Abstract
Capacitor Partners is a technology consulting and product management firm. As part of our product research process, we were commissioned to analyze the accuracy, experience, and pricing of mobile scanner applications, which use OCR technology to read, process, and present the information provided on a business card, into a digital format.
In the following report, we list the applications we are using in our test, the card parameters we are testing, and the testing methodology that we have designed to ensure the most reliable results possible.
We conclude the winners in a total of five categories, accuracy and reliability, overall speed, the versatility of features, pricing, privacy, and security, and then declare the overall winner.
Our results and input are all enclosed in the following sections.
In search for the best business card scanner
Card Scanner Selection
We tested and compared the scanning capabilities, UX, and pricing for some of the most known and powerful business card scanners that are available for iOS and Android devices.
The apps we are testing are CamCard, ABBYY, Covve Scan, and Biz Connect. All four apps:
Are available for both iOS and Android
Have 4+ ratings on both stores
Are free to download and try, although they offer in-app purchases and subscription packages for unlimited use and the unlocking of features.
The following table shows the reviews, ratings, and downloads for each one:
App | Google Play | Apple Appstore (Number of downloads not available) |
CamCard - Business Card Reader | Reviews: 149k Rating: 4.5 Downloads: 10 million+ | Reviews: 90.7k Rating: 4.7 |
Covve Scan | Reviews: 4.4k Rating: 4.7 Downloads: 100,000+ | Reviews: 1.2k Rating: 4.8 |
ABBYY Business Card Scanner | Reviews: 7k Rating: 3.9 Downloads: 1 million+ | Reviews: 19k Rating: 4.6 |
BizConnect - Business Card Scanner & Card Reader | Reviews: 3k Rating: 3.8 Downloads: 100,000+ | Reviews: 85 Rating: 4.7 |
*The data was updated on June 17, 2022.
Disclaimer:
Our list of apps is not exhaustive. There are plenty of scanner apps available on the market, and we encourage you to carry out any further tests via our methodology, which is disclosed later on in this report.
The tests we conducted aim to provide detailed insights on the accuracy, UX (user experience), and other factors that are taken into consideration when deciding which card scanner app to use. We disclose all of our testing methods and results, in full transparency. All four apps have undergone the exact same tests.
This research was commissioned by Covve in an effort to better understand how their card scanner app could improve, and how it compares to the leading competitors in the market. Our testers were unaware that the study was commissioned by an organization during the time of testing, removing any potential bias during the testing process.
Business Card Characteristics
In total we tested 80 cards that varied in the following ways:
Orientation: The card orientation was either in landscape or portrait.
Language/Characters: The card language was either in Latin or non-Latin characters.
For Latin characters, the language was English or French.
For non-Latin characters, the language was Greek.
Layout: Each card had a unique layout. Some featured corporate logos, others had two addresses or even a mix of languages for that matter.
Out of the 80 cards, 69 cards were in landscape and 11 were in portrait orientation.
To add, we also tested some of the most challenging card layouts, for a good reason. Text detection, segmentation, and recognition can be heavily influenced by factors such as the types of fonts, images, and even the design of the card. This has been previously highlighted in research such as “Text Recognition in the Wild” by Chen et al.
Ultimately, considering that each card scanner application recognizes a variety of different languages, we wanted to truly test the multilingualism of each one. According to each application’s website:
For this reason, we increased the complexity in the types of alphabets we were testing, focusing on European and North American markets.
All 11 portrait cards were in English
Out of the 69 landscape cards:
37 in English
19 were in non-English Latin alphabet (French)
13 were in the non-Latin alphabet (Greek)
Testing Methodology
In total, we thoroughly examined more than 5,500 fields, three times. One with the Levenshtein distance test, and twice with the Fairness Test.
Our equipment was extremely simple. A 2021 smartphone. The same equipment anyone would use to scan a business card at a meeting or a networking event. We went as far to scan each card by holding it at a 90-degree and a 45-degree angle, only to include the best result.
Our testing process consisted of 7 steps:
1) Transcribing the original text: The text from each card was manually transcribed. We then counted the character length for each field, spaces included.
2) Scanning the cards: Each card was scanned with all four apps. The scanned output was recorded for comparison.
3) The Levenshtein Distance test: We then compared the transcribed text against the scanned output from each app and assigned an L-Score for each field. Allow us to explain:
a) The L-Score refers to the Levenshtein Distance scoring. By definition: “The Levenshtein distance is a string metric for measuring the difference between two sequences. Informally, the Levenshtein distance between two words is the minimum number of single-character edits (i.e. insertions, deletions or substitutions) required to change one word into the other.” In a nutshell, we use this algorithm to determine if the scanned output differs from what is written on the card.
b) Let’s look at an imaginary example:
i) The email on a business card is busyperson@superbuzycorp.xyz ← This is the original text we transcribed. It consists of 28 characters.
ii) The scanned output from the apps was:
App 1: busyperson@superbuzycorp.xyz
App 2: busypeon@superbuzycorp.xyz
App 3: busyperson@superbuzycorp.x
App 4: busypon@supeuzycorp.xyz
iii) In this manner the L-Score in each case was:
App 1: 0 (since there was no distance between the original text and the scanned output)
App 2: 2
App 3: 2
App 4: 5
As you may have already guessed, based on the L-Score the most accurate output is from App 1, and the least accurate output is from App 4. You can try calculating the scores by yourself, here.
4) The Fairness test: Is the Levenshtein Distance test sufficient to determine the accuracy of the apps? Not really. That is the reason we subsequently introduced the “Fairness test”.
As mentioned above, the Levenshtein Distance detects differences between the input of two different fields and provides a score accordingly. The higher the L-Score, the greater the distance and the scan’s inaccuracy.
However, there are some drawbacks to this method, which we had to correct manually.
L-Score Unfairness: Sometimes the apps may slightly alter the scanned output, but without changing its true meaning. They may remove a comma, a space or provide the text in capital letters. Although the text or number may appear differently, the output may still be valid. For example, let’s say a telephone number on a card is provided in the following format: +1 (0) 555 666 777. When scanned, the output may be recorded as: +1(0)555666777. It is effectively the same number, just formatted differently. In this case, the scan will receive an L-Score of 4. This is unfair since the output is correct. In this case, we change the L-Score (or Levenshtein distance) back to zero. Hence the scan is 100% accurate.
Accuracy Unfairness: The scanned output may make no sense, yet receive a higher accuracy score than it should. For example, let's say someone’s name is Michael Johnson. Yet the scan returns “Hael Jnson”. The Levenshtein Distance test will automatically assign a score of 6 or an accuracy score of 60% (see below for how we calculate the accuracy score). This is unfair since the accuracy is actually 0%. The name that was returned from the scan did not have anything to do with the actual name on the card. In such cases, we set the accuracy score to 0%.
5) Determining the accuracy score per field: Once the L-Scores had been revised, we proceeded to calculate the accuracy of each scanned output.
a) The accuracy was determined via a simple formula:
[1-(L-Score of the field / Number of Characters of the field)] x 100
b) Therefore given that the email in our previous example consisted of 28 characters, the accuracy scores for each app are:
App 1: 100% accurate → [1-(0/28)]x100
App 2: 92.85% accurate
App 3: 92.85% accurate
App 4: 82.14% accurate
c) In case a field was invalid, the accuracy score was automatically set to zero percent. When is an output invalid?
i. If the L-Score was larger than the original text count. This occurs when the scanned output contains more characters than the original text, due to a processing error where an app attempts to automatically correct errors in the scan and ends up outputting lengthier words or numbers. Therefore if the numerator was larger than the denominator, and the output was negative. For example:
L-Score = 20
Character Count = 10
[1-(20/10)]x100 = -100% ← The output is invalid
ii. If the output made no sense, as previously mentioned in step 4.
6) Score weighting: Each piece of information carries a different level of importance. For this reason, we clustered the data in three categories: The Fundamental fields, the Extended contact details, and Non-Contact Information. Each category carried a different weight:
Fundamental fields (weight of x4): The most important data you need in order to contact a person such as their name, at least one email address, and at least a phone number.
Extended Contact Details (weight of x3): Every other contact information card provides, including second or third email addresses and telephone numbers.
Non-Contact Information (weight of x1): Any non-contact information, such as the organization name, job title, physical address, and websites. This information is still necessary but does not restrict a user from reaching out to a contact if they have the information provided in the other two categories.
7) Calculating the final scores: Last but not least, the final score for each app was calculated in the following steps:
We calculated the accuracy score for every field, on each card.
We then calculated the average for each of the three clusters (Fundamental Fields, Extended Contact Details, and Non-Contact Information) and produced a weighted average of all three, using the weight values we mentioned earlier. This is a sample of our table:
Card Info | Field 1 | Field (n.) | Fundam. Fields | Extent. Contact | Non-Contact Information | Weighted Average |
Card 1 | Accuracy% | Accuracy% | Average% | Average% | Average% | Weigh.Av.% |
Card 2 | Accuracy% | Accuracy% | Average% | Average% | Average% | Weigh.Av.% |
Card (n.) | Accuracy% | Accuracy% | Average% | Average% | Average% | Weigh.Av.% |
| | | Total Average | Total Average | Total Average | Total Weighted Average |
This process was repeated for four groups, providing the accuracy of each app in scanning:
Cards in the English language
Cards in the French language
Cards in the Greek language
Cards in all languages
The results follow in the next section.
Results: CamCard vs ABBYY vs Covve Scan vs BizConnect
Accuracy & Reliability
The results consist of the following categories.
The overall accuracy
The accuracy in scanning the Fundamental Fields, Extended Contact Details and Non-Contact Information
The reliability index shows the percentage of times an app managed to scan the fundamental fields from a card with 100% accuracy. Therefore represents how reliably an app can record the very essential information of a contact without the need of the user to correct them.
Please note that as we did not detect Greek among the languages CamCard supports, we provided two overall accuracy scores for the app:
One including the scanned cards in the Greek language
One excluding the scanned cards in the Greek language, which is the score we kept.
We used BizConnect’s post-verification results for our analysis. See below for more details on what we mean by post-verification results.
Latin Characters - English Language Results
| Overall Accuracy | Reliability Index | Fundamental Fields | Extended Contact Details | Non-Contact Information |
Covve Scan | 97% | 96% | 99% | 96% | 92% |
BizConnect | 95% | 93% | 98% | 93% | 89% |
ABBYY | 75% | 48% | 81% | 72% | 59% |
CamCard | 79% | 57% | 82% | 78% | 68% |
Latin Characters - French Language Results
| Overall Accuracy | Reliability Index | Fundamental Fields | Extended Contact Details | Non-Contact Information |
Covve Scan | 95% | 95% | 97% | 94% | 90% |
BizConnect | 92% | 95% | 97% | 87% | 84% |
ABBYY | 69% | 47% | 77% | 61% | 53% |
CamCard | 55% | 26% | 59% | 52% | 43% |
Non-Latin Characters - Greek Language Results
| Overall Accuracy | Reliability Index | Fundamental Fields | Extended Contact Details | Non-Contact Information |
Covve Scan | 90% | 87% | 98% | 87% | 67% |
BizConnect | 64% | 23% | 66% | 73% | 43% |
ABBYY | 39% | 15% | 44% | 45% | 8% |
CamCard | 48% | 15% | 52% | 60% | 4% |
*CamCard is excluded from this comparison, as it does not support the Greek language.
Overall Accuracy Results
| Overall Accuracy | Reliability Index | Fundamental Fields | Extended Contact Details | Non-Contact Information |
Covve Scan | 96% | 96% | 99% | 94% | 87% |
BizConnect | 88% | 80% | 91% | 87% | 79% |
ABBYY | 66% | 41% | 72% | 63% | 48% |
CamCard | 72% | 48% | 75% | 70% | 61% |
CamCard *including the scans in the Greek Language | 67% | 42% | 71% | 66% | 50% |
*CamCard’s overall accuracy results exclude the scores from the cards that were scanned in the Greek language, as shown in the last row of the table.
Highlights
In the English and French language categories:
Covve Scan and BizConnect are head to head for the first position, and ABBYY is competing with CamCard for the third position. Covve Scan comes first in both categories with a slight difference from BizConnect.
Camcard scores the 3rd position in the English language scans, and ABBYY comes 3rd in the French language scans.
In the Greek language Categories:
Although all apps can accurately read in a good amount of languages, Covve Scan has clear superiority over the Greek language scans. It scores an overall accuracy of 90%, whereas BizConnect which holds the 2nd position scores only 64%. ABBYY scores below 50%, while CamCard is excluded from the test.
In the overall accuracy category, which includes all languages:
Covve Scan comes first with a 96% score in its overall accuracy and reliability, and a 99% accuracy score for the fundamental fields.
BizConnect comes second, with an accuracy score close to 90%. It is worth noting that BizConnect is conducting post-scanning verification. In other words, once a card is scanned, BizConnect verifies the correctness of the fields and changes any omissions. This verification may take several minutes, as you will read in the User Section that follows. The accuracy score includes this post-scanning verification.
CamCard comes 3rd with a 72% overall accuracy score and ABBYY 4th with a score of 66%.
Accuracy & Reliability Winner: Covve Scan - The clear winner in the accuracy and reliability category is Covve Scan. However, the battle does not end here. Let’s continue our comparison, in the areas of User Experience, Feature Versatility, and Pricing.
User Experience
The user experience assessment of each app consists of two key factors. The ease of use and the versatility of features.
Ease of Use
To determine the ease of use, we focused on the number of steps required to scan a card and the time required to complete the process.
The number of steps required to complete the process
Covve Scan
Detect/adjust the edges of the card
Take a photo of the card
Confirm the scan or edit the edges / retake the photo
The information is scanned, presented, and saved
BizConnect Scan
Detect/adjust the edges of the card
Take a photo of the card
Confirm the scan or edit the edges / retake the photo
The user has the option to add location, tasks, groups or activity
The card is saved and sent for the verification
The user is prompted if the card is verified or not
ABBYY Scan
Select auto capture to detect the edges or manually select the edges yourself
Detect/adjust the edges of the card
Take a photo of the card
Confirm the scan or edit the edges / retake the photo
The information is scanned and presented
CamCard Scan
Detect/adjust the edges of the card
Take a photo of the card
The information is scanned and presented
The user is prompted to save or edit the information
Time required to complete the process
Speed is one of the most important factors that can make or break UX (user experience). According to UX Planet, in 2009 Google traced that slower loading times by 100 to 400 milliseconds, could lead a user to perform 0.6% fewer searches.
During our scanning process, we recorded the time it would take to scan a card with each app. We did so for 20 cards and produced the average time to scan for each app. Here are the results:
The total time in processing the scan:
Average recorded time to process the scan (in seconds) | |
Covve Scan | 2.12 |
CamCard | 3.86 |
BizConnect | 7.92 |
ABBYY | 6.21 |
The total time in detecting the card, adjusting the edges, taking the photograph, processing the results, and saving the scan:
Average time to complete the process (in seconds) | |
Covve Scan | 7.51 |
CamCard | 10.54 |
BizConnect | 11.93 (Not including verification time) |
ABBYY | 15.40 |
Overall Speed Winner: Covve Scan - Although the steps were mostly the same for all apps, Covve Scan was significantly faster in completing the process and in providing the final results. We do not discount the fact that ABBYY and Covve Scan prompt the user to double-check the scan before processing it, nor the fact that BizConnect runs a post-scan verification. Both elements are important in evaluating the accuracy of the scan. Nonetheless, we look at the practicality and the purpose of each app. In a networking event, you need to scan and store a card almost as fast as you would exchange it in its physical form. You do not always have the luxury of time to stop and correct things in a scan, nor to wait for the app to correct its mistakes. In the latter case, Bizconnect sometimes took up to 20 minutes to verify the validity of a card, and on different occasions, it failed. For this reason speed and accuracy go hand in hand. It is exactly what an effective networker is looking for when exchanging cards digitally. Hence, our winner in this category is Covve Scan.
Versatility of Features
In the following section, we explore the features that we liked or disliked in each app. This assessment is subjective to the tastes of our testers. We acknowledge that each user’s preferences may vary.
Covve Scan
Liked | Disliked |
Allows the users to export their contacts in excel, google, and outlook for free. | Does not always automatically recognize the edges on a card, so the user has to select the edges manually. Especially when the card is held at a 45-degree angle. |
Offers seamless experience, the user can start scanning without having to sign up. | Limits the user's ability to scan more than 10 cards, unless they purchase a subscription. |
Allows tags and notes in order to remember additional details about the person and group them as you need (e.g. client, partner, lead). | |
Integration with Covve’s wider ecosystem of apps and pCRM. | |
BizConnect
Liked | Disliked |
Offers a post-scan correction process, where the inputted text is reviewed and several errors from the scan are corrected. This does take a few minutes to complete, however it is worth it, as it improves the accuracy. | If the post-scan correction process fails, the user receives an email that explains the reasons behind the ‘transcription failure’. This can be an advantage, but feels a bit unnecessary. |
Very fast and accurate in recognizing the card’s edges automatically. | The user is not able to view the entire text on a field, if the text is too long. You have to edit the card to view the entire string of text. |
Adds map hyperlinks for scanned addresses, which can save the user a lot of time from copying and pasting the address in a map app. | |
ABBYY
Liked | Disliked |
Offers good card editing features. The app magnifies the field you are editing, so you can detect minor errors, like an extra dot in an email and more. | Difficulty in scanning the card edges. |
Allows the user to scan documents, as well as cards. This may come in handy when scanning brochures or leaflets at conferences. | Issues in taxonomizing the input to the right fields. |
It can scan multiple cards simultaneously. | On several occasions, it automatically assumed the country that a business operated in. For example, a business operating in the UAE, Russia, and Greece. The card holder’s office was in Greece, however, ABBYY added Russia as the country of choice. |
Also allows the users to export their contacts in excel, google, and outlook for free. | Performed poorly when scanning the non-English languages we tested. |
| Asks permission to use the user location, which feels unnecessary for a card scanner app. |
CamCard
Liked | Disliked |
Allows the user to scan up to 500 cards for free. | Asks for access to the user’s contacts. This is unnecessary. It would make sense to ask for access if the user wanted a pCRM app. The primary purpose of the app in this case though, is to operate as a scanner first. |
Prompts the user to add the backside of the card, and verifies if the information is correct before storing the card. | Hides long strings of text, such as the name, which can be truly frustrating. |
Allows the scanning to be performed in batches not just single cards. This is extremely useful. | Performed poorly when scanning the non-English languages we tested. |
Very fast and accurate in recognizing the card’s edges automatically. | Has the tendency to split a single address into two different addresses. Indeed, this drops the accuracy of the scan, but it also makes it really hard for the user to correct the error or copy the address on a map application. |
| Performs poorly on complex card layouts. |
| Issues in taxonomizing the input to the right fields. |
Versatility of Features Winner: CamCard - We were impressed by CamCard’s generosity in providing 500 card scans in its free package, its ability to automatically detect the card edges faster than any other application, and its capability in scanning the cards in batches, which is also offered by ABBYY.
With the above in mind let’s move on to compare the pricing for each application.
Pricing
All four apps are available to download and sample for free. However, they offer a number of subscription options.
Pricing (as presented in the app, excluding discounts) | |
CamCard |
|
Covve Scan |
|
ABBYY |
|
BizConnect |
|
*The data was collected in June 2022
Pricing Winner: Covve Scan - Covve Scan provides the lowest fee at $29.99 per year, and offers a lifetime premium for $44.99.
It is worth noting that each app offers different features and capabilities, via their different subscription packages.
Last but not least we review the privacy and security provision of each application.
Privacy & Security (Based on the claims of the app publishers)
| DPA (Readily Available) | Publishes Declares GDPR Compliance | Other Security Certifications |
Covve Scan | ✓ | ✓ | |
BizConnect | ✓ | ✓ | |
ABBYY | ✓ | ✓ | SOC 2, HIPAA |
CamCard | | ✓ | |
Privacy & Security Winner: ABBYY - Based on the information available on the publishers' websites and in the app stores, all apps declare that they comply with GDPR regulations and use 256-bit data encryption. We were unable to locate the CamCard’s DPA (Data Processing Agreement). We also note CamCard’s ban in India, alongside other Chinese apps, following cybersecurity concerns by the Indian government. We award ABBYY as this category’s winner due to its security accreditations.
Discussion of Results
Our overall winner is Covve Scan. The app scans with the highest accuracy, is the most reliable of the four, offers the fastest processing time, has the most seamless experience, and comes at the best price. It also carries the highest ratings on Google Play and the Apple AppStore.
With that being said, Covve Scan is the new kid on the block, and although very promising in becoming a significant player, it can learn a great deal from the strategies and features of its competitors.
CamCard is the dominant market app at the moment, and there is a reason for it. Offering 500 card scans for free is actually an ingenious and extremely attractive market penetration and retention strategy to keep your competitors at bay.
Given that these scanner apps are designed to process and store large amounts of personal information, meeting the legal requirements on privacy and data security is a must. Each app must do as much as possible to keep its users and their data secure. Let’s not forget that with a card scanner app, you are not processing your data, but the data of other individuals as well. Moreover, ABBYY’s compliance and security certifications of SOC2 and HIPAA make it stand out from the competition
Additionally, BizConnect’s post-scanning verification process is impressive but can take some time to complete. We experienced the results before and after the corrections derived from the verification process, and the improvement in the results was evident. Nonetheless, it goes without saying that since time matters in networking and meetings, the speed of the verification process is important. Covve Scan’s AI stands out in this respect, as it is robust and returns instant results.
As suggested at the beginning of this report, you are encouraged to conduct your own research and tests before making a decision. We truly hope that this analysis has aided your decision in selecting a card scanner for your use or by your organization.
For any inquiries contact us via our homepage at www.capacitorpartners.com
コメント