Case Study | Delectable

Deliver an Awesome User Experience

"The popularity of Delectable's app is growing and our users demand accuracy. Delivering an amazing user experience has been vital to our success. CrowdFlower has made it possible for us to continue delighting the people using Delectable while scaling over 500 percent. We love CrowdFlower." - Alex Fishman, CEO, Delectable

The Company

Delectable is the maker of an iPhone app that helps you “[r]emember the wines you’ve tasted, and discover wines you’ll love.” It also helps users buy wines easily – all in one place.

Wine lovers enjoy knowing the details behind the wines they drink – for instance, the winery, its region, the vintage, ratings, and tasting notes from fellow wine drinkers. Delectable is a fun, practical, and convenient way for wine lovers to keep an inventory of the wines they have tasted alongside their own notes on a particular bottle.

The Challenge

Cataloging an Infinite Dataset and Training Algorithms

The number of wine labels in the world is seemingly endless and only continues to grow. California alone has more than 60,000 registered wine labels.  Delectable has more than one million wine labels in its database already. When Delectable users take a photo of a wine bottle label and upload it to the app, the company’s technology identifies a great deal of information about the wine. However, identifying the vintage year – something that is particularly important to the discerning wine drinker – is extremely difficult. Delectable also uses a proprietary computer vision solution to match identical wine labels, but it wanted to add another verification layer to ensure accuracy.

In order to deliver an exceptional experience to app users, Delectable has to provide accurate information about the uploaded wine label within minutes, not hours. It’s no surprise that Delectable experiences a high volume of uploads and heavy app usage in the evenings and over weekends. The company needed an efficient, cost-effective solution that could accurately identify information on wine labels uploaded – especially during these peek user-times.

The Solution

An On-demand Workforce Ready to Label Data at Anytime

Delectable turned to CrowdFlower’s platform and our on-demand workforce. Through CrowdFlower’s API, the company set up an ongoing job that continuously sends newly uploaded photos of wine labels through the platform for verification and categorization by online workers.

Workers are asked to verify that the photo is a wine label. If it is, the workers are asked whether the label matches another wine label in its system and then to verify the vintage year from a drop-down menu. To ensure accuracy, three to seven people categorize each wine label. The number of workers varies depending on a few factors, including the degree to which each worker’s answers match and the worker’s accuracy track record. CrowdFlower’s platform delivers one aggregated answer back to Delectable’s app via a webhook. This results in near real-time processing of the wine labels for app users. Delectable also uses this label matching data to continuously improve its algorithms for its computer vision technology.

The Results

A Scalable Business Model and Happy App Users

Since Delectable began using CrowdFlower’s platform, it has been able to decrease the amount of time it takes for a user to obtain information about a wine label he or she uploads by 45 percent and does it in a cost effective manner.