XRay string screening for past company and current job title- why doesn't this work better?

[Wahidur Rahman]

Just as background- I recruit for a lot of accounting roles, so of course sourcing candidates with a Big 4 background is always very helpful. In this case, I'm looking for someone like that with 'audit' in their job title now. This string works, sort of....

site:linkedin.com "current * * * audit" "past * * * deloitte" "location * greater boston area" "people you know"

I get 51 results. (Past company doesn't necessarily need to be Deloitte- I just wanted to get this up and running, then I can add other company names). My issue is that, when I do a similar Advanced Search in LinkedIn, I receive 144 results. (Note: because 'Greater Boston area' is not a term to my knowledge in LinkedIn, I do a proximity search of 35 miles from 02110- Boston. Not sure if that makes a difference here). I have a network of about 2,000 mostly accounting and finance professionals, mostly in and around Boston. If I search for a different current job title, it gets even worse. This string-

site:linkedin.com "current * * * internal audit" "past * * * deloitte" "location * greater boston area" "people you know"

-gives me literally 1 result. When do a LinkedIn Advanced search, I get 24 results. Why the difference?

My question: If I receive 144 results from my network, I can only assume I would find even more outside my network by XRaying Linkedin. So why does a similar search return about two-thirds less profiles?? I've played with this string a dozen different ways and can't seem to get better results. What's the difference between the two searches?

[Dave Galley]

Wahidur,

The problem is in how Google is parsing the "past * * * deloitte" part of the string. For some reason it is not matching this to profiles that it seems like it should. The only profiles matching are those where deloitte is within 2 'words' of past, and this is only one profile. If you were to make it more like 5-8 *s, you would get as many as 10 results, then it starts falling off again (you get different results, so eventually you will hit most of them, but not in one search).

I would recommend not relying on "past * * " to find past employment.

LI internal search is better for the specific case of "current this, past that" type searches of LI profiles.

The Greater Boston area may also limit a little bit - there are some zips within 35 miles of Boston that show up as other 'areas' in the LI place label system (parts of RI and NH fall inside the range and get dumped under the nearest in-state metro, mostly). But this isn't the main problem.

[Wahidur Rahman]

Thanks Dave, I really appreciate your tackling this. Sorry to doublepost it in the LinkedIn group, I was just hoping for a magical solution....

Kind of frustrating that either Google or LinkedIn is not 100% cooperative with the XRay search here. I'll play around with Bing a bit this weekend and see if I can get different results. The multiple ***** doesn't seem to ever work for me, 3 seems to be my magic number and I get diminishing returns after that. I just can't understand what the problem is with the 'past' command in XRay- LinkedIn just doesn't want to play ball?

Anyways, thanks again for helping out!

Dave Galley said:

Wahidur,

The problem is in how Google is parsing the "past * * * deloitte" part of the string. For some reason it is not matching this to profiles that it seems like it should. The only profiles matching are those where deloitte is within 2 'words' of past, and this is only one profile. If you were to make it more like 5-8 *s, you would get as many as 10 results, then it starts falling off again (you get different results, so eventually you will hit most of them, but not in one search).

I would recommend not relying on "past * * " to find past employment.

LI internal search is better for the specific case of "current this, past that" type searches of LI profiles.

The Greater Boston area may also limit a little bit - there are some zips within 35 miles of Boston that show up as other 'areas' in the LI place label system (parts of RI and NH fall inside the range and get dumped under the nearest in-state metro, mostly). But this isn't the main problem.

[Dave Galley]

The way Google syntax works, you can read each * as "any number of words", so * * * is that, three times, which should be a minimum of three words, but it doesn't always work like that in practice.

Google also does odd things with text that is split over multiple HTML elements in the same page, when it comes to indexing and search, and the behavior is not always consistent (sometimes * * is a good way to 'jump' over a weird HTML based gap and then any number of words, and it's also why we need a * in the "location * greater" construction for LI profiles). Sorry if that's getting a bit technical, though. :)

Wahidur Rahman said:

Thanks Dave, I really appreciate your tackling this. Sorry to doublepost it in the LinkedIn group, I was just hoping for a magical solution....

Kind of frustrating that either Google or LinkedIn is not 100% cooperative with the XRay search here. I'll play around with Bing a bit this weekend and see if I can get different results. The multiple ***** doesn't seem to ever work for me, 3 seems to be my magic number and I get diminishing returns after that. I just can't understand what the problem is with the 'past' command in XRay- LinkedIn just doesn't want to play ball?

Anyways, thanks again for helping out!

Dave Galley said:

Wahidur,

The problem is in how Google is parsing the "past * * * deloitte" part of the string. For some reason it is not matching this to profiles that it seems like it should. The only profiles matching are those where deloitte is within 2 'words' of past, and this is only one profile. If you were to make it more like 5-8 *s, you would get as many as 10 results, then it starts falling off again (you get different results, so eventually you will hit most of them, but not in one search).

I would recommend not relying on "past * * " to find past employment.

LI internal search is better for the specific case of "current this, past that" type searches of LI profiles.

The Greater Boston area may also limit a little bit - there are some zips within 35 miles of Boston that show up as other 'areas' in the LI place label system (parts of RI and NH fall inside the range and get dumped under the nearest in-state metro, mostly). But this isn't the main problem.

[Wahidur Rahman]

Thanks again Dave.

I'm fairly new with Boolean sourcing, but I feel that XRaying LinkedIn is fairly overrated. For example, for the same search as above, I ran this string with no past company info:

site:linkedin.com "current * * * internal audit" "location * greater boston area" "people you know"

It returns 208 profiles. Yet a LinkedIn Advanced search with 'current title Internal Audit', within 35 miles of 02110 (downtown Boston), returns 328 profiles. I understand the stuff with the "Greater Boston area" and the mileage- but I'd still imagine an XRay search would return more profiles than my internal search, not over a third less.... My personal LI network is only 2000 people, it's not like it's gigantic.

My conclusion from this is simply to use LinkedIn internal 90%+ of the time- that there are just too many issues with indexing to really get more results out of XRaying. Just my thoughts

Dave Galley said:

The way Google syntax works, you can read each * as "any number of words", so * * * is that, three times, which should be a minimum of three words, but it doesn't always work like that in practice.

Google also does odd things with text that is split over multiple HTML elements in the same page, when it comes to indexing and search, and the behavior is not always consistent (sometimes * * is a good way to 'jump' over a weird HTML based gap and then any number of words, and it's also why we need a * in the "location * greater" construction for LI profiles). Sorry if that's getting a bit technical, though. :)

Wahidur Rahman said:

Thanks Dave, I really appreciate your tackling this. Sorry to doublepost it in the LinkedIn group, I was just hoping for a magical solution....

Kind of frustrating that either Google or LinkedIn is not 100% cooperative with the XRay search here. I'll play around with Bing a bit this weekend and see if I can get different results. The multiple ***** doesn't seem to ever work for me, 3 seems to be my magic number and I get diminishing returns after that. I just can't understand what the problem is with the 'past' command in XRay- LinkedIn just doesn't want to play ball?

Anyways, thanks again for helping out!

Dave Galley said:

Wahidur,

The problem is in how Google is parsing the "past * * * deloitte" part of the string. For some reason it is not matching this to profiles that it seems like it should. The only profiles matching are those where deloitte is within 2 'words' of past, and this is only one profile. If you were to make it more like 5-8 *s, you would get as many as 10 results, then it starts falling off again (you get different results, so eventually you will hit most of them, but not in one search).

I would recommend not relying on "past * * " to find past employment.

LI internal search is better for the specific case of "current this, past that" type searches of LI profiles.

The Greater Boston area may also limit a little bit - there are some zips within 35 miles of Boston that show up as other 'areas' in the LI place label system (parts of RI and NH fall inside the range and get dumped under the nearest in-state metro, mostly). But this isn't the main problem.

[Dave Galley]

This is an issue of the right tool for the right job.

As far as Google is concerned, all the pages it has indexed from LinkedIn are just big chunks of text for you to search over.

LinkedIn, however, 'knows' things about each profile - there is some structure to the data, and it can use that to allow you to search in specific parts of the profile (job title, location, company, etc).

But LinkedIn does not let you search for certain patterns of words like Google can. Try finding people who claim to speak French with native or bilingual proficiency using LI internal search, for example.

So if you mainly want to search using the structured fields, then by all means, LI will do this better. But if you are looking for things in the general body of the profile, more keyword-only search, then X-ray can do let you do some things that internal search can't.

[Wahidur Rahman]

Interesting. Thanks again Dave!

[Randy Bailey]

Hello Dave,

I am getting back into sourcing, so have been dealing with some of this.  When in LI, the key words I use, doesn't search the entire text?  What is it searching when I use a keyword search as part of my search?  

I haven't seen the point yet of doing X-raying, as LI has generally worked pretty well.  I have a fairly large network, but even that doesn't matter so much any more, if you are using LI Recruiter.  I do wonder what am I not seeing, who isn't showing up.  Obviously, if they haven't included some information that I am searching on, they aren't going to show up, but am I seeing everyone that has the specific key words?

Thanks for the informative responses!

Randy