Right now, selecting apps from the mobile app stores is a bit of a crap shoot. All people really have to rely on are the claims of the app maker and the user reviews.
Both both are deeply flawed and subjective. App makers often have no idea how to describe their apps in a simple way that helps consumers, JT Klepp of app self-publisher CodeNGo tells me.
The Palo Alto, Calif.-based app marketing and analytics firm Appbackr is now building a new version of its Android app-scoring framework that deals with this situation. The scoring system takes its cues from the way books, music, and movies are reviewed online, says VP of product and data Vy Nguyen.
The system applies a numeric score to Android apps, which could help people find great apps lost among the sometimes hundreds and thousands of options in the stores. Samsung’s, Amazon’s, and Firefox’s app stores use these scores.
The way Appbackr derives app scores involves several steps.
The company decompiles the code of an app to see how well it was designed. It looks at the application package file (APK) that Android apps use, and APIs used by the app developer to pull data from outside sources. It also studies the memory and power usage of the app.
Appbackr uses an artificial intelligence approach to simulate the way a person might use an app. It also has people use the app for a time and records the person’s screen taps. After a computer ingests this data, it learns how to operate the app. It can then test the app quickly and on multiple devices.
Finally, Appbackr uses machine learning to read both the developer’s description of the app and the user reviews. It looks for phrases in the app description that define the features and functions of the app and then compares those with the features and functions that users express in the reviews. In this way, the system can quantify how well an app does what it promises to do.
Nguyen stresses that the current app market has become an exclusionary, winner-take-all place. So when apps like Flappy Birds rise to the top, they have a way of staying there, to the exclusion of similar gaming apps that might be better. Successful apps can make a lot of money, helping their developers to update and improve them.
Appbackr is still working on the automation and machine learning aspects of the rating system, testing it with a small number of apps. As Nguyen will tell you, the question is whether or not the system can be scaled up to rate thousands of apps, on multiple platforms, quickly and accurately. It’s a tall order.
Everyone agrees that a merit-based way of reviewing and scoring apps might give lesser-known but good apps a better chance. “I think these things that surface the good apps are a big thing to me,” said Samsung director of developer relations Hod Greeley.
But Greeley is concerned about whether app buyers will take the time to use the reviews. “You’re only going to get people to read the app descriptions for a certain amount of time. People say, ‘I do want to know if this [app] is of value, but I’m only willing to read the description for 30 seconds’,” Greeley says. “Well, what do you want?”
App store curator Mario Tapia of Momentum Venture Partners questions whether a large app quality framework belongs in a world with so many bad apps. “At the end of the day, there really are only a few good apps, and the rest of them are just junk,” he said.
Both Nguyen and Greeley strongly disagree. As an example, Greeley says he just discovered a Twitter client that he loves. It was an app I’d never heard of.
Nguyen stresses that there are lots of developers who really care about delivering great apps, and those people deserve to have a chance to rise to the top and get noticed.