Keywords: crowd sourcing, big data, panel management
Crowdsourced panels can provide a targeted source of data for difficult to measure retailers. To achieve this, panels require engineering recruit and retain the appropriate panelists. Often attempts to select desirable panelists have unintended consequences. Panelists that are rejected from screened recruitment methods can find alternative methods of registering, leading to misleading estimates of “organic” growth of the panel. In addition, allowing compliant members to invite new panelists for a premium reward incentivizes panelists to commit fraud. Once panelists have successfully been recruited there are a variety of techniques to predict future compliance. Activity in the first three days, such as winning a prize, is highly predictive of future compliance. Choosing to save virtual currency for a cash out, as opposed to entering sweepstakes, was also predictive of future compliance. Conclusions were drawn from panelist data from the Shopprize Android app, a partnership project with Nielsen and OurCart. Shopprize panelists upload food shopping receipts images in exchange for digital money. Analysis was completed using R.