Page 39 - ITU-T Focus Group Digital Financial Services – Consumer Experience and Protection

ITU-T Focus Group Digital Financial Services
Consumer Experience and Protection

NOTE: These KPIs and their technical basis are currently not standardized and therefore cannot be assessed
in a comparative manner.

This list clearly contains elements which are not primarily related to mobile network behaviour or performance;
they also relate to the performance of underlying banking processes and implementations. So, the list can
probably be reduced to elements which are assumed to be primarily linked to mobile networks.

There is, however, a connection. If, for example, a connection loss occurs during a transaction consisting of
a number of roundtrips estimated to complete a DFS transaction, this may have different results depending
on a particular implementation of such banking processes. Therefore, it is assumed that the robustness and
stability of such processes against failures which are typical to specific basic services of mobile networks will
also have an effect on overall QoS of DFS.

5.2 Technological components of DFS

As outlined in other parts of this document, there are some services and functionalities within existing mobile
networks which can be used – with a further selection by available features of mobile devices - to realize DFS.

From the concept of a “pyramid of needs” and assessment of the end to end KPI for DFS, a clear hierarchy of
quality requirements can be derived.
The topmost requirement will be the integrity of a transaction. Integrity in DFS is the clear and reliable
assessment if a transaction has been successful or not. This is seen as even more important as the overall
success rate of an implementation. If a transaction is erroneously assessed as being successful or failed, the
objective damage (e.g. to a person’s financial condition) will be larger than a case where a transaction has
to be repeated due to a detected failure. The same applies to a transaction which is erroneously assessed as
unsuccessful, which would result in duplicate transfer due to a repetition of the process.

From a QoE point of view, the situation can be more complex. Assumed there are two implementations, one
of them being stable and robust in the sense of low (ideally zero) probability of false positives or negatives,
but slow; the other one faster but more sensitive to such errors. Unless the false-assessment error will be
quite large, it is likely that in the customer perception, the latter will appears as the “better” one. It follows
that in this area, considerations beyond a mere competition according to market rules need to be undertaken.

An end to end approach needs to be taken because the overall robustness of a particular implementation
depends on several factors.
Assume that there are two alternatives, one of them requiring N1 roundtrips, each having a time duration of
T1, and a success rate per roundtrip of S1; the other one characterized likewise by characteristics N2, T2 and
S2. Clearly, there are several interactions with typical network properties. For instance, if the transaction is
performed while the actor is moving (e.g. in a public transport vehicle or as a passenger in a car), the change of
network conditions during a transaction influences the overall success rate. This links the time scale of motion-
related impairments to transaction characteristics. If the typical overall duration of a DFS transaction (T1*N1
and T2*N2) is above the typical time during which network properties show degradations, the probability of
failure increases. In a more general view, the overall success rate of a DFS transaction can be expressed as S1N1
and S2N2. So even if an individual success rate per roundtrip of a specific implementation (where the motion
profile can be factored in) is lower, the resulting E2E success rate may be higher if the number of roundtrips
in this implementation is sufficiently smaller.

The same linkage between characteristics includes the times involved. For instance, if a transaction fails (in a
“proper” way, i.e. with correct assessment of the result), the negative impact on QoE will assumedly be smaller
if this result is obtained in a shorter period of time, as a follow-up try can be started and completed faster.

31