Projects with which we have been significantly involved with include:
| Defence quantitative risk analysis (QRA) Individuals now working for Andcer have determined the likely cost outcomes and associated cost risks for many major projects. Some of the larger... see below | Defence Fixed High Frequency Radio Replacement (cost modelling, and data discovery) The New Zealand Defence Force (NZDF) is replacing its legacy domestic high frequency radio system. This system... see below | Defence light aircraft Beechcraft 200 (economic cost studies) The New Zealand Defence Force had a fleet of B200 aircraft (more). These aircraft were leased, and were mainly used for aircrew... see below |
| NZDF sizing of aircraft engine inventory (QRA-logistics) High end military aircraft engines are expensive. Too many engines on inventory incurs excessive costs and creates a capability that... see below | Telecommunications (regulatory cost model) Telecom New Zealand (TNZ) (now Chorus and Spark) provided their regulator, the New Zealand Commerce Commission, with a cost model showing the costs of... see below | NZDF high capacity mobile communications (demand and economic cost model) WGS (Wideband Global Satellite communications) is a high capacity communications system owned and controlled by the... see below |
| Defence policy (optimisation) Policy and policy related discussions are often performed without the benefit of quantifying the economic cost and benefit of numerical optimisation. Policy is... see below | Statistics: tools to detect fraud [the value is in the detail] Curiosity is probably the analyst's best friend in examining data for fraud or in any modelling. Analysis of data is about finding... see below | ACC organisational performance measurement At the time of this study, ACC (the Accident Compensation Corporation) had approximately 33 main branches. Each branch was largely autonomous in its... see below |
| Police-organisational performance measurement The New Zealand Police have many Policing Areas distributed around New Zealand. This study examined the contribution of each Area to the overall... see below | Department of Labour (currently MBIE) Immigration Office Performance Data Envelopment Analysis (DEA) was used as the base technique to assess Immigration Office organisational performance. ... see below |
Curiosity is probably the analyst's best friend in examining data for fraud or in any modelling. Analysis of data is about finding patterns. Any discovery that is different from the general pattern might reveal some really interesting truths, either in creating insights into valid phenomena or occasionally detecting fraud.
Individuals now working for Andcer have been involved in projects to detect fraud.
Statistics have a real-world application. The assignment was to create a predictive model of factors that correlated with complimentary write-offs and refunds from a dial-in help desk. We devloped a model and summarised the data so that Telecom could investigate initiatives to produce better supported services and understand the drivers of write-offs.
When calls arrived, they were randomly distributed via ACDQ (Automatic Call Distribution and Queuing). After the primary analysis was complelted a final analysis was undertaken to check how effective the ACDQ was. There was expected to be no association or repeating pattern between the called number and the answering operator. This was a massive data set. There would have been unobserved factors such as callers making contact say only during the weekend, when operators working during the week could not answer the call. It was expected that the test of no association, ie that the relationships were random, would fail.
The test did fail, but not in ways that were expected. There were strong associations involving some operators repeatedly provided write-offs to the same customers. A little more investigation showed that these write-offs were to their home phone.
When using statistics, it is important that results used in analysis should be statistically relevant. This is necessary but not sufficient. Results also need to be of interest in a business sense. In the above example, the association was not random and would probably never have been random. However, some of the results were of "Business Interest".
The top three skills required and used for this project were: