The most effective way to identify errors in your data analysis is through the field visits. The remote collected raw data looks neat and complete until you identify the errors in the data analysis.
Check Equipment Placement
Counting sensors and cameras is not enough to produce meaningful figures, as long as they have been correctly set up. A quick check will confirm that the footfall counter, for example, has been set up straight and hasn’t been moved since the last visit.
Spot Miscoded Locations
When data collection takes place across different locations, it is easy for location codes to be misspelled, for example after a store refit or changes to addresses. However, on a field visit you can check that all the data for a certain location has been assigned with the correct location code.
Reconcile Sensor Readings With Reality
Sometimes the numbers from the sensors can look correct, but when the surveyor is on site and observes the situation, it becomes clear that the numbers don’t match up with what is really happening. Such deviations are caused by so-called sensor drift or by an obstruction of the sensor.
Confirm the Sample Matches the Brief
There are many factors that can affect a survey and it’s easy for a survey to be taken out of context. This is especially true when the intended population is counted or questioned in a survey and they don’t match the intended population as specified in the survey brief. In such cases, it is much easier to determine how survey methodology affects data quality when conducting a survey in person as opposed to relying on a spreadsheet for the survey.
Catch Environmental Changes
Changes to signage, construction in the vicinity, or other environmental factors can all affect customer behaviour at a store. A Data Analysis Company who carries out field visits can pick up on changes such as these whilst analysing data and amend their assumptions as necessary.
By combining site checks with desk-based analysis, it ensures that the numbers it works with are accurate and based on reality, rather than just the figures in a dataset.