Start a new topic

Getting ready to Score, first did lots of Previews

I'm about to Score after running lots of Previews. We've been doing predictive models for about 2 years, but before then we put together ideas for an engagement score and came up with about 14 things to look at and had an assigned a point value (we never got around to scoring or testing it because it seemed like it would take too much time to run...yay ScoreOmatic!).

On first Preview, the point allocation didn't make very much sense(putting too much weight onto coming to an event, not enough of giving us data, etc.) and I played around a bit with the points. It would be helpful to get more of the analysis on the Preview end before Scoring, or a way to look up a Constituent to see where they landed. It takes a little while to figure out how to get a better even distribution of points and that way more testing can be done before anything is committed (reducing back and forth).

Also, I tried loading up all my queries in the profile before entering points and I couldn't do that. But that would just be a work-style preference for me.

Overall, I think this is great! I can see so many uses and a lot of time saved.

Hi Corbett,
You can open a constituent from the preview screen by clicking the folder next to their name. Also, what other info would you like to see in the preview?
I meant looking up a Constituent, not opening one. For example, if I'm scoring a group of 10,000 people and I want to see were Bob Bestalum lands, I have to scroll or page down to where I think he'd be.

The info that might be nice in the preview is a summary Constituent count by score that allows you to quickly see the distribution. Something like:

Score 100 - 50 Constituents
Score 99 - 95 Constituents
Score 98 - 75 Constituents
Score 97 - 5,000 Constituents

Average and mean are good too, but something like this would tell me "Ah ha, why is everyone scoring 97?!?"

Hi Corbett,

Great feedback, thank you! We'll add a default score of "1" for each new row so that you do not need to complete each before moving on to the next one (or, perhaps more correctly, it will be completed with the default score and the user can change it if they wish). It won't be as fast a change, but I can definitely foresee a visualization that shows a bar for each individual score and number of constituents in it (vs. the percentile view). Another one we thought might be interesting for the preview is a visualization to show how many constituents got points for each line in the profile, Query #1 matched 1234 constituents, Query #2 matched 432 constituents, etc. Any other visualizations you can think of that would give insight?

I also like the idea of a quick summary visualization of how the constituents scored on the profiles. This is also something that we wanted to look at immediately after we scored our constituents to see how they all fell out. I think a lot of organizations use this back-and-forth method for determining whether their scoring system is working by looking at some of their key constituents to see whether they scored high enough to make sure the model is working the way they wanted to. We had to be very careful to make sure that we were not forcing our predictive model or scoring to give us the output that we thought it should give us. For example, looking at one of our highest major gift donors and the attributes or variables that they possessed and then using our scoring to say those are the attributes or variables that all major gift donors should possess or score high in to be a predictor of future major gift donors. That is why we chose to use a regression analysis to create a predictive model for us to determine the variables that were our highest correlation to a major gift donor, Of course that is based on the major gift donors that we have at this point in time. However, with that being said I still like the idea of seeing a breakdown of the scores especially so that if you're looking to make actionable decisions on a certain segment you can quickly look at the counts!
Login or Signup to post a comment