Page 15 - Methodology for inter-operator and cross-border P2P money transfers

effects are detected during a testing campaign. There may be undelivered transactions where money
Otherwise mis-classifications may result, such as is deducted from the sender’s account, along with
attributing effects of such limitation as functional transfer charges. In such cases, it will typically be
failures of a MoMo service. required to fill a complaint with the MoMo service
As the starting hypothesis for systematic testing, it is operator. If this complaint is successful, money will
assumed that a guard time is typically in the range of be returned at a later point in time (depending on the
10 to 30 seconds. process and the MoMo operator’s terms of service,
When testing is done manually, it is assumed that the transfer charges will not be refunded.
system can handle all testing speeds which can be Retrieval of lost money is understood as a second
realized by human testers, as even an experienced stream of activities outside the scope of this meth-
tester will not work significantly faster than an expe- odology. Functionally, even if money is returned later,
rienced regular user of DFS. Therefore, no special it will reduce the available credit for further tests.
requirements to slow down testing are applied. Therefore, in all cases of disappeared money, inser-
In fully automated testing, it would also be possi- tion of fresh money may be necessary to keep up the
ble to use the high degree of repeatability of such necessary level of credit for further testing.
control to determine the appropriate guard time by The matter of transaction failures needs special
probing, i.e., by systematically varying the guard time consideration. In that case, it is assumed that a
and check for respective effects. typical user seeks confirmation, by e.g., calling
There is a second category of effects which need or messaging the recipient (i.e., using an external
to be considered, namely the possibility of a means of communication). Also, in many cases, the
service-specific local memory (analogously to a receiver would issue a receipt confirming incoming
browser’s cache) which stores information related payments. The sending party might wait for that
to previous transactions. The effect would be that in statement and inquire.
subsequent transactions, such information would be In any case, in particular in testing modes where the A
read from local memory instead of obtaining them party has no direct visibility of events on the B-party
by an over the air request to the service. This could (this issue is also discussed in subsequent sections),
then impact related measurement values or KPI. reasonable and appropriate measures and conven-
As long as effects are quantitative rather than qual- tions, adapted to the actual scope and goals, shall be
itative, it may not practicable and is not necessari- considered and set-up as part of a testing campaign.
ly required to exclude frequency-dependent effects
entirely. However, respective effects need to be 8�6 Automation of tests
recorded and documented carefully as part of the The methodology in the present document describes
reporting in order to understand their impact on the testing in a generic way, i.e., service tests can be done
testing conditions. manually as well as in an automated way. It is under-
stood that automation of tests is desirable to achieve
8�3 Re-initialization after unsuccessful transactions a greater degree of repeatability, and less variation
If a transaction fails, in particular after a time-out in quantitative data values due to inaccuracy of e.g.,
condition has occurred, it shall be ensured that the manual time measurements.
service and the device or application are in the typi- Automation can have different forms with respective
cal neutral starting state again, i.e., that no memory degrees of automation up to fully automated test-
of previous error states remains in the system. ing. Using the multi-stopwatch concept as described
in this methodology is the next step of evolution,
8�4 Disappeared Money significantly improving the robustness of testing with
It is possible that during a transaction, the amount respect to manual event logging.
of money deducted is not correct with respect to The next step may be to still use manual operation
transferred amount and fees. This includes the case of transactions but to record low-level activities on
that the amount is correct but sent to a third party the DFS device itself, e.g., from recording of Layer
by an error in the system. From an end customer 3 messages or IP-level activities. The ultimate goal
perspective, this is either a loss (if too much money would be a system which executes the whole MoMo
is deducted), or an unjustified gain (if money is cred- process automatically. Respective implementations
ited but not deducted on the other side of the trans- require, at least, an extended level of access to plat-
action. For simplicity, we use the term “disappear” form devices (“rooting”, e.g., having system-level
for both variants of this kind of effect. access) and substantial technical efforts, in particular

Methodology for inter-operator and cross-border P2P money transfers 13