Dynamic checkout page optimization using machine-learned model

1. A method for processing payments made via an electronic payment processing system, the method comprising:

obtaining training data from a data source, the training data relating to prior purchases made via the electronic payment processing system, the data source including, for each of the prior purchases, a checkout page of a purchase transaction presented by a device user interface of the electronic payment system in a check out flow of each of the prior purchases;

extracting one or more features from the training data relating to the prior purchases, the one or more features associated with a negative user action invoked in relation to at least one of the prior purchases;

for a real-time purchase not included in the training data, the real-time purchase transacted at a real-time device user interface rendered by the electronic payment system presenting a real-time check out page in a check out flow, feeding the one or more extracted features into a transactional behavior model, the transactional behavior model trained via a machine learning algorithm to produce a transactional behavior score indicative of a probability that the real-time purchase will invoke a negative user action taken by the user pursuant to a conclusion of the real-time purchase;

for the real-time purchase, and based on the transactional behavior score and user data entered during the real-time purchase, causing a dynamic optimization, during the check out flow, of the real-time device user interface presenting the checkout page, the dynamic optimization of the real-time device user interface including insertion in the real-time check out page of a targeted remedial action, specific to the user, to reduce a probability that the real-time purchase will invoke a negative user action taken pursuant to a conclusion of the real-time purchase;

receiving, via the real-time device user interface, a request to invoke the remedial action; and in response to receiving the request to invoke the remedial action, triggering the remedial action to be invoked.

2. The method of claim 1 , wherein the data source is confined to data contained in or associated with the checkout page in the purchase transaction funnel.

3. The method of claim 2 , wherein the invoked negative user action includes one or more of a refund request, a chargeback request, and a return request.

4. The method of claim 3 , further comprising confining the training data to information extracted from the checkout page and associated with the invoked negative user action.

5. The method of claim 1 , wherein the training data includes a data structure, the data structure including a first user interaction section and a checkout page data section.

6. The method of claim 5 , wherein the first user interaction section includes data sourced exclusively from the checkout page and relating to one or more of:

a time period between a loading of the checkout page and a taking or completion of a payment action;

a number of times or frequency a customer viewed the checkout page before taking or completing a payment action;

a number of typos or other mis-entries corrected prior to taking or completing a payment action;

a detection of an omitted field completed prior to taking or completing a payment action;

a number, type, or frequency of a mouse movement;

a detection of a payment denial;

a detection of a decline of a payment instrument;

a detection of a substitution of the payment instrument;

an IP address associated with a prior purchase or the real-time purchase; and

a local user time of the prior purchase or the real-time purchase.

7. The method of claim 5 , wherein the checkout page data section includes data sourced exclusively from the checkout page and relating to one or more of:

a detection of a first-time customer;

a detection of a repeat customer;

a time period of an interval between a prior purchase and the real-time purchase, assessed relative to a customer average for the same interval;

a user cart size associated with the real-time purchase, assessed relative to an average user cart size for a purchase;

a locality or default currency of a payment instrument to give an indication of a customer's location relative to a merchant;

user order or shipping API data;

a selection of a shipping service or shipping rate;

embedded metadata associated with a user action;

a detection of a refund; and

a detection of a support ticket or request.

8. A system for processing electronic payments, the system comprising:

at least one processor; and

a memory storing instructions that, when executed by the at least one processor , cause the system to perform operations comprising, at least:

obtaining training data from a data source, the training data relating to prior purchases made via the electronic payment processing system, the data source including, for each of the prior purchases, a checkout page of a purchase transaction presented by a user interface of the electronic payment system in a check out flow of each of the prior purchases;

extracting one or more features from the training data relating to the prior purchases, the one or more features associated with a negative user action invoked in relation to at least one of the prior purchases;

for a real-time purchase not included in the training data, the real-time purchase transacted at a real-time user interface of the electronic payment system presenting a real-time check out page in a check out flow, feeding the one or more extracted features into a transactional behavior model, the transactional behavior model trained via a machine learning algorithm to produce a transactional behavior score indicative of a probability that the real-time purchase will invoke a negative user action taken by the user pursuant to a conclusion of the real-time purchase;

and

for the real-time purchase, and based on the transactional behavior score, causing a dynamic optimization, during the check out flow, of the real-time user interface presenting the checkout page, the dynamic optimization including insertion in the real-time check out page of a targeted remedial action, specific to the user, to reduce a probability that the real-time purchase will in fact invoke a negative user action taken pursuant to a conclusion of the real-time purchase.

9. The system of claim 8 , wherein the data source is confined to data contained in or associated with the checkout page in the purchase transaction funnel.

10. The system of claim 8 , wherein the invoked negative user action includes one or more of a refund request, a chargeback request, and a return request.

11. The system of claim 10 , wherein the operations further comprise confining the training data to information extracted from the checkout page and associated with the invoked negative user action.

12. The system of claim 8 , wherein the training data includes a data structure, the data structure including a first user interaction section and a checkout page data section.

13. The system of claim 12 , wherein the first user interaction section includes data sourced exclusively from the checkout page and relating to one or more of:

a time period between a loading of the checkout page and a taking or completion of a payment action;

a number of times or frequency a customer viewed the checkout page before taking or completing a payment action;

a number of typos or other mis-entries corrected prior to taking or completing a payment action;

a detection of an omitted field completed prior to taking or completing a payment action;

a number, type, or frequency of a mouse movement;

a detection of a payment denial;

a detection of a decline of a payment instrument;

a detection of a substitution of the payment instrument;

an IP address associated with a prior purchase or the real-time purchase; and

a local user time of the prior purchase or the real-time purchase.

14. The system of claim 12 , wherein the checkout page data section includes data sourced exclusively from the checkout page and relating to one or more of:

a detection of a first-time customer;

a detection of a repeat customer;

a time period of an interval between a prior purchase and the real-time purchase, assessed relative to a customer average for the same interval;

a user cart size associated with the real-time purchase, assessed relative to an average user cart size for a purchase;

a locality or default currency of a payment instrument to give an indication of a customer's location relative to a merchant;

user order or shipping API data;

a selection of a shipping service or shipping rate;

embedded metadata associated with a user action;

a detection of a refund; and

a detection of a support ticket or request.

15. A non-transitory machine-readable medium comprising instructions which, when read by a machine, cause the machine to perform operations for processing payments made via an electronic payment processing system, the operations comprising, at least:

obtaining training data from a data source, the training data relating to prior purchases made via the electronic payment processing system, the data source including, for each of the prior purchases, a checkout page of a purchase transaction presented by a user interface of the electronic payment system in a check out flow of each of the prior purchases;

extracting one or more features from the training data relating to the prior purchases, the one or more features associated with a negative user action invoked in relation to at least one of the prior purchases;

for a real-time purchase not included in the training data, the real-time purchase transacted at a real-time user interface of the electronic payment system presenting a real-time check out page in a check out flow, feeding the one or more extracted features into a transactional behavior model, the transactional behavior model trained via a machine learning algorithm to produce a transactional behavior score indicative of a probability that the real-time purchase will invoke a negative user action taken by the user pursuant to a conclusion of the real-time purchase;

and

for the real-time purchase, and based on the transactional behavior score, causing a dynamic optimization, during the check out flow, of the real-time user interface presenting the checkout page, the dynamic optimization including insertion in the real-time check out page of a targeted remedial action, specific to the user, to reduce a probability that the real-time purchase will in fact invoke a negative user action taken pursuant to a conclusion of the real-time purchase.

16. The medium of claim 15 , wherein the data source is confined to data contained in or associated with the checkout page in the purchase transaction funnel.

17. The medium of claim 15 , wherein the invoked negative user action includes one or more of a refund request, a chargeback request, and a return request.

18. The medium of claim 17 , wherein the operations further comprise confining the training data to information extracted from the checkout page and associated with the invoked negative user action.

19. The medium of claim 15 , wherein the training data includes a data structure, the data structure including a first user interaction section and a first checkout page data section.

20. The medium of claim 19 , wherein the first user interaction section includes data sourced exclusively from the checkout page and relating to one or more of:

a time period between a loading of the checkout page and a taking or completion of a payment action;

a number of times or frequency a customer viewed the checkout page before taking or completing a payment action;

a number of typos or other mis-entries corrected prior to taking or completing a payment action;

a detection of an omitted field completed prior to taking or completing a payment action;

a number, type, or frequency of a mouse movement;

a detection of a payment denial;

a detection of a decline of a payment instrument;

a detection of a substitution of the payment instrument;

an IP address associated with a prior purchase or the real-time purchase; and

a local user time of the prior purchase or the real-time purchase.