3 ms·

Marco is correct in pointing out that the "top" list approach has a few flaws.However, I think that the "top" list (in addition to Apple's featured list) is the

by credo 13y ago

Marco is correct in pointing out that the "top" list approach has a few flaws.However, I think that the "top" list (in addition to Apple's featured list) is the best solution among a list of sub-optimal solutions. He seems to suggest the top-lists should be replaced by a list of apps that Apple classifies as "great software". However, Apple already shows a list of featured apps and this list already has a huge impact on app rankings (and IMO Apple will never be able to create a perfect and comprehensive list) From my experience, I'd say that removing the top-lists will hurt the ecosystem (most users as well as many developers). For example, one of our iPad apps was #1 in the Finance category for six weeks. The media first mentioned this app several days after it became #1. Apple also featured the app a few days after the app became #1 (the app was never priced at the lowest price tiers). The app got initial traction organically and the #1 ranking helped it get even more traction. It isn't realistic to expect Apple to be able to pick a list of "great apps". Following up on my example, we launched a new version of the iPad app a couple of weeks ago, but Apple isn't featuring it on the "new and noteworthy" list even though the app has hundreds of reviews with 2/3 of them being 5-star reviews and another 1/4 being 4-star reviews Relying solely on Apple's featured list and on media-connections or advertisements may work for one class of developers, but for many other developers, the (category-specific) top-lists are helpful. I also think that the top-lists are helpful for users. Removing the list isn't going to make it any more easy for users to search through a million apps in the app store.

salmonellaeater 13y ago
Another option is to fix algorithm used to sort the top-list. It's similar to search results, in that a purchase from the top list should be attributed to two* factors: 1) the app's quality, and 2) its current position in the list. Basically, you can separate the effect on purchases of the position in the list from the effect of the app's quality, and then sort apps based on their quality rather than raw purchases. *Maybe I should say, at least two factors. You can go really deep when constructing rankings.
- __alexs 13y ago
  So apps should have a sort of ELO ranking where they fight with apps in the same category (possibly multiple categories simultaneously) for dominance and the fight is implemented as which app gets bought when displayed to a user at the same time as another similar app?
  - salmonellaeater 13y ago
    The ELO ranking is more suited to explicit head-to-head matchups, but its statistical basis could be adapted to app rankings. You could basically collect statistics for each app position (#1 averages 1000 purchases/day, #2 averages 500, etc.) and then estimate the quality of each app based on how its purchase rate differs from the expected average for that position. Google and other advertisers face a similar problem with sorting ads, which suggests some other relevant inputs: http://en.wikipedia.org/wiki/Quality_Score http://en.wikipedia.org/wiki/Quality_Score

by credo 13y ago

salmonellaeater 13y ago

Another option is to fix algorithm used to sort the top-list. It's similar to search results, in that a purchase from the top list should be attributed to two* factors: 1) the app's quality, and 2) its current position in the list. Basically, you can separate the effect on purchases of the position in the list from the effect of the app's quality, and then sort apps based on their quality rather than raw purchases. *Maybe I should say, at least two factors. You can go really deep when constructing rankings.

__alexs 13y ago
So apps should have a sort of ELO ranking where they fight with apps in the same category (possibly multiple categories simultaneously) for dominance and the fight is implemented as which app gets bought when displayed to a user at the same time as another similar app?
- salmonellaeater 13y ago
  The ELO ranking is more suited to explicit head-to-head matchups, but its statistical basis could be adapted to app rankings. You could basically collect statistics for each app position (#1 averages 1000 purchases/day, #2 averages 500, etc.) and then estimate the quality of each app based on how its purchase rate differs from the expected average for that position. Google and other advertisers face a similar problem with sorting ads, which suggests some other relevant inputs: http://en.wikipedia.org/wiki/Quality_Score http://en.wikipedia.org/wiki/Quality_Score