Monthly Archives

2 Articles

Posted by knomee on

Scoring Quests with Knomee

 

Knomee assign a score (from zero to 100%) to each quest, that is displayed on the main screen in the upper right corner. If you click on this number, you access the quest score view which is illustrated with the image above. Here you see a quest with a score 68%. To follow-up on the previous post, this is a quest whose target is “fitness” (that is measured using labels who are unique to the user of this quest), while the three factors are respectively the amount of ingested fat, ingested cereals and the rest heart rate as measured by the Apple Watch.

 

This score is made of three parts :

  • The quantity of data : a third of the score tells you  if you have accumulated enough data to get significant insights. Depending on your data, Knomee will wait until you have between 30 and 60 measures to give you a high score on this part. In the example that is show in the picture above, you can see in the green box that this first component is 100% (there is enough data, as told by the comment)
  • The insight score : this tells you if Knomee has found interesting insights using the three factors, or the time, day and location of the measure. If this sub-score is high, you will see that the relevant factors are coloured. In this case, “fat” and “cereal” are positively correlated, so they are coloured as green. The “chart view” is where you will get more details, but if you click on one of the factors, you will get a short insight summary as a forecast. In this example, the strength of the insights is not very strong so the sub-score is 25%.
  • The forecast score : this says if Knomee is able to forecast your successive measures well. Recall that whenever you enter a new measure, Knomee pre-position the sliders to its “guessed” value. This is mostly offered as a way to make the app faster and more playful, but Knomee keep the score on its forecasting capability. In this example, Knomee is on average 10% away from the actual value, which translates into a good third subscore of 77%. A good forecasting score is a sign of Granger-causality.

 

At first you will simply look at the score on the main user screen (the measure capture screen is the home screen), but after a while you will probably venture to this quest score screen to understand your quest score better. You can see that the rest of the screen displays a description of your quests, with colors to attract your attention on the relevant factors.

What does my score tell me?

  • A score below 50% says that either your quest is too young … or that it is not very significant. There may be many explanations, but most often it means that your causal diagram hypothesis is false. Put more bluntly, it means that your factors do not seem to have much influence on your target. As said repeatedly, this is a critical feature: Knomee helps you distinguish between your “hunches” and “data-supported causation schemes”. As it turns out, we are often “fooled by randomness”.
  • A score over 70% says the opposite: your “causal hypothesis”, i.e., your quest, is definitely interesting. By navigating through the various screens, you are likely to find interesting insights. If you have allowed Knomee to send you notifications, you will receive one of these insights daily.
  • A score in between means that your quest is interesting but there are probably many other factors influencing your “target”.

Knomee has no ambition to know you or to understand you. It is your job to understand yourself better through this self-tracking and these insights. The ambition of Knomee is to help you craft interesting self-tracking quests. The score is a great tool to help you during this journey. You should try quests and drop those whose score stay low. You should play with factors to see if some new factor improves or decreases your score. There are three stages of using Knomee:

  • At first, play with the default quests to get familiar with the app. Although we tried to select three quests with a broad range of interest, you are likely to get bored quickly or to say “this is not for me”. Remember that there is a “quest library” so you can substitute many other quests to start this first learning stage.
  • Knomee starts to be interesting when you define your own quest. We have tried to make this as simple as possible, and we shall continue to work on our design. Knomee is not “another self-tracking app”, it is “your own tracking app”. It takes five minutes to define a new quest, and then you can enjoy a tool that is unique to you. However, our experience suggests that you need to play with the existing quests (stage 1) before moving to stage 2.
  • There is only a finite amount of self-knowledge that your will extract from a given quest. After a delay that varies from a few weeks to a few months, you will be done with that quest. There are two ways to keep using Knomee : try a new quest, randomly, once in while … or continuously optimize your quest by changing the factors, the scales, or the data sources (switching from declarative values – when you input your data- to automatic values – when the value is read from a connected device through HealthKit).

Once you reach this third quest, you should consider sharing your quest with others. The screen shown above has an “envelope” button that allows you to share a recommendation through email. In the future we plan to make quest sharing simpler and more seamless.

 

 

Posted by knomee on

What is a Quest and why does Knomee use quests ?

Quests are central to the Knomee experience. On the negative side, they make Knomee a self-tracking app with a steeper learning curve, that requires a little bit more time and effort, compared, for example, with one of the activity or mood tracker that one may find on the App Store. On the positive side, quests give more sense to tracking, they make the tracking experience more fulfilling and they help Knomee give you more meaningful feedback about yourself. A quest is a group of things that you want to track, each of them is called a tracker. Trackers are defined by one thing (weight, sleep, activity, mood, etc.) that you track, either by entering the value (using the sliders in Knomee’s interface) or by importing the value from “HealthKit”, the Apple service on your iPhone that collects all data from your connected devices (wristband, watch, scale, sleep monitor, etc.) or your iPhone itself.

A quest is what is called a “causal diagram” (albeit a simple one) in the scientific world. This means that a quest represents a causal hypothesis that you make, about yourself. A quest has one target and one to three factors. When defining a quest, you tell Knomee that up to three factors (say, the time you go to bed, the amount of steps that you walked, and the richness of the dinner that you had) have a causal influence on something that you care about, the target tracker (in this example, it could be the number of hours that you slept).

Self-tracking is good for you but boring. This is not an opinion, this is a scientific fact. It is proven that self-tracking helps you both to know yourself better and to help you change your behaviour towards a goal. It is also proven than most people stop self-tracking quickly, from a few days to a few weeks. Knomee was created to tackle this challenge, and it is a hard one.

The only thing that makes this self-tracking worthwhile is learning about yourself. This is why we selected “self-tracking with sense” as our motto. We came up with the quest idea for two reasons. First, a “quest” is “indeed a quest to know yourself better” and to see if your “causal hypothesis” happens to work. In many cases, using Knomee is a way to see if doing some particular effort is “worth it”.  Second, a causal diagram is a powerful tool to orient the machine learning and statistical analysis. It makes the problem of “making sense from your data” easier and helps us keep everything on your phone (hence our guarantee of full privacy).

Quests don’t last: you formulate a hypothesis, you learn (or you don’t) from it and you move to other things. We have made it easy to add and drop quests to Knomee. Our usage statistics show that this is not properly understood yet. Many users start with the pre-defined quests and never venture to create their own. This is a shame since it is unlikely that those pre-defined causal diagrams apply to you.

Quests are meant to be shared: the future of Knomee is to make it easier to share successful quests with others (sharing the model, the causal diagram, not your data). In a reciprocate way, it would be nice to be able to look at quests that have been successful for people who have the same “target” goals. For instance, although everyone is different, it would be nice to have access to a collection of successful quests from people who tried to improve the quality of their sleep.  Currently, sharing is possible but cumbersome: you send the model/quest description to a friend through email … and she/he may import it.  The quest library is another way to benefit from quest sharing but this feature is still in an infancy stage.

You may wonder why  no other tracking app is using quests. This is because we take our mission “self-tracking with sense” seriously. It is clear that it is hard to make sense with one single tracking dimension. You know this already if you are using a tracking app or if you are looking at your health app on your iPhone. At first seeing all this data and this nicely shaped charts is exciting, but you get rapidly bored because there is not much value there. The chronology (looking for weekly, daily and hourly patterns) is the most interesting part, but only a few apps do a decent job at it. If you “self-track” regularly, you will notice that the interesting questions arise from the combination of factors. There are a few Knomee competitors that upload all data in the cloud to search for any interesting combination or pattern. We already said that using quests (simple causal diagrams) makes the analysis simpler and suitable for a “device-only” solution (everything on your phone) but there is another reason for using quests. You are in charge, you know better than anyone what questions are interesting for you.