When Evernote added a business-card scanning feature to its iOS apps last December, I thought of my desk.
That furniture long ago became the dumping ground for every card I collected at a conference, briefing, lunch, reception or other work-esque event. The resulting stack never shrank: Typing their info into my contacts app was too much work, and I couldn’t recycle most cards after connecting with their owners on LinkedIn, because too many of them forgot that using a business-networking site entails providing current contact data. And I’d seen enough griping about LinkedIn’s CardMunch app, the most obvious solution, to shy away from that.
[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":1513737,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"business,mobile,","session":"C"}']Like CardMunch, Evernote promised that its app could digitize cards’ text (but via software alone instead of slower human transcription) and cross-reference that with their owners’ LinkedIn profiles. It proved glitchy at first, but then Evernote followed up with a May announcement that it would provide a year’s free use of this premium feature (normally $45/year) and a June update adding such improvements as capturing the back of cards.
So it was time to get to work.
AI Weekly
The must-read newsletter for AI and Big Data industry written by Khari Johnson, Kyle Wiggers, and Seth Colaner.
Included with VentureBeat Insider and VentureBeat VIP memberships.
One thousand, three hundred and thirty-three cards later, here’s what I learned about how Evernote’s ambitious experiment in machine vision intersects with my favorite form of portable graphic design.
It works most of the time
Of those 1,333 cards scanned on my first-generation iPad mini, the app captured 1,179 cards with sufficient fidelity to require fixing nothing beyond a letter or two in an e-mail. It almost never flubbed phone numbers, while street addresses were trickier but less risky in practice. You’re far more likely to search for somebody via name or employer than address, and either way the card’s photo can set you straight.
(Note that my iPad has one of the lesser cameras Evernote recommends for this feature; the company was sufficiently concerned about first-gen iPad minis’ cameras that it disabled auto-detecting a card for them. When I asked about that downgrade, product manager Martin Cheng said it would be restored. An Android version, meanwhile, remains on the to-do list.)
Of the remainder, 107 cards had at least a name or company digitized in searchable form, but entailed either retyping most or all of an e-mail address (without which Evernote can’t find a LinkedIn connection) or more than two tries with the iPad mini’s camera. The app found so little data on 47 cards that I had to hit the keyboard to get anything searchable.
Different shaped cards didn’t bother Evernote, and oddball materials required less fussing than I feared. Most translucent-plastic cards scanned properly on a white background, and one of Steve Wozniak’s etched-metal cards only demanded I hold the iPad at an angle where the room lighting heightened the contrast between its text and its background. The same drill was required to avoid glare on cards with glossy finishes.
The LinkedIn integration was something close to magical, not just fetching people’s titles and employers but correcting spelling errors in names caused by Evernote’s own scanning. I also found it weirdly fascinating to see which startup types remained at the same company or had pivoted elsewhere.
[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":1513737,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"business,mobile,","session":"C"}']
Exciting? Not so much. But the process took less time than I’d feared: I averaged about 100 cards an hour and could have gone a little faster if I hadn’t been taking notes.
Character flaws
One of the most common scanning glitches is also among the easiest to avoid: Placing a descriptive character or other text too close to an e-mail or phone number. Dozens of cards yielded numbers beginning with “0” because Evernote saw an “O:” (as in, “office:”) right next to somebody’s digits, and many others were digitized with e-mail addresses starting with “e-mail” or ending with a Web address.
The placement of text on a card caused trouble too. Splitting an e-mail into two lines ensured the app missed it, while having some text run vertically or otherwise offset from other text invited further confusion.
And the app often missed text near a card’s edge. Evernote frequently skipped e-mail addresses listed on the last line, while part of the street address on VentureBeat writer Dean Takahashi’s card got scrambled because a suite number bumped up against the right margin.
[aditude-amp id="medium2" targeting='{"env":"staging","page_type":"article","post_id":1513737,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"business,mobile,","session":"C"}']
Font fails
I wasn’t surprised to see the app ignore handwritten corrections and text in script, blocky or other artsy fonts in microscopic point sizes, or fonts with shadow or other graphic effects. And the guy who chose Comic Sans… apparently, humans aren’t alone in rejecting that font.
But the app had a surprisingly difficult time discerning other typefaces. It couldn’t handle the mix of thick and thin strokes in Bodoni and the dense small-caps fonts of some Congressional staffers’ cards — one “mail.house.gov” became “mail.liotjse.gov.” And the abstracted simplicity of some sans-serif typefaces like Futura led to it mixing up a’s for o’s or i’s for l’s.
Clashing contrast
The app can deal with dark text on a light background or the opposite, but not in-between contrasts. Cards presenting some info in a colored area — for example, an older Evernote employee’s card displayed its former Mountain View, Calif., address in white text on a green box — risk having the app ignore those highlights.
For some reason, Evernote rarely read white text on a blue background. And woe betide a card with a muted contrast plus off-axis text: I hate to break this to my county’s economic development authority, but between its cards’ lettering appearing in a shade of brown and at a 45-degree angle, Evernote can’t stand them.
[aditude-amp id="medium3" targeting='{"env":"staging","page_type":"article","post_id":1513737,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"business,mobile,","session":"C"}']
Two faces
Evernote can’t do much with the back of a card beyond take a photo: It can’t digitize that image, and you only see this option after saving the contact.
Most of the time, you only lose a Web address, company logo or foreign-language version of the card. Alas, one of the world’s largest electronics manufacturers issues cards with people’s names and titles on one side but other contact info on its opposite.
I hope Samsung doesn’t try to resolve this by buying Evernote.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn More