Page 18 - Methodology for measurement of Quality of Service (QoS) Key Performance Performance
Indicators (KPIs) for Digital Financial Services
P. 18
b) The expected response may not occur. This essentially the point of observation (POO). For a POO on the A
a matter of time-out condition. Without additional in- Party side of a use case implementation, action refers
formation, the A Party cannot determine if the request — to an activity performed on the A side (by human ac-
the data sent to the service — or the response of the tion or some programmed activity) while event refers
service has been lost. to something incoming (e.g. a message received via a
mobile network).
If there is, in a particular implementation of a test or
the DF service, no sending failure information on the A ■ NOTE: In older literature, the term PCO was used
side, case a) cannot be technically distinguished and all (Point of Control and Observation). The newer term
interruptions appear to be of type b). POO reflects the fact that in most cases, respective
In any case, the A side has information of the last data comes from sources which do not allow control
successful step, and the next one attempted. In case of anyway (e.g. IP layer traces); also in general it is bet-
failure, this information can be output together with the ter to not intermix control and data layers.
failure information and used in subsequent processing.
Trigger Point ID (TPID) = <Service and use case code> _<Type>_<Index>
4.1.5 Time corrections for human interaction where
If interactions require human input, time measure- <Service and use case code>: in the present document, always DFSP2P
ments will need adjustments. The top-level phase for
set-up (see Top-level phases) consists, as shown in <Type> is either
Event and action flow, of a series of prompts for infor- - AE event observable on the A side.
mation items, and respective input by the user. There- - AA action to be performed by the user on the A side.
fore, a time measurement for the whole set-up phase - BE event observable on the B side.
will contain elements which depend on the user’s typ- - BA (not used) Action to be performed on the B side.
ing speed which is clearly not useful for an objective
measurement. <Index> is a continuous index, three digits, leading zeroes. Please note
that numbering is not necessarily consecutive, i.e. choice of index does
If time measurements are sufficiently fine-grained, it not carry meaning by itself.
is possible to separate human interaction-related time
spans from time spans caused by network or service re- For practical purposes in cases where the use case con-
sponse. For instance, if a prompt to enter data appears, text is clearly defined, there is also a short-form TPID
the user needs some time to read the prompt, enter the being used that omits the service and use case code
requested information, and send it to the service. The and the related delimiter.
service then responds with the next prompt until all re-
quired steps are made. 4.2.2 Trigger point IDs used
When the DFS event flow is monitored and record- The following list of events has been derived from video
ed manually, the granularity of time measurements, analysis of an actual DFS P2P money transfer, for two
and their accuracy, is limited. Therefore, it may be diffi- variants:
cult to separate service response times. Time measure-
ments for larger groups of activity — such as the entire a) App based. This category also includes browser
set-up phase as shown in Figure 4-2 — will inevitably based web applications. Typically such applications
contain human-interaction times. It can be expected use https or other secure protocols..
that after some initial training, the time to enter data b) USSD based (typically used on feature phone based).
will be quite constant from transaction to transaction. The actual platform was, however, a smartphone in both
However, time measurements should be expected to cases.
be of limited accuracy. For further reference see also Event and action flow.
It is plausible to assume that service response times
Table 4-1 shows the trigger point ID for the MoMo
in the set-up phase are of interest nevertheless. A pos- P2P transaction model used.
sible way to create respective data — at least on an av- With respect to the considerations discussed in Time
eraged basis — is to record a number of interactions by corrections for human interaction, and Special consid-
e.g. video and determine a typical “typing time”. erations for manual testing and time-taking the table
For a practical example, see the extended table in
clause 4.2, and the definitions provided there. also contains color-coding describing the nature of the
phase between respective trigger points.
The fields marked in blue identify parts of the event
4.2 Trigger point IDs
flow that relate to user activity. They are to be read
4.2.1 Trigger point ID basics in the following way: Beginning of the user activity is
A trigger point ID is a short-form notation describing a marked by the TPID preceding this element; the end of
specific action or event. The difference between action user activity is marked by the TPID assigned to the re-
and event is somewhat arbitrary and also depends on spective element.
16 • Methodology for measurement of Quality of Service (QoS) Key Performance Indicators (KPIs) for Digital Financial Services