Siri is so aloof; Saga wants to get to know you

A new personal assistant app called Saga makes its way into the iOS App Store (s aapl) on Tuesday, and it wants to get to know users better than they know themselves. Like the recently unveiled Google Now, Saga tries to take the Siri concept a step further by proactively telling users what they need to know when they need to know it. Doing this requires lots of data and algorithms capable of learning what users are really up to when they’re out and about.
Andy Hickl, co-founder and CEO of Seattle-based A.R.O., the startup that built Saga, doesn’t see the app as a Siri replacement as much as a “chatty little brother or sister, or sidekick” that makes suggestions you might come up with on your own if you had the time. Assuming users grant it permission to track their location, Saga will know where they are, how long they’re likely to stay and how long it will take to get home if they leave now. Near lunchtime, it will suggest good places to eat based on the places users (or maybe their friends) frequent. It will tell users how far they’ve traveled recently and breakdown where they’ve spent the most time.
In some ways, Hickl said, Saga is a “quantitied self play for the things you do and places you go on a regular basis.”

Gotta train that model

What any good quantified selfer (or data scientist like Hickl, who also founded answer engine Swingly) knows, however, is that increased accuracy often goes hand in hand with more data. Saga is no exception, provided it’s good data to start with. That’s why initially, Hickl explained, apps like Saga require a “burn-in” period to teach them who you are.
Especially in dense urban areas, it can be difficult for an app to determine whether you’re at a coffee shop, the Italian restaurant across the street or in your apartment six floors above it. (Here’s how another Seattle startup, Placed, is able to pinpoint users’ locations.) If that restaurant is one of the most popular in the city, Hickl said, Saga likely will initially assume that’s where you are. Usually, though, after about 48 to 72 hours of feedback in the form of you telling the app where you really are, it’s able to learn your routine and make more-accurate predictions.
Another thing that apps like Saga need time to learn is how to gauge users’ intentions. “We don’t have sensor data that lets [Saga] know whether I’m buying beer or buying Benadryl,” Hickl said.
What he means is that if someone’s at Walgreen’s at 11 p.m., he could just as easily be buying a 40-ounce bottle of Miller High Life on his way to a party as he could be buying a box of cold medicine for a sick wife. Saga will instinctively flag suggestions based on both possibilities, but knowing that someone rarely leaves the house past 9 p.m. on weeknights — especially to go to bars or clubs — will skew the model toward the latter choice.

Always more useful

In the future, Hickl said, Saga hopes to become even smarter by becoming a platform for other apps with unique user data, and by bringing in friends’ data to offer more-curated suggestions. He also wants to figure out a way to get Saga off of just mobile phones and onto new devices such as wearable displays. Inventions such as Google Glasses can’t just be a new way of finding out the weather or checking email.
Essentially, though, it’s just about making that app smarter and more useful and then going with the technological flow. A.R.O. has a lot of good ideas for Saga, Hicki said, “[but we] recognize a lot of our good ideas are 2012-era thinking and could be laughable in a few years.”
A.R.O. also has to figure out a way to make Saga stand out against what will certainly be an onslaught of similar apps, including the aforementioned Google (s goog) Now that will likely find its way natively onto million of Android phones over the next couple years. Saga’s privacy-by-design features (it has mechanisms “to nuke [your data] from orbit,” Hickl said) and A.R.O.’s refusal to sell user data might help — especially against privacy scapegoat Google (Saga is coming to Android at some point) — but making use of location data is the new black for mobile developers. This saga is far from over.