Recruitment job search engines - 07/2001

A good recruitment website is more than a pleasing web design, lots of candidate services or even a large selection of jobs. Candidates will revisit a site only if the functions work. At the heart of the site must be a useful search engine - it is usually the first function a candidate will try. This article discusses search engine technology for online recruitment.

Job searches are prone to producing inaccurate results or much worse, no results at all. The attention span of even the most determined online job-hunter is short and if candidates have to constantly modify the search criteria he/she will quickly give up and probably never visit the site again. There are many reasons for poor results, for instance, data entry of accurate job details is obviously prone to human error. Also, the implementation of the engine can sometimes be too literal, only returning exact matches rather than near misses. Inaccurate results from search engines are even less useful using push-technology such as jobs by email or SMS (Short Message Service to mobile telephones) because the candidate does not have the opportunity of re-adjusting the search criteria once the results have been delivered.

There is a balance in collecting enough information to produce quality results and making it too complex for the candidate. Remember the 3-click rule - the principle that access to any feature of an application, or each logical step in a process, should require no more than 3 clicks. Specifying a job is obviously going to take more than 3 clicks. However, simple legitimate uses of cookies (small pieces of data that a site can send to the usersí computer) can be used to quickly login the candidate and run a previous search that the user took many clicks to set up and quickly obtain the latest results.

The presentation of search results is important and should contain useful summaries sorted on relevance. Most WWW search engines list in order of relevance and even show a snippet of the document on which the match was made so you can decide whether to follow the link or not. Many job sites however, expect us to laboriously click thruí each link and read the whole of the job description one by one. Often job summaries are sorted on posting date, or salary rather than relevance to the search criteria.
What is needed is fuzzy matching, predictive searching and data-to-data matching. When a candidate searches for a salary of 25,000, he would probably also like to consider jobs that fall just outside that range e.g. 20-24K. Jobs that have a salary range that encompasses his desired salary should return a rating of 100%, those just outside something less than this.

Fuzzy text searching has been well known for WWW search engines - this involves detecting common misspellings, alternate word endings, soundex (sounds-like) and other algorithms. Predictive searching uses contextual information, for instance, a computer programmer working in aerospace is quite likely to be interested in ADA programming jobs (you have to know the sector!). These do not have to be included in the current search but may be suggested to the candidate, particularly if only a few results have been returned. Alternatively, they could be included albeit with a low rating. Data-to-data matching involves extracting the contextual information from both the candidatesí information (C.V.) and job specifications and doing a two-way match. Unless the information has been entered in a structured fashion, this is extremely problematic. Mostly, candidates are not interested in doing this - although some companies send CVs and job advertisements to the Far East for manual processing.

When specifying your next online recruitment search engine, bear in mind Albert Einstein memorable maxim ëThings should be made as simple as possible, but not any simplerí. More sophisticated software and simpler for the user.

www.careerbox.co.uk