Sujith Nair posed the following question in response to Automated Job Tagging in Action: “How do you ensure that advertisers posting same job advt with similar title can be filtered out?” Great question Sujith!
In CareerBuilder Suffering from Tag Spam I stated that CareerBuilder, “…primarily due to ‘tag spam’ …dropped to last place in a new job search shootout against Jobster, Indeed, and SimplyHired.” Interestingly, even though CareerBuilder suffered from tag spam, they returned no duplicate entries. Compare this to a rate of roughly 10% duplicate ads returned by SimplyHired.
Two lessons here:
CareerBuilder, by charging for ads, effectively discourages duplicates and near-duplicates. The ROI for advertisers on duplicate ads approaches zero. Even Craigslist, the granddaddy of free classifieds, has moved to impose low fees on some categories and locations to reduce spam. Quality is tied to price – and businesses built on free services are going to struggle with spam. In contrast, TheLadders leads the pack in data quality by having two staff members review each and every ad before posting. Charging for services increases quality, and reduces duplicates.
Vertical job search sites aggregate free content – some from high-quality sources (e.g. CareerBuilder and Monster) and some from lower-quality sources (e.g. Craigslist and GoogleBase). No disrespect intended towards Craigslist and GoogleBase; but their pricing model (free or low-cost) yields a higher percentage of marginal content. Thus, the verticals become infected with this lower-quality content.
So, to reduce duplicates and near-duplicates, the three best choices are: 1) charge for each job ad; 2) aggregate only from trusted employers or sources that charge for each job ad; or 3) charge end-users for adequate data cleanup.
If you care about quality, free is the wrong price.
PS: Free test-drives are a completely different animal. Potential customers should have the chance to determine the personal value of a service before plunking down their money.
Dear robert,
Thanks for the reply. However, i beg to differ on the solution provided here. Though introducing a price tag is a big brake on the spams, but its not a scallable option. There are plenty of advertisers, mostly consultants who don't mind paying for advertising. Here, its a catch 22, as you cannot filter their content and at the same time they get a license to post as many jobs (most of the times duplicates) because they are ready to pay. They are ready to do anything to increase their listings' visibility.
Also, manual screening is out of question. Its not scallable....
How do we tackle these advertisers, who are now "clients"? I know there won't be any fool proof solution for this, but i guess the solution lies in the combination of policies supported by technology. What do you think ?
thanks,
sujith
Posted by: Sujith Nair | July 26, 2006 at 10:04 PM
Hi Sujith,
I agree with you that there isn’t a foolproof solution; and a combination of policies and supporting technology can reduce the duplicates. A couple additional thoughts …
If the advertisers are willing to pay, then it seems you have an opportunity to help them solve their problem in a way that doesn’t flood your site with duplicates. Can you offer a special service targeted at these advertisers … perhaps banner ads, featured listings, or an ad listing optimization service? If so, then you’d be working together rather than at cross-purposes.
On the technology front, Indeed seems to be doing the best job of handling duplicates and near duplicates once they’re in the database – they suppress them during the initial display, but provide a link to display all.
Best wishes!
Bob :-)
Posted by: Bob Wilson | July 27, 2006 at 07:05 AM
Hey bob, i believe i am not stretching this discussion too long. Its an interesting topic and therefore i cant help but take this discussion further :-) You have a point and i agree that increasing visibility of these clients through a different section like text/banner ads etc can give them a better ROI. But care should be taken that organic results are not influenced by results of those paying more.
Its a good idea to screen all duplicate advts behind a link. But now the big question is, which one to show and which one to hide. If the recent advt is shown and others put behind a link, then the relevance may go for a toss. Also, this needs to be done real time, ie, when the results are rendered.
What say? Sorry, i m playing devil's advocate here ;-)
Posted by: Sujith Nair | July 27, 2006 at 08:55 PM
Hi Sujith,
My thought is that there isn't a single 'right' answer. If we're talking duplicates (or near duplicates) then I'd display the most current first ... this way if there are subtle updates to the ad, users will see them, and it reinforces to the job seeker that your content is fresh.
If there are multiple duplicates with the same date, coming from different job sources, then I'd let the sources bid on which link is displayed while the others are hidden behind the dup link. It may turn out that different sources will pay different amounts to be listed first depending on the occupational category ... e.g., if a tech job, Dice might pay more to be listed first than Monster.
In the end, your decisions need to balance the needs of your customers, and support your business model.
Best wishes,
Bob :-)
Posted by: Bob Wilson | July 28, 2006 at 06:46 AM
free is hardly hurting craigslist -- http://internetinc.com/Craigslist-2008-revenue-to-reach-81-million
Posted by: Eric Shannon | April 11, 2008 at 02:19 PM