Yelp, Angie's List, Amazon -- these companies and many others offer user reviews of local businesses. But the authenticity of many of those reviews is questionable, which is why an Israel-based company is offering a new service that it claims provides "honest, authentic referrals."

The company, WhoDoYou, uses machine learning and natural language processing to collect and sort public conversations -- mostly from Facebook, and mostly positive recommendations -- about painters, lawn services, restaurants and the like.

Free Service

The free service currently offers over half a million searchable results for local businesses from thousands of cities in over four dozen countries. Each recommendation can be traced by the reader back to its source. Users do not need to be registered to see the reviews. Those who log in with Facebook credentials can privately get recommendations from the conversations of their own circle of friends.

While WhoDoYou has not officially released its user population, company officials claim the site – and its associated Android app -- has experienced 10 percent growth week over week for the last 10 weeks. An iOS app is expected soon.

Looking for a painter in Livingston, N.J., for instance, the site shows a series of comments gleaned from Facebook. Some recommendations appeared genuine, such as:

  • J M. answered: Marcus McNamara 908 419 8065 Good friend from Summit High. Magnificent Painter Tell him I sent ya Bob.

While others seem to be promotional plugs from painting services:

  • David C. answered: Call BBJ remodelers (732)356-1512 we can help you out

'Very Few Complaints'

Yoav Schwartz, CEO and founder of WhoDoYou, told CMSWire that these recommendations were taken from a page on Facebook where the request was for "interior painter recommendations."

"You can see that people [on Facebook] are asking the list who they should use as a painter, and this person [David C.] is suggesting his company, BBJ Remodelers," he said. "Presumably that has value to the community group, and we prefer not to hide posts that may be useful. We would treat differently a post that is made TO the group, saying something like –'You should use BBJ Remodelers - the best contractor out there...' "

2014-26-June-WhoDoYou-Images.jpg

A screen from WhoDoYou.com

Have there been complaints from the posters for reusing their public comments without permission?

"We've had very few complaints about using posts," he told us. "In fact, it's more common for providers that are no longer in business to request that we remove their listing because they're getting calls from people who found them on WhoDoYou."

Future Plans

Currently, the up to five stars "Bottom Line" ranking for each vendor is auto generated and based on the system's assessment of the recommendations. "Since WhoDoYou primarily scans for positive recommendations, most providers have either a four out a five star rating. But that will change over time as we invest more deeply in differentiating the syntax and sentiment of a recommendation," Schwartz told us. 

Does the service misrepresent customer assessments of a local business by focusing on positive recommendations?

Schwartz told us his company doesn't "have a policy of taking primarily positive recommendations." Instead, he said, it takes comments from one friend asking another for a recommended provider, which generally is about "a provider that he or she has used and would promote."

Occasionally, he said, "people will respond by saying who not to use, and indeed these kind of responses are added periodically when they come up. But it is an extreme minority of cases'' because most people want a recommendation, not a pan.

Spamming, Positives

Couldn't a vendor spam Facebook with promotions — and wouldn't WhoDoYou reflect that spam?

"Without divulging too much proprietary info [about our algorithms]," he said, "we search for patterns of questions that are typically used when one friend asks another for advice or recommendations."

Although he acknowledged that mistakes can be made by parsing so much unstructured data, Schwartz said that "the vast, vast majority – by our calculation the precision is greater than 96 percent – of posts you'll see are legitimate responses from one friend to another recommending a local provider."

The entire online recommendations scene, which appears to be full of fake posts, is ready for a rebirth. If it can find a sustaining audience and maintain credibility, companies like WhoDoYou could become key parts of that next wave.