​
enc_U2FsdGVkX1+Zi//k7+PNW1WEmd9lIiCUbCRDYk/Uucw=_enc

​
enc_U2FsdGVkX1+fLGBopf8R7yxxS/Rtt/uM8IGvY+XijWY=_enc

The session scoring is done in three steps:
  • Get all the questions scores (see the “questions scoring” section) and compute the average score using each question coefficient ;
  • Review each question that has a score of 0 and that is marked as eliminatory, if such a question if found then the session final score is also set to zero ;
  • Apply the geolocation scoring and the watchlist scoring (see the corresponding sections) each of this scoring can set the session final score to zero.

​
enc_U2FsdGVkX1/iZzf3HxiCb49h8NSJYdGNY7G9Qk3Yb9o=_enc

Let’s take the following scenario:
  • Question 1 - Coefficient 4 - Say your first name out loud (type “verification”)
  • Question 2 - Eliminatory **- **Scan your Identity Document (type “identity”)
  • Question 3 - Eliminatory - Scan your face (type “face”)
  • Question 4 - Eliminatory - Say these 4 digits out loud (type “antibot”)
With the following scores:
  • Question 1 - 60%
  • Question 2 - 0%
  • Question 3 - 100%
  • Question 4 - 100%
To score the session step by step:
  • The average score including the coefficients is 63% ;
  • But we have one question with a score of 0 and that was eliminatory, so now the session score is 0% ;
  • We loot at the geolocation scoring and the watchlist scoring and apply the related actions (see the corresponding sections)
The session score is now 0% and if there is a label that the algorithm can set itself then a negative label will be set.

​
enc_U2FsdGVkX19r6L/ZiYFI2J9Bll4CLcNB6lIk+cVfYbQ3g6LovfePICszIebieaBa_enc

​
enc_U2FsdGVkX19KkgNLWXTrdqe3H2DAcATbd3QO9u94ax4=_enc

Each question has a specific scoring system. We provide this list of question types:
  • Identity: the identity document of the customer will be recorded.
  • Face: the face of the customer will be recorded.
  • Information: the customer is asked a question, the answer can be tested against a list of disallowed answers.
  • Verification: the customer will have to provide an information by voice that will be compared to another information retrieved during or provided before the session.
  • Antibot: a challenge has to be solved by the customer to ensure it is not a robot.
  • Upload: a document must be uploaded by the customer, nothing is scored here.
Note: Each question can be asked to the customer up to 3 times if something went wrong. After these 3 requests, the question score will be set to 0%.

​
enc_U2FsdGVkX1/eWXQbJ4v355aIj63QiVlBe1jwvYtcFD3TG2qmWuiJzN/ZIbF5uE04_enc

A list of checks are done on the identity document provided during the session. These checks can have 4 status: success (green), fail (red), consider (orange) and unknown (grey).
  • Issuing date: We check that the document issuance date isn’t in the future.
  • Expiration date: We check that the document has not expired.
  • Birth date: We check that the birth date is present, not in the future or too far in the past and is consistent across the document.
  • Minimum required age: Some identity documents cannot be owned under a certain age (driving license for instance), if the birth date doesn’t match the identity document requirements, this check will fail.
  • MRZ: The MRZ was found and the content was correct (checksums). To validate a MRZ we also check its font (monospace font OCR-B).
  • Face: A picture of a face was detected and this face respects the requirements of an identity document (orientation, no accessories, neutral face). We also check that the face gender and age match the information provided on the identity document, by default this check has no effect on the final result.
  • Specimen Detection: The document doesn’t appear to contain the specimen word in it nor contains any already known specimen identity document in our database.
  • Image Quality: The image quality is good (sharp, coloured, good dimensions) ;
Each of this check status can trigger or not a score to 0% depending on the scenario configuration. If none of these checks trigger a score to 0%, the question score will be the average of the scores of the match between the identity document first name, last name and birth date and the corresponding information provided by the customer before the session. Names match scores uses mainly the Soundex algorithm. You can learn more about Soundex here: https://en.wikipedia.org/wiki/Soundex. Birth date match is an equality match.

​
enc_U2FsdGVkX1+JBoedN8ZRquf1GHHuStJ3g7Qhq0GjYHOPX7sFrLgzGRFl2qNJsGVy_enc

Like on the Identity question scoring, a list of checks are done on the faces provided during the session. These checks can have 4 status: success (green), fail (red), consider (orange) and unknown (grey).
  • Face detected: A face was detected ;
  • Match reference: The detected face matches the reference face (the reference face is the face extracted from the identity document) ;
  • Not a known celebrity: Our face analysis provider can detect if the customer appear to be a known celebrity (only when using AWS Recognition service);
  • Quality: The image quality is good (sharp, coloured, good dimensions) ;
  • Liveness: The faces doesn’t appear to be a “stolen” image (obtained from the orientations of the face).
Each of these checks status can trigger or not a score to 0% depending on the scenario configuration. If none of these checks trigger a score to 0%, the question score will be the score of the face match between the reference face and the provided face.

​
enc_U2FsdGVkX18fHK5EvCEnxMCRn5SI4LG0czrCWFhoAsSNI48kHkOGGqpD2AGJ4TuV_enc

If the customer’s answer contains one of the disallowed words, the question score is set to 0. If not, the question score is set to 100%.

​
enc_U2FsdGVkX1/jVZWTlB9dT9RbSw+TMAmDoAscnmGhkSOsenJLUn6r4D5vNWLQE7DL_enc

If the verified item is a number (Age), we test if the customer’s answer is equal to the item. If not, the question score is set to 0%. If the verified item is a string (First name, Last name), we apply a Soundex algorithm to get the question score. You can learn more about Soundex here: https://en.wikipedia.org/wiki/Soundex .

​
enc_U2FsdGVkX1/10i8RccWW9b1CfLbQDPw6nfx57o16Nv8h0gDBn8dAzqLvfaR613Pw_enc

If the customer’s answer by voice matches the challenge (serie of numbers to spell for example), the question score is set to 100%. If not, a score of 0% is set.

​
enc_U2FsdGVkX1/rksfXb5t7vzGzuXlANV1dnmNPSA/L337fhA9epgZT6qu0Yx4HaQkE_enc

If a document is uploaded a score of 100% is set. If not, a score of 0% is set.

​
enc_U2FsdGVkX18jqMRHWsNguWMCVHizbJvm7qi7eus1cbG0GfiD2Fl9aDK2yvylGIYrd14geDRf2PdVchCiqcZKW/VF4R1Dd769s/yxDuyEbW8=_enc

​
enc_U2FsdGVkX1/Yj77RWNYZDGVD6hWcExWaSMldkIrxgdHRWEEJIXPvyyylh0alLPon_enc

enc_U2FsdGVkX1/ylBHqHOzpe5Dc5Sx1zH0vAqmCCEjbTebNrSnhKgIGN6Em1s/e9vd7yYoZ2o/jl9Oh0E5gZgAWMx6vOzIMChRcdEDZdzFVCBIRX1gPYKyJoseWFZwE3q1hTrBioUqfR/puGiVMeUBYjw==_enc enc_U2FsdGVkX1+OB1W0tHrhvfeEy3AG7+fBccpFNj02m23dLVt04lB+LTYfMt8f2T0xVwNtm8I1boDSRnOQSrIEEbugvxIoWiEMfsQK/o8n+FGdwPwsWwYSlLalkYYLoFgCo6IVAgNh2vu7ubLXcmVoSg==_enc
  • Precise geolocation was disabled: when the user refused to share the precise location provided by the browser at the beginning of the session.
  • IP country doesn’t match customer defined country: when the IP estimated country doesn’t match the customer defined country. We use a daily updated database to retrieve this information. We are not able to detect proxies.
  • IP location to geolocation max distance (in km): if the precise geolocation was enabled, we can ensure the user IP and the user current location are the same in this specified radius.
  • IP location to geolocation distance exceeded: when the previous point is false.

​
enc_U2FsdGVkX1+fSvsAQ0pnFAbCIbo+FTZjxNBMqPDsUa19loO04FY+7CRnUy95pDJm_enc

Depending on the Name Search result against our watchlists, we can either set the session status to 0%, do nothing, or add a flag to the session (no impact on the score). To know more about the Name Search algorithm, read the corresponding document. Go!Scan uses the default Name Search settings used in the “namesearch” feature (80% name matching for fuzzy match option).