Still Using Keyword Searches to Find Applicants?

A very interesting thing happened the other day when a friend of mine, a director of human resources at a large national healthcare organization, stopped by in a state of panic. What you are about to read may be very disturbing to you, too – especially if you’re still searching resumes by entering keywords. She was very upset about her current online applicant tracking system and was certain that it was broken. She started by entering a series of keywords and clicking “Search.” Her results returned hundreds of applicants that met the keyword criteria. She pointed out five key applicants and told me to watch what happens when she entered a different series of keywords and clicked “Search” again. The same five applicants appeared at the top of the new search. She clicked on one of the top applicant’s resumes and didn’t see anything unusual in the body of the text or an unusual number of keywords. In fact, we didn’t see many of the keywords that were entered for the search. Hmmm… She opened up another applicant’s resume that “always” seemed to pop up regardless of the keywords entered and didn’t notice anything unusual here either. Perhaps her system was indeed broken. I asked her to highlight the applicant’s entire resume and paste it into a Word document. I then asked her to highlight the text and change “all” of the font color to black. An amazing thing happened. There between the paragraphs were all of the keywords that were not visible (or printable) until they were highlighted and the font color changed from white to black. The mystery was solved. It appears that applicants have become really creative in their attempt to find a job. By entering numerous keywords in their resume, then highlighting the keywords and changing the font color from black to white to match the background color, the applicants are finding much greater exposure when a keyword search is performed by a prospective employer. That means recruiters must now do much more work to find the truly “qualified” applicants. Multiply that by thousands of applicants using this “trick” to get ranked by search engines, and you can see how it could quickly cripple any system that relies on keyword searches. In essence, a recruiter would now have to manually read every resume in the database in order to identify a handful of key prospects once more and more applicants start using this “trick” to foil keyword searches. Vendors of keyword matching systems will need to filter each resume for “hidden” text, but will not be able to do anything to the resume since this would be considered tampering with the content. At least recruiters will be able to instantly see which applicants are trying to “pepper” their resumes with keywords and delete them from the database. We called a few other friends in HR and asked them to do the same thing. Two recruiters found similar results. It doesn’t appear that this practice is widespread at this time, but it will undoubtedly gain popularity as more and more applicants learn of this technique as they try to gain every possible advantage in finding a job in a soft job market. Perhaps employers will have to tighten the process for accepting resumes, the same way airports have tightened passenger security. One possible solution is to move from keyword searches to position profiles, in which applicants are asked to complete a series of position-specific questions that directly target the position requirements and bypass keyword searches altogether. The drawback: Applicants will need to complete a position-specific questionnaire for each position that interests them. On the bright side, at least recruiters will know that an applicant is truly motivated if he or she takes the time to complete a position-specific profile instead of simply pasting a resume to a job opening. A resume can still be accepted, but its importance in the initial screening process is secondary to the profile. It may take a while for companies to come up with the right strategies to combat what clearly has the potential to be a growing problem. Either way, the solution to the problem begins by recognizing its existence and analyzing what kind of impact it’s already having on your recruiting processes.

Article Continues Below

Joe Stimac (jstimac@accuhire.com) is President and founder of AccuHire.com and a frequent speaker and trainer on recruitment and selection issues. He is often quoted by leading business press including the New York Times, Chicago Tribune, and Employment Review, and is the author of Winning Career Strategies and other books. Mr. Stimac and his team are the creators of AccuHire HCMS, the first Internet-based applicant screening solution that pre-screens applicants against position-specific criteria and automatically generates highly targeted, performance-based interview questions for recruiters and hiring managers.

Topics

7 Comments on “Still Using Keyword Searches to Find Applicants?

  1. After reading this article I was very concerned with our resume scanning process. We are currently using an applicant tracking system in which the resumes are scanned and put through an OCR (optical character regognition) process which then converts the content of the resume to text. We did two tests based on the information in your article. The first test involved changing a block of resume text to white and scanning the resume. When the resume went through the OCR process the white text was ignored. The white print did not appear as text in our system and no skills were extracted from that information. In the second test we added additional, easy to identify, skills to the bottom of the resume. After the OCR process we verified that the resume was not given credit for the skills that we added. I did not find that our applicant tracking system picked up any hidden skills from the resume.

    You can read the original article at:
    http://www.erexchange.com/a/d.asp?cid=955087FFAAED4CCAA949554AECB91B39

    Post your own Article Review
    http://www.erexchange.com/p/g.asp?d=M&cid=955087FFAAED4CCAA949554AECB91B39

  2. Our system cost about $275 dollar (for main program) Plus another $175 (for email scanning module) PER PC WORKSTATION.

    It is currently being used on an internal network by four recruiters PLUS is accessible via email packets to TWO satellite offices in other locations.

    WORKS much better than most systems I’ve seen demonstrated costing between $5,000 to $50,000 dollars.

    You can read the original article at:
    http://www.erexchange.com/a/d.asp?cid=955087FFAAED4CCAA949554AECB91B39

    Post your own Article Review
    http://www.erexchange.com/p/g.asp?d=M&cid=955087FFAAED4CCAA949554AECB91B39

  3. This is one of the reasons we still prefer a system which allows keywords and skills to be entered manually for searching. That way the resume can be reviewed by a person and the actual skills deciphered. It is more work but well worth it in the long run.

    You can read the original article at:
    http://www.erexchange.com/a/d.asp?cid=955087FFAAED4CCAA949554AECB91B39

    Post your own Article Review
    http://www.erexchange.com/p/g.asp?d=M&cid=955087FFAAED4CCAA949554AECB91B39

  4. If candidates are learning to fool systems that use full text searching or basic word extraction using some of the methods described in this thread, it seems to me that the answer is to do away with these search methods to begin with (or at the very least to use them as a secondary search criteria to positively include or exclude certain words).

    In my humble opinion, the newest generation of search tools using quantitive based neural network technologies (I intentionally don’t refer to this as artificial intelligence because the term has been abused by some in the past) is the answer to this challenge. Unlike anything the HR world has seen to date, these tools mathematically analyze a large number of documents (such as resumes) to develop a real understanding of words and concepts in those documents. This understanding is not based on a fixed vocubulary, lexicon or static full text index, but is rather derived from millions of different character and word combinations that typically appear in documents. Like the human brain, the system develops millions of connections that create true artificial intelligence and dynamic learning as it is exposed to more and more documents (it will even learn your industry specifics over time).

    The specific solution I am most familiar with is developed by Engenium Corp out of Dallas Texas and will probably increasingly become a part of the ATS landscape as awareness of these technologies spreads (assuming the unstructured format of the resume continues as the principle descriptor of candidate qualifications and experience – a topic which is outside the scope of this thread). Engenium Corp’s HireReasoning solution was just named Top Human Resource Product of the Year for 2001 by Human Resource Executive Magazine so look for more to come on this front (www.engenium.com).

    These kinds of technologies are a big step forward in searching technologies that go a long way to reducing the full-text searching headaches that are becoming more commonplace as candidate populations grow in numbers and engenuity…Until the traditional resume is eclipsed by something better, these technologies present a compelling value proposotion.

    You can read the original article at:
    http://www.erexchange.com/a/d.asp?cid=955087FFAAED4CCAA949554AECB91B39

    Post your own Article Review
    http://www.erexchange.com/p/g.asp?d=M&cid=955087FFAAED4CCAA949554AECB91B39

  5. I agree with Bertrand’s comments that AI may be a good keyword solution in the future. However, In my opinion, I believe that the future may go even further and not use keywords of any type and actually put the work back on the applicant by asking them to select a specific position when applying and completing a position-specific questionnaire.

    We settled on this approach after extensive tests of methods to identify the best applicants for each position.

    In our final design, each applicant is screened and ranked according to position-specific criteria…no keyword searches, no resume scanning. It’s a direct match everytime. Recruiters simply select a position to fill, view the list of pre-screened, ranked applicants (review attached resumes) and start scheduling interviews.

    In our research we found keywords to be very ineffective since many users weren’t well trained in their use and applicants were “loading” their resumes with irrelevenat keywords in order to get noticed. This put a lot of work back on recruiters who had to “read the list of identified/qualified applicants.”

    The sad part is that many great applicants were being missed because they submitted “standard resumes which contained “verbs” instead of nouns and were often missed in keyword searches. I was at a client’s site conducting interview training when a HR manager came up and told me that her friend didn’t even make the list of applicants for a position for which she was very well qualified. We looked at her resume and realized she wrote a standard-“action verb driven” resume instead of a “noun-driven”–electronic resume.

    The question we struggled with was. “How certain are we that our keyword search identified the “best applicants?” The answer was—not very confident since so many variables came into play (user training, biased search strings, keywords in the resume, verbs instead of nouns, keyword loading —and now “hidden keywords”).

    The other problem we identified was–just because our keyword found a “great applicant” we still had to contact them to see if they were “interested” in the specific position. Many times a great keyword match resulted in a decline by the applicant who was not interested in a specific job/location.

    I would be very interested to learn about the other problems recruiters have experienced when they conducted keyword searches.

    Thanks in advance!

    You can read the original article at:
    http://www.erexchange.com/a/d.asp?cid=955087FFAAED4CCAA949554AECB91B39

    Post your own Article Review
    http://www.erexchange.com/p/g.asp?d=M&cid=955087FFAAED4CCAA949554AECB91B39

  6. Hidden keywords in resumes are only one of the problems with keyword searching. Keyword searching has so many limitations that it will soon give way to the new matching technologies that have recently been introduced to the recruiting industry.

    While there will never be artificial intelligence that can take the place of a human, there are technologies available that can be much more productive than keyword searching. We’ve found that we can eliminate 70% of the time it takes our clients to read resumes by utilizing an advanced matching technology.

    The technology reads the job description for you, identifies the key “concepts” not “keywords” of the job description. (The difference between concepts and keywords is very detailed and I don’t want to bore you with it, however, just know that it involves PhD’s, natural language process, sentence structure, word stemming, etc). The technology does the same for resumes and allows for much better matching results than keyword searches. These technologies are available and will soon become much more prevalent in recruiting. Keyword searching will be a thing of the past.

  7. Enough with the keywords already.

    Whatever happened to picking up the phone and ASKING the applicant?

    I for one do not believe the SOFTWARE should be excluding or making candidate matching decisions which I am supposed to be making.

    Recruiting and interviewing/candidate selection is more art than science .. the software is a tool which should produce a group which “comes close” that’s it.

    I don’t want it to be any more intelligent than that as I can handle it from there thank you very much.

    (hope you realize this is humorous as well as serious) – Frank

    You can read the original article at:
    http://www.erexchange.com/a/d.asp?cid=955087FFAAED4CCAA949554AECB91B39

    Post your own Article Review
    http://www.erexchange.com/p/g.asp?d=M&cid=955087FFAAED4CCAA949554AECB91B39

Leave a Comment

Your email address will not be published. Required fields are marked *