payment data prediction at test time

I have the payment data of the client. I want to predict the prob of customers paying late with target classes being 0-30 days, 30-60 days, 60-90 days, and 90+ days based on this paper.

The features I have are as follows:

And at test time I have values as follows:

Data Description:-

1. Amount: Amount to be paid
2. Payment terms: a payment that needs to be done under the specified date
3. Diff in days: Difference between clearing date and invoice date.
4. Paid invoices bef order: Number of paid invoices before new order was created in next data 5. point
5. Paid invoices late: Number of paid invoices that were late
6. Ration of paid in which were late: paid invoices late/Paid invoices bef order
7. Sum of inv bef order: sum of invoices before new order creation
8. Sum inv late: sum of invoices that were late before new order was created
9. Ratio of outs inv: Ratio of 8/7
10. Avg days late outs: Average days late of all outstanding invoices that were late prior to a new invoice for a customer

the invoice information at test time doesn't have a clearing date hence the 'diff in days' and 'avg days late outs' column has nans.

How can I feature my test data if I don't know the clearing date? And how can i predict a future datapoint when I don't know such information which can help me in engineering features?