With regards to the availability (GL 2), we believe that during first period post initial roll-out of the dedicated interface, provisions should be made to allow for larger planned downtime in comparison to PSU, if it is scheduled for off-peak hours of interface usage. This will allow ASPSPs to roll out improvements to their dedicated interfaces faster, leading to overall improvement in interface quality and functionality.
With regards to measurements of performance (GL 2.3), we believe the guideline should mention that performance KPIs are measured at the API/PSU endpoint on ASPSP network, to ensure network latency that is beyond control of an ASPSP does not affect the measurements.
With regards to publication of statistics of performance (GL 3), we believe that an overall maximum timeframe for publication of quarterly statistics should be included in the guideline (for instance, 1 month after end of quarter).
With regards to publication of performance KPI (GL 3), we believe that more clarity on how to publish statistics of performance KPI. We suggest that the guideline defines the published values with more precision. Our suggestion is to have a KPI in form of 90-th percentile of response time per each day (24 hours) per each API.
We believe that current guidelines for stress testing are too vague, operating in terms such as unusually high" or "multiple". As, per our understanding, the underlying reasoning for this guideline is to ensure ASPSPs keep their interfaces robust and able to cope with peak loads and support future growth, we think the stress test definition should be adjusted per size of ASPSP business. Our suggestion is to guide national authorities to define figures of stress testing (such as concurrent access, high number of requests, large volumes of data) in terms of a multiple of ASPSP current number of accounts/payment transactions/other applicable existing business KPIs. ASPSPs could include their baseline figures in official reports they submit to competent authority. We also believe stress tests should be periodical. For example: ASPSPs should test 5 times the maximum number of concurrent requests they had experienced via PSU during last year. In case actual figures approach the tested threshold, or after 1 year since last stress test, ASPSPs should perform a new test with adjusted figures (5 times the maximum number of concurrent requests over dedicated interface or PSU, the larger of both)."
We fully agree with EBA's assessments on monitoring.
We agree with existing obstacles. We believe that in addition to them, it is important to ensure that no commercial obstacles are imposed for operations performed via the interface, such as unreasonably high or discriminatory pricing per transaction for a particular institution.
We agree with EBA's assessment for design and testing. Since some market initiative standards consider definition of their own certification and compliance process, we suggest that EBA guides competent authorities to pre-approve market initiative certification plans and reports. With this approach, an ASPSP can simply submit a standard-compliant certification testing report, which will encompass both deviations from the standard (GL 6.4b), coverage of abilities mentioned in GL 6.2 and disclosure of weaknesses mentioned in GL 6.3
We agree with GL 7.1 but are concerned that GL7.2 can put unreasonable demand on communication of interface via channels such as trade bodies, conferences and market actors. We suggest to limit the list of communication channels to web site and social media. GL 6.1 already prescribes publication of interface specifications on ASPSP web site, and this is where all parties interested to integrate with the interface will search for details on it. Hence, we believe it is reasonable to publish details on availability of testing facilities on the website and expect it will reach widest relevant audience.
We fully agree with EBA's assessments and Guideline 8.
We fully agree with EBA's Guideline 9 and the Assessment Form.
We do not have any particular concerns.
We agree with the level of details, except for specific points mentioned in responses to questions above.