Getting Started with Smartscan Feedback

What is a Feedback ID?¶

Every time Smartscan processes a document, it returns a unique Feedback ID. When your user reviews or corrects the data, you can send that final, confirmed information back to us using this ID. Instead of just sending us the corrected suggestions, you send the final version of the data - whether the user approved Smartscan’s suggestion or manually adjusted it. This data is the fuel for a more intelligent integration.

Why implement the Feedback Loop? When you share the "full truth" (the final values confirmed by your end-users), you're providing the data we need to:

Power your Quality Metrics Dashboard.: We transform your feedback into real-time stats, showing you exactly how Smartscan performs for your specific users and fields, as perceived by the users.
Improve model accuracy: We use some of these insights to train the models on your specific documents. While not an automated loop, it allows our team to proactively improve your data extraction quality.
Understand user intent: Seeing how your users interact with our suggestions helps us align our AI with their actual goals.

How it works¶

Feedback can be submitted as soon as a document is successfully annotated, but you have time to provide feedback on a scan within 90 days. This timeframe is designed to capture timely insights and enhance the efficiency of our feedback loop.

Capture the feedback ID: Every successful annotation returns a unique identifier, known as the feedback_id. We can use the Feedback IDs to look at particular documents.
Send the Truth: Within 90 days, send a request with the final values reviewed by your end-users.
See the Results: View your automation rates and field-by-field accuracy on your Quality Metrics Dashboard.

To prevent a 404 error, ensure you aren't sending feedback for the same document twice, or attempting to send it after the 90-day limit.

Privacy & Data Control¶

Data retention: Data is kept for 365 days to assist in model training. So we handle a rolling 12-month deletion automatically, but the API ensures you can meet your own specific GDPR commitments to your users as described below.
Batch delete by tags: By including tags (like a unique customer ID) in your request, you can call a delete on that tag at any time to remove user data immediately. We highly recommend using tags when you send feedback. This allows you to call our delete endpoint at any time, perfect for batch deleting data for offboarded customers.
Deletion Policy: Check our Data Deletion Policy for details.

Error handling¶

To ensure a smooth feedback process, it's crucial to be aware of potential errors:

404 Not Found Error: Attempting to provide feedback after the 90-day window will result in a 404 Not Found error. Users should adhere to the recommended timeframe for optimal engagement.
409 Conflict Error: Rapidly submitting multiple feedback calls for the same document may result in a 409 Conflict error for some calls. Subsequently, this may lead to 404 errors once the initial call is complete and temporary stored data is removed.

Example request¶

POST /v1/feedback:create

https://api.stag.ssn.visma.ai/v1/feedback:create

Authorization - Bearer Token

Token: demo

Body - raw (json)

Body

{
    "id": "65226d1b-1c70-425c-9288-a29106928ef2",
    "tags": [
        "foo",
        "bar",
        "baz",
        "qux"
    ],
    "true_values": {
        "currency": "DKK",
        "document_number": "1202942",
        "document_date": {
            "day": 1,
            "month": 1,
            "year": 2019
        },
        "payment_due_date": {
            "day": 1,
            "month": 1,
            "year": 2019
        },
        "total_excl_vat": 225.0,
        "total_incl_vat": 300.0,
        "total_vat": 75.0
    }
}