Field data is collected quickly today and, no matter how well a field data collection program is set up, errors can be introduced in the data collection process that have not been picked up by automated flags.
Check Timestamps for Rushed Responses
The surveyor who fills out 40 questions in 4 minutes is probably not reading the questions. He or she may be backfilling the form out later, in which case the form will look perfectly normal – all the questions filled out in the correct format – but the answers will all be made up. Human reviewers should be looking for completion times that are below the minimum realistic time for filling out a form.
Watch for GPS Drift
There is also the issue of GPS ‘drift’ and how this can affect the accuracy of GPS logged information, especially when used to record visits within retail premises. Indoor GPS signals can fluctuate significantly and records made in this way can end up being logged as being within the building when in fact they have been recorded as being somewhere else entirely within the building. These errors can be easily identified by cross referencing location information with the site boundaries for the individual store(s).
More on Data Analysis Company can be found at https://shepper.com/.
Identical or near identical answers have most likely been copied from a previous entry by the surveyor instead of them having entered the information themselves. The amount of time that it takes for a reviewer to check through two forms from the same operative and find all of the instances of duplicate and/or copy and paste is minutes at most.
Look for Logic Conflicts in a Single Form
A store that has been recorded as closed should not have any recorded stock levels. While these types of errors may pass through automated validation because each individual field has been completed correctly, a human reviewer reading through a completed form will very quickly pick up on such discrepancies.
Review Photos in Context
An automated check of images against a set of rules confirms that an image has been uploaded and that it conforms to the basic rules for a file of that type. However, it cannot confirm that the images show the correct fixtures or that they have been taken at the correct location. A sample of images should therefore be reviewed against the brief for the collection of data. The quality standards for collected data published by the UK Government Statistics Service set out why it is necessary to verify accuracy, consistency and completeness of data in collection.
However, any reputable data analysis company will make QA by a human a core component of their work, since it is often difficult to tell and distinguish between error (noise) that appears to be correct and real insight within a data analysis dashboard.
