Why Knomee almost got killed

1. An application that failed to find its Product-Market Fit

Knomee was launched in 2017 on the iPhone AppStore, after 18 months of lean startup development. Self-tracking was a hot topic a decade ago and there were still a number of concurrent apps on the AppStore with similar ambitions.  Thanks to a moderate effort to make the app visible through this web site, through the web site, a twitter account and a Facebook page, we have managed to welcome a small but continuous stream of newcomers who downloaded the app. However, retention did not occur : almost every user got bored or discouraged rapidly and stopped using Knomee. Consequently, Knomee development effort slowed down considerably in 2020, until the beginning of 2023 where the possibility of killing the Knomee project and removing the app from the AppStore was seriously considered.

Knomee has managed to keep a handful of satisfied users over the years, and they have experienced the value promised by our "self-tracking with sense" motto.  Over the past five years, some our users, including our lead developer, have found "quests" (set of self-tracking indicators) that really helped them improve their health or keep a more fulfilling lifestyle.  Not only these positive experiences have reinforced the scientific knowledge that self-tracking, although tedious, is useful, but the insights provided by Knomee are a big motivational help.
We have also validated the hypothesis that flexibility is key: you may use Knomee for a short amount of time, to track for a couple of weeks or months something that is very specific to you.

Still the verdict is clear : Knomee in its 2022 version is too complex to attract users. The effort it takes to set up a few relevant quests (to each her  or his own), and to leard to use the richness of self-tracking data visualization and analysis is simply too much. There are too many concepts, the underlying machine learning algorithms are too sophisticated (not only most users don't see the point, but they slow the app down and often cause crashes when the iPhone is out of memory). The user interface, which also reflects the deep functional capabilities of Knomee, was found to be difficult and hard to use.

2.  Why Knomee deserves a second life

Although the decision to fix Knomee and to continue is based on a strategic analysis that is sumarized in the next section, there are also some narratives, that are observations that make us emotionally attached to the "availability of something like Knomee" :

• Self-tracking is indeed tedious at best and often boring, hence "tracking with sense" is mandatory. The insights that a tracking app such as Knomee may provide (correlation, delays, geolocation, cumulative analysis) are precious.

• There are other "customizable self-tracking apps" that supports combining personal note taking (tracking qualitative things such as mood) with HealthKit logging (automated self-tracking leveraging your smartphone sensors) ... but they are slow and it takes too much effort to use them daily (we tried most of them).

• Although the concept of "quest" is not intuitive, there is now a wealth of "self-tracking sets" with a proven track-record to help you improve your sleep or energy, to keep your weight under control, or to better manage your mood.

• During the last 6 years, thanks to the constant improvements to the Apple watch, Knomee has become a powerful to manage some health conditions such as Asthma (using blood O2 as an indicator, a "breathing-health proxy") or some heart conditions (using HRV and rest heart rate as proxys). By self-tracking at the same time the prescription ordered by your physician and the health proxys from the watch, one can create a positive reinforcement loop, one that has shown to be of significant value.

• Last, we have recently started to add automated tracking for swimming sessions at the pool, because swimming as a sport is poorly served compared to other sports that have beaufiful apps such as running or cycling.

3. Strategic Kiss Analysis :  Keep, Improve, Stop, Start

We will return on the feature characterisation of Knomee 3.0 when the application is launched. Today we want to share our "strategic analysis" that was made early this year.

Keep:

• The "Quest structure" is the heart of Knomee - it is made of a goal (the target tracker) and a set of one to three supporting values that you expect to help you reach your goal. A quest is your causality hypothesis, it makes self-tracking more interesting.

• Simple "all in one" user interface with thumb sliders, designed for speed. One screen is enough to see how you are doing and to enter a new set of values (you manage a full quest - 2 to 4 trackers - with one screen, and you can swipe easily from one quest to another.

• Obviously, we stick to our commitment for Full Privacy : no servers - everything on the phone. It is actually a distinctive feature of Knomee, most equivalent apps - that also come with a freemium subscription - rely on cloud exposure of your tracking data.

Improve:

• • The last four years have shown that a great part of the value of a self-tracking application lies with a proper HealthKit integration. We have redesigned the querying module to be more robust and more efficient through better parallelisation of queries.

  • The first version of Knomee was built with the idea that Artificial Intelligence would provide all the necessary insights. The last few years have shown, on the opposite that user-led data visualisation and analysis is the true source of self-knowledge. We have improved and enriched the analysis for hourly, weekly, geolocation, correlation (with delays and cumulative integration) patterns for all trackers.

Stop:

• • We have completely removed the complex machine learning algorithms and replaced them with a simple linear regression. It turns out to be enough for the forecasting use case that is built into Knomee. Because of our commitment to full privacy, we were running our complex algorithms on your smartphone ... and simplicity now translates both in better performance and robustness

  • Knomee was implementing "Granger causality." thanks to its complex forecasting, we now rely on user-led, simpler data analyses as explained in the "improve" section.

  • We got rid of Face ID integration, which added complexity and was  not really used. 

  • Last we removed some screens, such as quest summary (which was redundant with the "edit" screen). We also dropped some animations and buttons in order to provide a simpler user interface.

Start:

• You should be able to reuse trackers from one quests to another, both the definition of the tracker and the access to the last recorded value. It is quite common to define a handful of quests, such as energy, mood and sleep, that are inter-related. Knomee 3.0 now supports sharing trackers between quests (which is also a way to escape from the 4-tracker max limitation).

• As stated earlier, Knomee is really useful when swimming. We have updated the distance tracker to sum up over 6 days (to represent the weekly distance). We have also added two trackers which are independant from the pool size: "best 100m" and "average 100m", which are quite useful one one alternates one's training in pools that are of different lengths.

4. Conclusion : Stay tune for Knomee 3.0, coming this month ...

The development effort to execute this strategy (i.e., kill a lot of code, improve a few section and add a few new features) is over. The new version is currently under test and should begin its external tests soon. We expect to publish Knomee 3.0 on the Apple AppStore before the end of November 2023.

We will let you judge if the new application is better suited to your need, but we can already say that it is both faster (hence more responsive) and much more robust.