By Piyush Khurana
A notable recent advancement within sampling has been automation. This has largely been through a technique called ‘Programmatic Sampling’.
Programmatic Sampling partners with API – a software intermediary that allows two applications to talk to each other – to automate the process of supplying sample. Resultantly, sample suppliers can program their targets, quotas, field times and potentially work more directly into panel databases.
Consequently, traditional notions of researchers waking up in the middle of the night to check quotas no longer exist. Today, researchers can set quotas and forget about them. And guess what? Researchers are now getting speedy and effective results compared to traditional sampling methods.
Prior to the invention of Programmatic Sampling, the process was very set – suppliers bid on projects, committed to feasibility, made phone calls manually and sent emails to confirm prices before surveys went live. Now more suppliers are using programmatic approaches. This has meant sample quality is higher. This is because programmatic approaches are less error-prone, faster, require little or no human intervention and increase participant engagement.
But programmatic approaches are not perfect. However, technology may have the answer again. This time through machine learning methods:
- Dropout rates are still high – machine learning can help solve this by understanding why people drop out of surveys and fixing it
- Data quality issues are still rife – machine learning systems should be capable enough to interpret the quality of data and clean it on the researcher’s behalf
- Unmet quotas – machine learning can do better at identifying where sampling shortfalls lie and recruit sample to fill these without human intervention
As you can see, programmatic sampling approaches have already greatly helped researchers. However, we still have sampling problems that programmatic approaches and machine learning are yet to solve. I have outlined how we can currently do this.
However, I hope to be able to add some new learnings about Programmatic Sampling from the forthcoming ESOMAR Global Congress!
By Piyush Khurana, Track Opinion