“I do not think ‘standard’ codes work.” That’s Jon’s response to my Targeting Your Job Search post.
Jon goes on to say:
“There are way too may job titles to cover all the opportunities on the internet. In addition jobs vary greatly based on each company. For example: Do you think a project manager for Coke needs the same skill set as a project manager for Amazon.com? There are simply too many variations out there to try to "box" jobs in.”
Good feedback Jon.
I have to disagree with you though, because many standard coding systems do work. Zip codes work for mail delivery. Clothing size codes work pretty well … at least better than if stores had racks and racks of clothes with no size codes. And libraries are organized using standard codes … it’s much easier to find the book I’m looking for when the books are in order by code.
Still, most standard coding systems used for jobs fail for six reasons:
2) Too many codes – paradoxically, having too many codes is no better than having too few codes. That’s why In Tagging Jobs for Search I disagreed with Charlene Li on the value of user-generated tags.
3) Poorly defined codes – it’s not enough to have a list of code titles. To be useful, each code should be defined and provide a list of commonly included job titles. Better yet, the definitions should list excluded codes. For example, look at the ONET definition for wholesale and retail buyers, except farm products.
4) Overlapping codes – if a job site lists the standard codes ‘manager’ and ‘sales’, and you have to choose one, which one is correct for a ‘sales manager’ job? To succeed, standard codes cannot overlap. That’s why ONET provides many different codes for sales-related jobs, as illustrated in the image at right.
5) Untrained users – standard coding systems are like foreign languages. Not surprising then that infrequent users aren’t fluent. So asking untrained users to assign standard codes results in unacceptable error rates.
6) Apathy – adding accurate occupation codes to jobs costs money; so as long as users remain satisfied with the limitations of current keyword-focused job search engines, there is little motivation for adoption.
With all of the above problems, I can see why Jon doesn’t believe in standard job codes. But by combining a well-researched occupational taxonomy (e.g. ONET) with an accurate automated coding system (e.g. O*NET-SOC AutoCoder) the problems are eliminated, and standard codes do work for grouping related jobs.
So, what does this mean for Jon’s project manager jobs at Coke and Amazon? First, we’re left to guess about the specifics of these two example jobs – is the job at Amazon for a software engineering project or a product warehousing project? Is the job at Coke an R&D project for a new flavor of Coke, or a cost reduction project?
In the end, no coding solution solves every problem (e.g. library books can still be hard to find); but accurately structured data provides a better starting point than a pile of unstructured data.