.
That means you can no longer perform some of the interesting diversity sourcing searches I’ve detailed in the past, such as searching for all of the HBCU’s in a single search, or searching for the 354 most common female names in the U.S. over the past 4 decades to find 65% of all of the women on LinkedIn in a single search.
From my preliminary testing, it seems that you can get away with searches up to around 1,300 characters with spaces before you start to encounter LinkedIn just spinning and never executing your search. With a first name search, this is what 1,281 characters with spaces looks like.
Top 10 Facets
Also missing from the LinkedIn’s new search interface is the ability to see the top 10 results in each facet.
I can’t be the only person who found the ability to see the top 10 companies employing certain types of people in a given market, the top 10 markets for specific skills, or the top universities by skill to be valuable, can I?
Now free users are limited to the top 5.
Thankfully, you can still view the top 10 results in each facet in LinkedIn Recruiter.
Advanced Search Operators
Alas, Voltron has laid LinkedIn’s Advanced Search Operators to rest.
LinkedIn has claimed that extending new search functionality to their mobile apps is something that they’re looking into, but for now, the mobile apps only allow users to search people but not within other categories.
Mobile search is a big deal for LinkedIn – did you now that 19 people searches are performed and 41 profiles are viewed every second via LinkedIn mobile apps?
What About 3rd Degree Connections?
While there was a bit of early buzz that users searching LinkedIn with a free account would not be able to search 3rd degree connections, you can in fact still search for them.
While some early testing showed that it appears LinkedIn’s default was to only return results from your 1st and 2nd degree network, all of my recent searches appear to default to “All,” which includes Group Members and “3rd + Everyone Else.”
Thankfully, I haven’t run across too many search anomalies yet, but I did find a few I think you (and the LinkedIn dev team) will find interesting.
I ran a basic search and took notice of the top 5 companies represented:
I then set about to see if I could use the -/NOT functionality to eliminate results from the top 5 companies in order to find the next top 5 (thus completing the top 10).
I started entering 1 company at a time in the current company field: -Microsoft, -IBM, -Cisco, etc.
This seemed to work quite well in removing those companies from the top 5, allowing me to explore the next 5 or more. But then I noticed that when I was excluding the company names in the current company field, the company names were being returned as positive hits and highlighted as keywords in the profiles. The same thing happens if I change it to -(ibm OR microsoft OR cisco).
Hmm. That’s not good.
The same thing happens when I try to exclude a term from the title field. As you can see below, I am excluding the term “engineer” from the title field, and while the term is excluded properly in most cases, there are a few random results where “engineer” is in the current title – as with Kevin below, the word “engineer” also shows up as a positive & highlighted keyword hit in summaries, headline phrases, etc. It doesn’t matter if I use NOT, AND NOT either – I’ve tried all 3 ways and get the same results.
This one is pretty strange – I ran a first name search for “Abigail” and got results with “Gail” and “Abby” on the first page.
I don’t know how much of a fluke this is, because I’ve tried other names as well as searched for companies and various I.T. keywords to see if LinkedIn is performing some kind of fuzzy matching but have yet to run into another instance where LinkedIn gives me terms other than the one I specifically searched for. Please let me know if you find any.
Also, it seems that the ability to search within groups from the main search interface is still being listed as a premium filter with the yellow “in” icon, yet I can search within groups with my free account. Maybe it’s actually free functionality now?
What I Would Like to See from LinkedIn Search
For quite some time I’ve been thinking about writing a post specifically about what I’d like to see from LinkedIn with regard to new search functionality, but I’ve never gotten around to it.
I’ll take this opportunity to at least highlight a few things I would suggest to the LinkedIn team:
The ability to specifically search within the most recent work experience listed. One word: Massive. Can I get an “amen?”
Stemming/root word/wildcard search. It would certainly be nice to not always have to construct massive OR statements, e.g., (develop OR developing OR develops OR development OR developed OR developer).
Not only bring back the top 10 in each facet – but enable them to be expanded to the top 25. Expanded facets yield incredible market and competitive intel/insight with the click of a mouse.
Ability to sort by keyword relevance not tied to relationship. If you can’t/won’t bring this back to LinkedIn for free accounts, at the very least, never get rid of the ability to sort by keyword only relevance in premium versions.
Keyword boosting – enabling users with the ability to determine which keywords are the most relevant to them.
Proximity search – enabling users to search for terms within a specific distance of each other, to achieve semantic search.