How data becomes insights – Part 2: Analysing data
From the first part of our mini-series “How Data Turns into Insights”, we already know that we process around 3 billion mobile signaling data every day, which first pass through a TÜV-approved anonymization process.
Part 2: Data analysis
n the second part, we will look at the processing of the data and how meaningful analyses can be made with a market share of less than 50% and an inaccuracy of several hundred meters.
In a first preprocessing step, the anonymized data is aggregated and analysed for obvious outliers. Such outliers can occur, for example, when mobile terminals register with different macro mobile radio cells in rotation because the transmission ranges of the cells overlap. In such cases, geographical jumps of several kilometres can occur within a few seconds.
After all “outliers” have been cleaned, the macro data (Summarized Signal Points) are divided into moving and stationary segments and extrapolated according to the regional extrapolation factors.
Aggregation of the anonymized data is another mechanism in terms of data protection, as at least 20 anonymized moving streams are combined. Using specially developed models and analysis procedures, we follow a probabilistic approach and survey in which large-area locations were visited in chronological order, taking into account various influencing factors. In the course of these analyses, we can provide our customers, for example, with source-destination matrices at municipality level, as well as a frequency landscape using the Statistics Austria grid (500m x 500m) in interactive dashboards, pdf reports, or csv files.
We will report on the provision and presentation of these anonymized analysis results next week.
