TagCow is a company that I quite frankly never thought we’d hear from again. We first discovered them in March when they launched a product that tagged things in photos, seemingly by magic. If it was automated, they’d done something with computers that even Google still uses humans to do – recognize stuff in pictures.
Then it turned out that TagCow was using humans also, leveraging Amazon Mechanical Turk to get photos tagged for cheap (to their credit, they never said one way or the other). People were paid $0.04 to tag a group of five photos.
It turns out that model doesn’t work, either. Founder Michael Droz emailed today to say “After…tagging over a million images for free we looked at our balance sheet and realized it was make money or close up shop time.”
But they didn’t close up shop. Instead, they started pitching companies on a tagging service, and at least two have now signed deals: Art.com and AutoByTel. Both use TagCow to tag images, for a fee. The company says October was a profitable month for them based on these deals.
Any site which has a lot of images can potentially benefit from having more metadata about those photos. Search results will yield much better results (and the site is better optimized for search engines in general), and perhaps a sale will be made that otherwise wouldn’t be.
TagCow, which says it can tag a million photos a day, and is looking for more customers.
They also say they’ll keep their user-facing service but are now charging a fee.
Mooo.
If theyre gonna have a cow in the logo, I expect it to come in 1 litre cartons
Weird way to crowdsource commercial image tagging.. I’m shocked that they can get do a million images a day.. who are the poor saps tagging photos?
Also, like Fabian, I think the number of images should be measured in liters.
Not sure about the image tagging for enterprise (except for the design department, if there is one). However, Powerpoint Slide tagging could be of a killer. Just my 2 cents.
Chris,
If you want to know a little about our workers here’s a link to a very interesting survey we did in June.
http://www.tagc...ker_survey_0608
Also, it’s worth noting that the real power in what we do or the secret sauce if you will is our ability to create templates that abstract the unique properties and meaning of our clients taxonomies and allow anyone to tag/categorize even the most challenging of images/products. We also use decision trees to send images through the tagging process multiple times using different templates each time.
The point is we don’t just throw enterprise images up on MTurk and get whatever random tags someone decides to add…we drive this process. We know exactly what the results and quality will be before we start. It’s very powerful service for any company with a lot of images.
Glad to see more web 2.0 features are gaining popularity in the enterprise. Check out our views on enterprise adoption of consumer tools here:
http://www.open...prisetrends.php
After spending so much money, you would have think they could have automated the images.
Wow…per their user survey 30% of these Tagcow workers have advanced degrees.