More on OR: the Google Boolean Dilemma

booleanstrings Boolean


Anthropology and Psychology study human behavior. Ethology is the study of animal behavior. As a Sourcer, I am finding myself more and more studying software behavior. (Is there a term for that?)

Studying a software application may sound odd. Isn’t writing software also called “programming” which means that its behavior is fully predictable?

There are two factors related to studying software behavior:

  1. We usually don’t have access to the code, and documentation covers its behavior only partially.
  2. If a user is interacting with the system, the user’s input triggers various code behaviors. Trying various inputs, we can derive, with some confidence, what a particular application does.

Sure enough, there’s code complexity, bugs, and other factors interfering with our study, but the knowledge we gain is worth the effort. It provides invaluable practical tips on using software.

This post is a result of studying Google search behavior. Here is a hypothesis on how Google search responds when we use OR statements.

It’s well known that Google has semantic search capabilities. As part of that, for every keyword in the search string, Google would also search for variations of the word (called stemming: manager/management) and synonyms (developer/programmer).

However, as our study shows, for keywords in an OR expression, Google stops looking for variations or synonyms. It searches for the exact expression as if we put the word in the quotation marks (meaning “no variations”).

Take a look:

Do you see what is happening? When a term is part of an OR statement, it is used exactly, with no variations or synonyms (search on the left). With no OR used, we see variations in the results (search on the right).

When I source, I prefer to let Google bring suggestions. I almost never user OR statements, leaving Google’s semantic power at work. For somebody who can sit down and list every possible synonym and variation you would want to see, the OR statement is for you.


Do not use OR statements for synonyms and similar terms; Google will bring in additional relevant results, and your job would be just to collect them.

If you have in mind variations of a term to search for, that you want to make sure are included, you can still search several times using each of the terms separately.

If you feel you need to be in tight control of the terms to use, or if you have a longer list of terms that are not synonyms (for example, names of target companies), use an OR statement.

We will soon be offering a fully updated webinar Boolean Basics; if you are interested, keep an eye on our schedule >>>

Search by Ideal Candidate? #LIR

booleanstrings Boolean

An “ideal candidate” is a notion of importance when sourcing talent. When we source, we always ask for one, or better, several profiles of “ideal” candidates. Even with a well-written job description and clearly specified parameters, such as location, target job titles, years of experience, salary range, must-haves and nice-to-haves – there can be some unverified assumptions as to whom the hiring manager and her team want to join them. We know from studies and experience that an “ideal” profile or resume is always of help, narrowing down what we are looking for. An “ideal profile” facilitates our communications with the employer, our client who is hiring.

An ideal profile can serve as the starting point for sourcing. Sourcers would study the profile and get a better understanding which keywords, title, and companies to look for. When we source, we’ve liked to look up LinkedIn member profiles “similar” to the one that is given. Too bad we no longer see the “similar” search in personal accounts. Whatever the algorithm for similarity was, it did provide some results worth reviewing, always around 100 profiles or so – that we knew how to narrow down to a location. LinkedIn Recruiter still shows “similar” profiles on request, and it’s a useful feature.

LinkedIn Recruiter also offers us to point to one or more “ideal” profiles and give us “matching” results. But this newest “search by an ideal candidate” available in LinkedIn Recruiter, while being heavily advertised, is not something that I am excited to use for sourcing. All it does, apparently, is extract several job titles, skills, and companies from the given profile(s),  put in the search dialog, and search for a Boolean OR of the terms. That may sound good on paper, but it doesn’t work well in real life. Old job titles go into the search along with new; important skills are mixed with unimportant. The number of results varies, to say the least! Here are some number for you – see for yourself:

  1. David Galley has only 4 (four) people “like” him (I knew David is unique!)
  2. Martin Lee is even more unique – three profiles (I knew that too!)
  3. So is Julia Tverskaya – three profiles (yes!)
  4. Me – less unique – 52 profiles
  5. Glen Cathey – 511 profiles like him
  6. 1,070 profiles like Suzy Tonini
  7. 10,868 profiles like Jim Stroud 
  8. 170,944 profiles like Phil Tusing
  9. 1,469,952 profiles like Balazs Paroczay (are you kidding?)
  10. David, Martin, Julia, and I together produce 13 (thirteen) profiles that are similar to the four of us. (What does that mean?)

Do these numbers (and suggested profiles) make sense? Sorry, LinkedIn, not really.

The question is, whether searching “by ideal candidates is a feature helpful to Sourcers in any particular circumstances. If LinkedIn simply extracts titles, skills, and keywords from the “ideal” profiles, without any extra intelligence – or machine learning – to improve how well it “understands” what we are looking for – this feature would not help those who search.

Which features are helpful, and which, just misleading? Join me for the webinar about LIR (LinkedIn Recruiter) and its variations such as Lite and RPS, this Wednesday, April 12th, to find out!





How LinkedIn “Loses” Your Potential Candidates and Leads

booleanstrings Boolean


Searching on LinkedIn? As illustrated by some puzzling search examples in the posts LinkedIn-Based IQ Test and Advanced LinkedIn-Based IQ Test, LinkedIn search has been missing some results that apparently do exist in its database. Thanks to all for the comments and investigating. Special thanks to Rachel Evans and Adam Kovacs who performed an in-depth exploration of the unfortunate “phenomena” and provided some insights!

Some searches I ran across in my own investigation lost as much as 98% of the results that should be there – but were not.

Here is what has been going on: for each search, LinkedIn “personal” search would seemingly randomly interpreting the keywords we type in the keyword box as people’s first or last names (for example, it interpreted this way the word Morgan in one of the examples), or company names (it switched to interpreting as a company name when we entered Morgan Stanley) – or as job titles. Based on my previous knowledge and tests, LinkedIn wasn’t trying to cut results so that we would pay for a more expensive account. To the best of my knowledge this was a massive bug, some random code, perhaps “left over” from the good old times of the Economic Graph ambition. It lasted a while.

Did the Software Engineers at LinkedIn read my past two blog posts?

Two weeks since the posts, it looks like LinkedIn is rolling out a search update addressing the issue. Some members still get partial results. Some other members are no longer seeing the gap in results as of this morning – so perhaps this outdated half-written code that was interpreting was finally removed. (Are there new bugs and inconsistencies? Probably; we’ll be investigating.)

Are you wondering whether your account still produces partial results due to the “semantic” bug? Here is how to find out. For this search,

Java enterprise,

if your account got the rolled-out fix, you will see 312 results; if not, you will still see 6 results (as it was for all at the time of my last post).

Generally, as we have seen LinkedIn search results being “missed” in the past, that mostly was due to the code and database design, specifically due to various unsuccessful attempts to semantically interpret what users want to search for.

It’s never boring, is it?

Whether you are using LinkedIn Recruiter (including Lite) or a personal account, or want to search outside of LinkedIn, we have a webinar for you to attend, to get helpful insights on what to expect from LinkedIn and outside of it, and how to search for prospects in the most productive fashion. Upcoming sessions include

Advanced LinkedIn-Based IQ Test

booleanstrings Boolean

In the previous post, LinkedIn-Based IQ Test, together with some folks who have posted insightful comments, we have observed that LinkedIn search attempts to interpret search terms. It recognizes the names of people and some (large?) companies if we enter them in the search box  – and automatically alters the searches it performs based on that. For example, search for Morgan, get people with the first or last name – Morgan. Search for Morgan Stanley, LinkedIn decides it’s a company name. (We didn’t find this behavior helpful or intuitive).

As much as we would want LinkedIn to search for skills (vs. “just” keywords), it does not do that.

Here is your next set of searches – can you explain these? The results would be the same for a basic, job seeker, or business premium accounts:

If you have one of the “professional” subscriptions – Recruiter, Lite, or Sales Navigator – you might want to look what the same searches produce there as well. You have guessed, the results will be different than in the “personal” search.

Josef Kaldec (whose hacks we loved watching at Sourcing Summit Europe) points to some interesting, related, behavior of the business network in his presentation.

David Galley and I will shed some more light on this mysterious LinkedIn search behavior in a repeat of the twice-sold-out webinar “Overcoming LinkedIn’s Limitations”– April 19, 2017.

I am also happy to report, that our new tool Social List finds many more results for some of those searches – and it costs less than Recruiter, Lite, or even Sales Navigator.

Now – who has some guesses, why do the above searches work as they do?


Social List Launch!

booleanstrings Boolean

Social List is Live

I am happy to let you all know – our new sourcing tool is up and running! We are getting our first subscribers, fast!

If you have not heard, Social List is an easy-to-use tool to instantly build lists of target social profiles, matching your search parameters, such as locations, employers, and job titles. We have developed the Social List Agents from our accumulated “know-how” of Internet search behavior, tweaking multiple search parameters “behind the code” to provide high-precision results.

I would like to invite you to see a demo of Social List 1.0, and ask any questions you might have, at one of the two webinar-based sessions I will hold:

  1. Social List Launch – Fri, Mar 24, 2017 9:00 AM – 10:00 AM PDT
  2. Social List Launch (2nd session) Mon, Mar 27, 2017 10:00 AM – 11:00 AM PDT

You can also view a video Demo and send us questions via the Contact page. And here is a detailed description of the tool – The Social List Difference.

If you are currently a Beta tester – thank you for helping us to shape up the tool! Your free access will be ending EOB Monday, March 24.

All new subscribers (at the introductory price of $49/month) will be receiving a two-day free trial, during which you can cancel and you won’t be charged, or you can continue the monthly subscription.

We will only be allowing a limited number of users in this first release. If you want your colleagues to be part of the launch, please tell them to hurry and sign up!


The Difference

Here is what our Agents can find, what you won’t find anywhere else:

User Feedback

Over the past month and a half, a group of Beta-testers has tried the tool – and everyone loved it. Here are just a few of our first user’s comments.

“Just found a fantastic candidate using your tool that I hadn’t come across with LI Recruiter. He literally had everything I was looking for and LI hadn’t returned him for me once.” – Katharine Robinson

 “This would be the first tool I would start my search with. I am very happy with the results – hardly any noise! – very focused and targeted answers. No whistles, no bells, just a straightforward, excellent search tool. A (very!) good job! ” – Karen Azulai

 “Fast, reliable, and simple list creation. Maximum results with minimal effort. Fun to find profiles I would have otherwise missed. I like that in addition to GitHub and LinkedIn (often the lowest hanging fruit), Social List offers custom search lists for Meetup, Zoominfo and Google Plus. — Sourcers at all levels will want to hop right in” – Maisha Cannon

What’s Next

We already have tons of ideas on scaling and expanding Social List, its integration with other services, building custom Agents for clients, and more.

I look forward to sharing the adventures of using Social List with you!

Thank you,

Irina Shamaeva

LinkedIn-Based IQ Test

booleanstrings Boolean

Whether Boolean, Semantic, or Machine Learning win the global search quality competition, remains to be seen. But a search system quality, by definition, depends on how well users can get results they want. For that, a user’s understanding of what to expect in the results is important (I hope you agree with the last statement).

Many modern systems have advanced search syntax – at a minimum, supporting the Boolean logic and the quotation marks for phrases. Systems also do some interpretation of the user’s intent when searching. For example, we can expect Google to: a) see if there are misspellings and offer corrections b) include synonyms for keywords without the quotation marks and for abbreviations; c) try to find pages where keywords are close to each other, in the same order, and rank those pages higher.

On ZoomInfo, we can expect that searches for VP and Vice President will return the same results.

On LinkedIn, however, there has been little interpretation of searches. It has had some title abbreviation recognition (VP = Vice President) on and off, different in Recruiter and personal, and it’s rather unclear where that stands. We’ve also long noticed that LinkedIn is interpreting people’s names – searches for Bob and Robert return similar (though not the same) results.

In any search, we expect that:

  1. If we add a keyword, the number of results goes down
  2. If we add a condition, the number of results goes down
  3. If we didn’t use the quotation marks, the word order should not matter.

So now, let me present you with an “IQ test”, based on the following searches, which produce unexpected results.

Question for you: what is LinkedIn “thinking” (i.e. what is the internal logic, where does any interpretation come in) when it produces these results? (And yes, it’s not what a user would expect.)

Email me the answers (or hypothesis) of why it works like that, or post in the comments. The first few correct responses will get a ticket at the upcoming

LinkedIn webinar

(sold out for this week but we have scheduled a new session).

Recruiting Recruiters – Sourcing Techniques

booleanstrings Boolean

A guest post by Martin Lee

Ironic isn’t it that some of the hardest people to recruit are recruiters?

Arguably they are the easiest people of all to find online by the very nature of the business. Active in advertising, networking, promoting, attending events and being “out there.” Is there a recruiter NOT on LinkedIn? ( I would love to know).

R2R / rec2rec (I think this is mostly a British term) and people responsible for internal hires place great value on their networks. Built up over a period of time from referrals and knowing who’s who in the market from experience. It’s old school, and I like it, but it shouldn’t be just limited to this. One person’s network will differ from another which limits the potential addressable market.

So for someone who recruits recruiters (agency or internal), the challenge lies in sifting through the gazillion profiles to find those genuinely good.

Let’s define “good” into some common requirements that are often looked for and then let’s apply a sourcing head into how we can search for them.

“Top billers” traditionally is a term for agency recruiters. Those that make the most profit for a company and naturally of interest.

Like all searches I start off as simply as possible, just try adding some keywords and phrases into platforms where the target people go, in this case, LinkedIn.


Lots of results from individuals and teams and as we are searching posts, not profiles – these are the latest results, people currently performing well.

No prizes for guessing who some of the 19 likes are from.

Of course, we could also run specific searches with a variety of terms: “over target” “promoted to” “exceeded quota” “won tender for” and many other. It would be worth creating a long list of these and saving as an alert if that’s what you’re interested in.

One of the challenges with LinkedIn is that you must think of all possible connotations of these terms. For example a search for “achieved 100% of target” would find different people to “hit my target” and what about all those people who were over 100% of target – do we really want to write “achieved 100% of target” OR “achieved 110% of target” OR “achieved 150% of target”, etc. And then there are the people who don’t mention it all; they just get on with it.

Cue Google.

The asterisk * “fills in the blanks” and finds any word(s). The two dots .. show numbers above the number specified i.e. 100.. will show all numbers above 100.

So we can run a search like this.

Additionally, we may be interested in people changing job title (often a promotion or further specialization) or returning to a company. We are operating under the assumption that if invited back they must have been successful before, and it’s a good assumption. Using the “new” LinkedIn:

Recruiters post jobs. Across social media, job boards and with their ATS’s – therefore isn’t that also a good place to look at their specific activity as opposed to what they say about themselves on a LinkedIn profile?

There are thousands of these platforms, but common ones include Taleo, Aplitrak (part of Bullhorn), Smart Recruiters and Indeed offers an advanced search option, which also allows you to remove employment agency jobs. There are lots to be found in here and get understanding who does what, where and to what level.

Or a search looking for recruiters at specific companies in a particular field?

Now all of the above is fine but here’s the thing – If the role you have is not a good match then no amount of clever sourcing on its own will do and you have to do your research first. Salary, bonus, benefits, responsibilities, travel, progression, culture & location are all factors recruiters will consider.

I’ll be digging into this important topic much more deeply at my upcoming

Webinar: Recruiting Recruiters 

Date: Tuesday, March 28th
Time: 1PM British Standard Time (Check your local time for the Webinar)

LinkedIn Tip Sheet Error And a New Hack

booleanstrings Boolean


We shouldn’t be mixing the level of service, quality of software, and changes in features and pricing as a reason to use or not use LinkedIn. OK, LinkedIn has taken us along for a ride, adding and removing features the platform used to have years ago and making it “news”. BUT – it would be silly and unproductive of us to walk away from the incredible pool of self-entered professional data. It’s just not an option. Besides, we may be getting an adequate search system back, after all!

Apparently, it is hard not just for us, uninformed users, but for LinkedIn itself to ride through so many redesigns and individual roll-outs within such a short period. Let’s take a look at the screenshot above, which is THE guide for advanced searching at the moment. This syntax, as in the example, doesn’t work; please take a note of it.

There needs to be one correction in LinkedIn’s search tip sheet. This syntax –

title:(CMO OR “chief marketing”) (wrong!)

– doesn’t do what you think it should do, i.e. look for one job title or another. Compare with searching for

a:(CMO OR “chief marketing”)

or just

CMO OR “chief marketing”

– and you will see the same results. That means that the operator in front of an OR statement does. The first search above does just a keyword search, not a title search. To search for one title or another, we need to write it differently than LinkedIn tells us –

title:CMO OR title:”chief marketing” (correct, at the moment).

Our friends from Social Talent have also noticed the discrepancy and reflected in a recent post. Unfortunately, the wrong tip sheet gets propagated by bloggers copying it:

It would be nice if LinkedIn either implements the syntax it documents or fixes the tip sheet to reflect what it does. Agree?

And here’s a new LinkedIn HACK for you.

Step 1. Let’s take the original string (like in the tip sheet) title:(CMO OR “chief marketing”). Now, replace the : for =

title=(CMO OR “chief marketing”)

Step 2. Run a search for people – for example, this.

Step 3. Append what we got in Step 1 to the end of your search URL, after a &. And here is a search for the current job title:

Given that there are multiple UI versions out there and things are still in flux, this may or may not work for you- sounds like it does, for most!

Tip. If you examine how we have changed the search URL to perform the job title search, you may come up with other “hacks” that will expand the search functionality.

We will tell you how to work with and work around the newly redesigned LinkedIn in the upcoming webinar – Wednesday, March 15, optional practice Thursday March 16, 2017.

Don’t miss it!


Googling for Recently Updated Profiles?

booleanstrings Boolean

Can we find web pages, including social profiles, that have recently been updated? (We all know that updates may mean warming up for a new employment opportunity).

On Google, you can set a date range for the search results, then sort by date. However, I have long noticed that, even if we set a date range from the beginning of times until now, we will be losing some search results. Apparently, for some pages, Googlebot cannot define the last updated date, and those pages would be missing from the results if sorted by date.

Greg Lindahl, Founder and CTO of (now defunct) search engine Blekko, acquired by IBM Watson in 2015, provides an excellent overview of web page dates’ challenge for search engines on Quora:

“There are two huge problems for date sorting of search results.

The first is that date sort — and I really mean date sort, not relevance sort of a date range like “past hour” or “june 3” — means you only get 1 bit of relevance, where something is included or excluded. This means you may get a lot of spam…

The second is what date should be assigned to a webpage. The first date it was crawled? The date on the page? If a page changes slightly, does it get a new date? If a website puts the current date and time on every page, what do you do?”

If we are X-raying a site for profiles, the “spam” issue is not as significant, since all the results would be profiles. But the second problem Greg states is there – depending on the site. A pages’ last updated date depends on “how well” the site “tells” Googlebot about the date.

Here is a TIP: to find out whether Google has the last updated date for a page, and what it is, X-ray for the page, while setting a wide date range.

It turns out, for the majority of LinkedIn profiles the date is a question mark and, therefore, the date is absent in Google’s index. Compare, for example, these two X-Ray searches:

Some LinkedIn public profiles do have a date, but that’s pretty inconsistent. (Of course, as an additional factor in identifying recent changes is the frequency of Googlebot’s visits to various profiles. I’ll talk about that in another post).

Bottom line, it’s impossible to Google for the most recently updated LinkedIn profiles.

X-Raying other websites? Take a look whether those sites are “better disciplined” in providing the dates. Please share what you discover!

Seven Fun X-Ray Strings for Tech Talent

booleanstrings Boolean

Those of us who search for Technical Talent, know how competitive the industry has been. Let me offer some X-Ray searches to help to discover potential IT candidates – and to help to grow your X-Ray skills. As always, pay attention to the URL structure and common phrases that are unique for the target pages, and you’ll find what you are looking for! I have included Google searches that that you probably have not seen on other blogs; take a few minutes to try and figure out why they produce the results they do.

Here you go:

  1. machine learning – Developer stories, that represent their professional history
  2. London Devops “last seen today OR yesterday” – recently active users (hoping that Googlebot has caught up fast)
  3. “member since” geeks 94105 – members of Meetup gatherings happening at the zip code 94105
  4. “10.. results for repositories written in Python” inurl:repositories – users who have written a lot in Python (though you may argue that this is not an X-Ray string J)
  5. inurl:repositories javascript golang Berlin – profiles of people who have written code in both JavaScript and Go
  6.*/resume developer “mountain view” javascript – built-in Github resumes
  7. resume angularjs Chicago – resumes on Github “pages”

To learn more about sourcing for Techies, consider registering for the webinar How To Find and Attract Technical Talentattendees get a month of sourcing support. We got rave reviews and will be repeating the webinar later this month. (Let us know if you can’t wait and want to get the recording from a few days ago instead).

For 300+ more Boolean Strings check out the second edition of the Boolean Book, fully reworked for the new year. Over 460 470 480 of your colleagues have obtained the e-book by now.