Page 28 - Methodology for measurement of Quality of Service (QoS) Key Performance Performance Indicators (KPIs) for Digital Financial Services
P. 28

Does the timestamp range match the time window    This was not entirely unexpected; manual mode was
             recorded for that location?                      deliberately chosen to provide the widest possible an-
                                                              gle of view and maximum transparency of the data ac-
           ❏  Check timestamps of background measurement data   quisition process.
             against respective location logs
                                                                The weaknesses of manual operation are manifest in
              Does the timestamp range match the time window   some main fields:
             recorded for that location?                        Time-taking of the transaction,  with  typical  time
           ❏  Decide on necessity to exclude time ranges.     scales in the order of a few seconds, introduces quanti-
                                                              zation errors at best where times have to be read from
              Does the location log indicate special events and con-  displays with a typical resolution of one second. Other
             ditions, which set the need to exclude data from the   effects come on top, such as potential errors due to time
             set?
                                                              offsets when using different devices. Time-synchroniza-
           ❏  Visualize timestamps of transactions: Are there any   tion of such devices can only provide limited protection
             gaps or unusually dense transactions during a period   as this compensation is itself quantized to one second
             of time? If yes, validate reasons.               steps (unless modified devices are being used).
                                                                Transferring readings to paper logs open up addi-
           ❏  (further check items to be added)
                                                              tional sources of error due to handwriting.
                                                                Further transferring of paper logs to electronic
           9.1.2 Tests on background test data
                                                              means—which is a prerequisite of data processing—is
           ❏  If GPS data are available, does the location indicated,   again prone to reading errors.
             and the GPS location match?                        All process steps are essentially dull and repetitive,
                                                              and are therefore vulnerable to human errors.
           ❏  Visualize timestamps of transactions: Are there any   Typical error patterns are:
             gaps or unusually dense transactions during a period
             of time? If yes, validate reasons.               a) Reading from the display: subsequent timestamps
                                                                10:49:58 10:50:02 logged as 10:49:02
           ❏  (further check items to be added)
                                                              b) Transferring from paper logs: number switches such
           9.1.3 Cross tests between data (after import)        as 1<>5, 2<->3, 2<->5, 3<->5, 1<-7, 4<->9 depending
                                                                on handwriting.
           ❏  Validate time stamps of DFS and background data
             for consistency.                                 c) Transferring from paper logs: eye-jumping to the line
                                                                above or below the actual one
           ❏  Validate consistency between network unavailabili-
             ty in DFS and background data. A possible consis-  d) Transferring from paper logs: Number-switching, e.g.
             tency  problem  exists  if  background  data  indicate   12:30:14 ->13:20:14
             network unavailability but DFS transactions work   Transferring from paper logs: simple typing errors.
             during a given timespan. If such periods of time    Of course, it is possible to extend the manual estab-
             exist, mark them in the database and seek further   lished data quality  assurance  procedures, which can
             clarification.                                   prevent or eliminate errors. However, this is a signifi-
           ❏  (further check items to be added)               cant cost driver and therefore needs to be considered
                                                              against automation or partially automation of the data
           9.2 Additional processing                          acquisition process. Some suggestions how this could
           With respect to some KPI definitions, additional check   be done are given subsequently.
           procedures may be done.
                                                              10.2 Recommended measures
           Examples are:                                      Fully automated DFS transactions would eliminate all
           1.  Check consistency of accounts throughout a se-  of the above mentioned sources of error. If budgetary
             quence of information SMS.                       conditions allow, this would be the method of choice.
                                                              It should however kept in mind that careful—and peri-
           2.  Check for “false negatives” (ref. Money Transfer False   odically repeated — validation of automated solutions
             Negative Rate MTFNR) by comparing account bal-   is part of the design of such a solution and therefore
             ance against transaction results.                needs to be considered in a cost assessment.
                                                                The next best solution — in case budgetary or other
                                                              considerations lead to the decision to not use full auto-
           10 LESSONS LEARNED
                                                              mation—is tool-assisted time-taking. A respective tool
                                                              would have the following basic functionality:
           10.1 Overview
           The manual capture of DFS TA has shown to be a major   -  Android app with a simple and user-friendly user
           weak point in the campaign.                          interface, e.g. showing a group of buttons with one
                                                                button per timer flag.


           26 • Methodology for measurement of Quality of Service (QoS) Key Performance Indicators (KPIs) for Digital Financial Services
   23   24   25   26   27   28   29   30   31   32   33