2 ms·
Oblique to this but still relevant. There's a retail chain I write the software for, including the website and online store. They send out occasional customer
by noduerme 5y ago
Oblique to this but still relevant.
There's a retail chain I write
the software for, including the website and online store. They send out occasional customer surveys, which I also built, and which are constantly reviewed by corporate management. So, they wanted to raise their Google review rating, which was hovering between 4.6 to 4.8 stars, varying by location.
The CEO called me with this idea that at the end of taking a survey, people who give us five stars in all ten survey categories should be shown a link to directly review the store on Google.
Me, as an engineer - I was like, okay let's sit down and review metrics for each survey question, weight them and figure out how likely someone is to give us 5, rather than 4 stars in aggregate, and use that to decide whether to give them the link.
He said no. The only people who should get the link are people who gave 5 stars on the survey in every single category. This seemed counterintuitive to me and I ran the numbers and found that only a small fraction of people did. But I accepted the dictat and did it his way.
End result: About a 30% increase in 5-star google ratings across the board. His intuition was right, that the people most motivated to take the survey and complete it with straight-A's would be the people who would take the time to write great comments.
You can and probably should infer from this that if you actually need to balance "average" ratings, the outliers on the top and bottom of the scale exert an inordinate amount of pull and tend to skew results toward the extremes.