Practice
Business card scanning
Business card scanning uses OCR to turn a photographed card into a structured contact — name, company, phone, email — ready to import into an address book.
A business card scanner does two jobs in sequence. First, optical character recognition (OCR) extracts raw text from the photo — modern engines like Tesseract's LSTM models or the on-device frameworks in iOS and Android handle this well for clean, Latin-script cards. Second, and harder, field classification assigns each text fragment to a slot: is "Mercer" a surname, a company, or a street? This step leans on heuristics and named-entity models, and it's where most scanning errors actually occur.
The practical workflow matters more than the technology. A stack of cards from a conference is a pile of dormant intentions; the cards that get scanned, annotated with where you met and what you discussed, and followed up within a week become relationships. The ones that stay in the drawer become recycling.
The quiet issue is where your scans go. Many free scanner apps process images on their servers, and several have built businesses on the data flowing through them — the card you scanned becomes a record in someone else's database. Checking whether OCR runs on-device or in the cloud is worth thirty seconds before you photograph a hundred strangers' details.
From photo to fields: the OCR pipeline
A typical pipeline runs four stages: detect the card and correct perspective, binarize and deskew the image, run character recognition, then classify the recognized lines into contact fields. Recognition accuracy on clean print is high; what breaks it are design choices cards are full of — stylized display fonts, white text on photos, vertical layouts, logos that OCR reads as letters, and bilingual cards mixing scripts. Field classification adds its own errors: titles like "Dr." migrating into first names, the company line landing in the address, mobile and office numbers swapped. The rule of thumb: always glance over a scanned record before saving, because a transposed digit in a phone number is invisible until the call fails.
The fine print of free scanner apps
When a scanning app is free and OCR happens "in the cloud," the cards you photograph are uploaded to the vendor's servers — and the people on those cards never agreed to that. Several popular scanners have historically used uploaded cards and synced address books to build people-search or networking databases, meaning the contact details a stranger handed you on paper ended up queryable by third parties. For anyone scanning in a professional context in the EU, that quietly makes you the person who passed personal data to an undisclosed processor. Prefer apps that state OCR runs on-device, read the data-sharing section of the privacy policy, and treat "free, cloud-based, ad-supported" as a description of the business model, not a gift.
Getting scanned cards into your personal CRM
The scan is the cheap part; the context is the value. Whatever scanner you use, export the results as vCard (.vcf) or CSV — both are universal — and enrich each record yourself with the two facts no OCR can see: where you met and why you'd want to talk again. Endearist imports both formats and runs duplicate matching on the way in, so re-scanning a card you already have proposes a merge instead of creating a twin; from there the card becomes a contact with notes, a follow-up reminder, and a relationship history rather than a JPEG in a folder. Done within a day or two of the event, while "the tall Dane who knew everything about batteries" is still attached to a face, this turns a stack of cardboard into an actual network.
Frequently asked questions
- How accurate is business card OCR?
- On a clean, conventionally designed card in Latin script, modern OCR reads nearly everything correctly; realistic field-level accuracy across mixed real-world cards is lower, because decorative fonts, logos, glossy reflections, and unusual layouts confuse both recognition and field assignment. Numbers deserve special attention — a single misread digit invalidates a phone number. Plan on a two-second human review per card rather than expecting unattended perfection.
- Do card scanning apps upload my contacts?
- Many do. Cloud-based OCR necessarily uploads the card image, and some apps additionally request full address-book access for "sync" or "matching" features. Historically, several scanner vendors aggregated this data into searchable people databases. Check two things before trusting one: whether the privacy policy says recognition happens on-device, and whether the app demands contacts permission it doesn't strictly need. Paid or on-device-first apps are structurally less motivated to harvest.
- Is scanning business cards GDPR-compliant?
- Handing over a card is widely read as an invitation to be contacted, so storing it for that purpose is generally defensible under legitimate interest in a business context. Complications start with what happens next: uploading cards to a cloud service makes that vendor a processor you're responsible for choosing carefully, and using the data for bulk marketing exceeds the original expectation. For purely private networking, GDPR's household exemption usually applies. This is general information, not legal advice.
Last updated: 2026-06-10
Tend relationships, not records.
Endearist is a local-first personal CRM. Free up to 25 contacts.
Start free