How to Turn a Business Card into a Contact

OCR guide · 6 min read

You come back from a conference with a stack of business cards. Every one of them holds a name, a phone number, and an email address that you will want in your phone eventually, and typing them all in is the kind of job that quietly never gets done. The cards sit in a drawer, and six months later you cannot remember who any of them were.

There is a shortcut. A card is just text on a small piece of paper, so recognition can read it, and the fields it contains follow patterns predictable enough to be sorted automatically into a contact file your address book will import.

What a vCard actually is

The file behind every address-book entry is a vCard, usually with a .vcf extension. It is a plain-text format that has been around since the nineties and is understood by essentially every contacts application on every platform: iOS, Android, Outlook, Google Contacts, macOS Contacts, Thunderbird.

Inside, it is unremarkable. A line for the formatted name, a line for the organisation, lines for each phone number and email address. Because it is a standard rather than a product, a vCard produced on one device imports cleanly into another without any account or sync service in between. Double-clicking the file on a desktop opens it in your contacts app; emailing it to yourself and opening it on a phone does the same thing.

That portability is the reason it is worth producing a vCard rather than just copying the text out. Copied text still has to be typed into fields. A vCard is already in fields.

Turning a card into a contact

The process has two stages that are easy to confuse. First, recognition reads the card and gives you its text. Second, that text is sorted into contact fields. Both have to go right.

1. Photograph the card straight on, filling the frame. Cards are small, so resolution matters more than usual — a card shot from across a desk gives recognition very little to work with.
2. Extract the text. You should get the name, the job title, the company, and the contact details as plain lines.
3. Have those lines sorted into a contact and saved as a .vcf file.
4. Open the file to import it, and check the fields before saving.

In Textquill this is the Extract menu, which offers a contact card option on any text you have pulled out. It produces the summary on screen and a downloadable file, entirely on your own machine.

Why the fields get mixed up, and how to spot it

Sorting text into fields is guesswork, and it is worth understanding where the guesses come from so you know what to check.

Email addresses are the easy case. They have an unmistakable shape, so they are almost never wrong.

Phone numbers are mostly reliable but occasionally over-eager. A number has to have roughly seven to fifteen digits to qualify, which rules out years, suite numbers and postcodes. A long reference number printed on a card can still slip through.

Names and companies are the genuinely hard part, because on a card they are just two lines of text with no marker saying which is which. The usual heuristic is that a company name often carries a legal suffix, and that a person's name is two or three words of letters with no digits. That works on most cards and fails on the ones where the company is a person's name, or where the designer set everything in the same weight so there is no visual hierarchy to fall back on.

Job titles are recognised by vocabulary. A line containing a familiar role word is treated as a title. An unusual internal job title will not be recognised as one and may end up as the company instead.

The practical consequence: always glance at the parsed fields before importing. A tool that hands you a half-filled contact you can finish in ten seconds is more useful than one that confidently fills every field with a guess, because a wrong contact in your address book is worse than an incomplete one — you will not notice it is wrong until you need it.

Getting a clean read from a small card

Cards fail recognition for reasons specific to their size and design.

Low resolution is the main one. A business card is about nine centimetres wide, and if it occupies a quarter of your photo the lettering is only a few pixels tall. Get close, or crop to the card and scan again.

Glossy stock reflects your light source straight back into the lens, wiping out a band of text. Tilt the card slightly, or move the light rather than the card.

Dark or coloured backgrounds reduce the contrast between ink and paper. Raising contrast before recognition usually rescues these. Reversed cards, with light text on a dark background, need the tool to handle inverted text.

Decorative type is the one that resists everything. A logo set in a script face is a picture as far as recognition is concerned, and the company name may have to be typed in by hand.

Doing a stack of them

For a conference stack, batching is worth the setup. Lay the cards out under even light and photograph them one at a time rather than several to a frame, because a single photo of six cards gives each card a sixth of the resolution. Work through them in one sitting while you still remember who is who, and add a note to each contact as you import it — where you met, what you discussed. That note is the part your future self will actually want, and it is the one thing no recogniser can extract from the card.

Keeping it private

Business cards carry other people's personal information, which is a reason to care where the processing happens. A card handed to you in confidence should not be uploaded to a service you have not vetted, and in some jurisdictions handling other people's contact details brings obligations of its own.

Recognition that runs entirely on your own device sidesteps the question. The card is read in your browser, the contact file is written on your machine, and nothing about the person on that card is transmitted anywhere.