The Dark Knight: A Movie‑lover's Lesson In Web Analytics

My daughter really wants to go see the opening of Mama Mia! this weekend. But, while a good play, it is not exactly what I am looking for in a movie. On the other hand, I am dying (heh) to see Heath Ledger’s posthumous appearance as the Joker in The Dark Knight.

“It got a ‘Must Go” rating on Fandango last night,” I pointed out to her. Ever the analysts daughter, she retorted, “And that was out of three people?”

Well, no, that were actually hundredds of people who succeeded in seeing it before it opened (Hmm, maybe it opened in other countries in other time zones, the way that you could get an iP

hone in New Zealand almost a full day earlier than here.) But it got me to thinking. If there are only five rankings: Must Go, Go, So-So, No and Oh No! — then how does anyone ever achieve a ranking at the ends of the scales? It’s like asking someone to take a survey and they can choose a number between 1 (lousy) and 5 (awesome) — unless everyone chooses a 5, how does anyone end up with an average of 5?

“Aren’t you assuming a lot about Fandango’s algorithm?” asked John Henson, famous creator of the GA Goal Copy tool. “Maybe it’s like the Google ratings,” pointed out SEO Jim Gianoglio, “They count more if you not only rate but also write a review.”

Well, analytics to the rescue. If you click through, you can actually see the rankings in buckets (sort of like the Google Analytics loyalty charts, but without all the misleading titles):

Obviously, you don’t have to get all “fives” to get a five. So let’s expand the system and pretend that Fandango weights all answers on a scale of 1-10, and you have to get between a 9 and a 10 to score a “Must Go.” And maybe each vote gets the top of its category (so if you vote “must go,” it is worth ten points, and if you vote, “go” it is worth eight points. We would have (in my made-up algorithm):

45*2, 26*4, 63*6, 96*8 and 991*10

all of which gets divided by the number of votes, 1221. For a weighted average, i.e. raking of 9.21376 (OK, that is a little overly precise given that I don’t know the algorithm.)

Late note: After publishing, I realized that this (made up) algorithm only works at the high end. What if you had a lot of Oh No! and a scattering of other rankings — if you gave a “two” to an “oh no!” ranking, you could never get a movie to rank, overall, as an “oh no!” So probably it is more of a sliding scale — but the concept is the same.

Well anyway, that is your web analytics movie lesson. Enjoy the weekend. Comment when you see the movie and tell me if I should go.