This project was completed in middle 2016.  If an individual user wants to try Azure in China, he is required to fill out a form which must include his name and ID number.  He must upload his ID card (both front and back side) to prove himself a citizen of China.  This process is called “Real name check”.   For an enterprise application, users are required to fill in their company’s name and Organization Code and upload their business license instead.

Challenges

• The poor recognition rate of APIX. APIX is a third-party tool, which can recognize the text and numbers in pictures. Our purpose was to use it to discern the name and ID number on the ID cards that users upload. Before using this tool, all the review work had to be done manually so that users were required to wait at least 24 hours after they submitted the trial form successfully.  Utilization of a successful tool should help us reduce labor costs.  The stated recognition rate on APIX’s official website was said to be up to 98%.  However, when we tested it with the ID cards users had previously uploaded, the recognition rate was only around 50%.   Further, the tool had been bought from another collaborative team before the test results. Following multiple conversations, it was determined that it was one of the best tools in the market.  Other competitive tools had been tested and produced even lower rates.  Additionally, CMBC, a bank, was using this tool.  We needed to decide if we were going to use it at all, and, if so, how to use it properly.  
• Complex scenarios. Below are several general scenarios in the trial process:
  1. New individual users apply via the trial for the first time and successfully pass all the procedures.
  2. New individual users who don’t pass APIX but submit the form successfully;  new enterprise users who submit the form successfully.  Both of these situations have to wait for the manual check. They can come again during the meantime.
  3. Those who don’t pass the manual check and have to re-upload the images.
  4. Users who have come more than once.
  5. PIA (Pay-in-advance) customers who’d like to try.
  6. API outage.

Solution

• First, the idea of using recognition instead of users’ filling evaporated.  After more testing, we discovered that the recognition rate mainly depended on the quality of the picture image.  For that reason, I put the requirements for the image directly next to the uploading areas along with samples.  Recognition was also influenced by minority ID cards as their language was on their cards.  The manual check process is still provided for those who upload their cards correctly but whose cards are not recognized by APIX.   This is a transparent process for these users.  They can also try again and submit successfully.  Finally, users can use the minimal functional image editor to rotate or crop their image in order to meet the requirements.

Real name check process for trial user

See the details →

Image editor

See the details →

Hi-fi Mockup on PC

Default

Form-fill

Error

Image-edit

Mobile

Enable taking a photo on mobile so that users choose to go to the photo album or take a photo directly after clicking the upload button. See in the mockup below. Use hints to remind users of aligning with the frame.