15 October 2019
When we set out to build the best business card and conference badge scanning app, we had 3 goals:
- accurately capture card and badge information with guaranteed accuracy,
- search across the internet for the specific entity on the card, and
- complete these within 2 seconds.
As someone who attended many meetings and conferences, I found myself feeling anxious about meeting people for the first time but not knowing much about that person. I’d ask questions like, “Where are your offices?” What I really wanted was a real-time “Things you have in common”. I was looking for an app that could scan a business card or a conference badge, get all the information correct, perform a search across the internet for that specific person (better than Google), return this information and things we shared in common, and all within 2 seconds.
Guaranteed Accuracy
To accomplish this vision around user experience, the foundation was scanning accuracy. After all, if Wingman could not correctly capture the characters on the card, group them into their fields and label those fields accurately, it would simultaneously frustrate the user who was just trying to capture a contact from the physical business card and prevent Wingman from performing the search across the internet with the correct information.
We evaluated a few different business card scanning mobile apps, and there are many! We scanned 100 business cards using the top 3 business card scanning products in the industry. We performed these scans under very controlled conditions – good lighting, stationary, and close to the card. The best results from these 3 mobile apps was 42 of the 100 cards captured all the information on the card correctly. The other 58 of the 100 had at least one incorrect field on the card. With these low numbers, we could not use any of these products as the foundation for our vision around user experience.
During customer discovery, we also learned that many users had either used an existing real-time business card scanning app and uninstalled it because “it took longer to review and correct mistakes than just typing it in from scratch or used a mobile app like Contacts+ that used humans to manually create the contact based on the scanned business card. The latter was not going to work for Wingman, as it would not satisfy our “within 2 seconds” goal.
We developed a set of machine learning algorithms to achieve accuracy that was close to human-level accuracy. However, we also realized that machine learning alone would not provide the guarantees that would build trust that we wanted with the Wingman brand. So, we developed a set of algorithms that could detect when a field was incorrect and had to be reviewed by, and possibly corrected by, a person. The combination of the machine learning algorithms, error detection, and human review has allowed us to achieve near-perfect accuracy on US-based, English language business cards. We are now deploying these machine learning algorithms on other languages.
Real-time Internet Search
Our second goal was to return search results for the specific person on the business card. We wanted our search to be better than Google’s search. Google performs great at full-text search, but it returns many false positives and fails to leverage the field-level context that is present on a business card. So, we focused our energy on finding the matching profiles for a person within social networks, including LinkedIn, Twitter, Github, Meetup, AngelList, and a few dozen other social networks.
We created a search algorithm that has helped us achieve significantly less false positives than a full-text Google search and a user experience within Wingman that allows the user to find the correct entity quickly when there is ambiguity in our search results.
Complete Within 2 Seconds
The final goal was to achieve the first 2 goals within 2 seconds. Without going into details, we figured this out. And while some older devices take up to 3 seconds, new devices complete a scan within 1 second.
The Wingman App is now on the Google Play Store and will be on the App Store later this week.
Please let us know about your experience. These are early days of a long journey and the faster we get your feedback, the better Wingman will be at, well, being your wingman.
