Hi all, I wanted to make a short video explaining a new screen I added to my mood tracking app “Changes”. This is the product of a lot of self-conscious procrastination but I’d be interested in your reactions to the result.
Let me know if there’s anything you’d like to know more about, ideas on how I could do better for my next video, and, ideally, any places you think I could share this to drum up some business!
Thanks, hope I don’t come off as too much of a “marketeer” - I’m doing my best to learn how to promote my work with media like this.
The tracking looks very well thought out. But a simple number like +10% doesn’t necessarily capture changes well, as it doesn’t take the amount of variance (or the number of data points) into account.
Hi Rain9dome9, I don’t know if you’re familiar with app development but suggesting I ‘rename’ this app is akin to suggesting I start again! I’ve spent many hundreds of hours on this so far, between programming and making content to promote it - so renaming it would mean throwing away a LOT of work.
I’m not really going for Google SEO anyway. My hope was always to get featured by Apple in the App Store (no luck yet sadly). Also, it’s more about the name you see on your home screen than the name you see on Google. But yeah, the reason I say show a search for “changes good to hear” at the end is to try to mitigate the simplicity of the name in a search context.
Yes, if the 10% were the only analysis available, you might have a point, but let me put it in its wider context.
Since the user tracks every entry they will have a good idea of how much data the figure represents
The number is not static - it will evolve over time and the user will develop an intuition about how it changes
There are more detailed data visualisations available within the app that show data aggregated by tags or location, and I have a “my story” visualisation that shows every single data point on a calendar. These are all presented in augmented reality to make it easy to move around and inspect the data - a mobile data visualisation innovation I am quite proud of.
I have spent a lot of time making the spreadsheet data export as useful as possible, including a series of YouTube videos and blog posts explaining how to cross-reference this app’s data with other data sources and visualise the data in different ways to draw different conclusions. (https://www.youtube.com/playlist?list=PLj666SNUOQu2zuQsHqkTF1JzqsJyJnYoT)
There are an infinity of reasons why mood-tracking data can’t ever be analysed in a perfectly scientific way. I have a degree in Physics so I’ve done enough proper science in my time to know this! Happiness data will always be extremely subjective. This is why the emphasis is on the stories we tell ourselves over some idea of a true scientific study.
Looks really cool Michael, I’m scratching my brain trying to think of where would be a good place to share it… I imagine it might take off in active self help communities? Like you could try targeted ads on Facebook for people who are in certain communities? Or even people who gather around Positive Psychology (look up Martin Seligman if you’re unfamiliar).
I bet there’s some inspiration / positivity porn Instagram pages that would happily post / story an ad for a small fee. And you could scale the price point right down to the size of their audience.
Have you got anyone using it yet? I imagine the best kind of advertising for something like this would be people finding it working for them. Maybe if you’ve got an easy way for people to share their data socially like a trophy (murky territory there - don’t want evangelicals) or even a referral system they can use if they like, tell a friend about it or something.
Even a small pdf cheat sheet that helps people overcome the barriers to tracking or comparing metrics, could be a gift for those who get a friend to sign up and make a change - target the people who you already know are willing to go a bit further than the average Joe (because they want to learn metric comparison) and reward them for getting a mate in on their journey.
I imagine it’s a tough space to be in, between marketing and trying to sell positive change because the authenticity of your brand and product comes across really clearly, so finding that balance, well I would hardly find it easy lol.
Maybe you could give the ap to some specific people, or even target like, the layer above your target market. Run a targetted ad on LinkedIn talking to doctors / case workers / therapists. Something like ‘have you got clients / patients right on the cusp of being ready to make change? If you use this code they can get a powerful, simple tool, at a discount’ idk, you get where I’m going with that.
Hi Agaricus, the data is stored as a local Realm database in the app.
There’s a secure private iCloud sync so users’ data won’t get lost if their phone goes missing.
I have three ways of exporting the data.
CSV (in a pretty useful format since I’ve been making these spreadsheet demo videos)
Structured JSON
The Realm database itself
The data export is locked with FaceID. You can export it without entries and tags marked “sensitive” and with location names truncated and geo-coordinates stripped if you ever want to share data with a third party for analysis (or… make a YouTube demo video with your data!).
I’ve saved all your suggestions in my TODO list and will let you know how it works out!
Regarding the social sharing, I have included some features to help people do this - not an easy feat considering the privacy considerations as you rightly point out. I have the concept of “sensitive entries” that are blurred out by default when you look through your data either on your feed or in the augmented reality data visualisations (and in the data exports). This was partly to make it easier for people to show off their data on social media. The AR visualisation stuff basically functions as a camera app so I’ve been hoping to see some interesting pictures come up on social media (not that I have checked for these yet!).
I have an eBook “Tracking Happiness” which functions as a kind of cheat sheet - it’s a ‘lead magnet’ for people to sign up to my mailing list where I post weekly(ish) advice on mood tracking and decision/change-making. The referral idea is interesting though I always have to navigate Apple’s App Review policies and there isn’t a good way to give different people different prices.
The tricky thing about all this is that it has a implicit slow-burn to user adoption. Nobody is going to download it and immediately recommend it - they need to use it for a while before they see the real benefits. I’m just hoping that as time goes on, I can gather momentum. But it’s hard because I don’t have a lot of patience for “selling” stuff - I’d rather just make more stuff!
True, but with this new screen you have opened the science box
Given your background, you’re familiar with using a p-value to judge if the 10% is just random noise. I prefer confidence intervals, as they are a bit easier to interpret. Or you could calculate a Bayesian “likeliness”, which probably makes the most intuitive sense (see e.g. SleepCoacher)…
Maybe this is clearer if I show you what it looks like with debugging turned on. I’m fitting a trend line to the low-pass-filtered data - the percentage is the resulting trend line’s gradient.
Cheers @michaelforrest I’d be real curious to hear what the results are if you did decide to try any of those ideas. I wonder have you had much of a tracking mentality applied to your business metrics?
Think in terms of ‘campaigns’, with specific strategies behind each sprint, set budget (money and time) etc?
I’ve been wondering about tracking social media growth and taking that every day scientific approach to the growth of a brand. A/B testing etc.
It’s interesting, the tracking mentality around marketing.
It feels more like obsessing over Instagram likes than gathering data toward a scientific end.
I do have one app that took off earlier this year and I had to limit myself to looking at the metrics once a week because it was interfering with my concentration so much (particularly with the App Store analytics coming out at an unpredictable time during the day).
It’s quite hard to start drawing conclusions with low volumes of traffic/sales/engagement. I think until I’m regularly spending money on advertising, A/B testing will be of limited benefit. And working alone it’s hard enough to put together one version of an ad, or piece of copy, let alone multiple variations!
Yeah I think the science would come in if I had money to spend. For now I’m still just having to throw things out there and see what happens.
Had a great initial response to this video anyway. Seems to have tailed off now so I guess the viral coefficient was <1.0 (or whatever they call it ).
Each “change” in the app results in a new trend line being started. This was kinda that whole point of the video, although perhaps I’m so close to it that I can’t tell whether that comes across! But yes, I’ve made three changes (the first trendline is before the first change) and the point is to get an overview of whether each change was beneficial or detrimental to my overall happiness.
If your mood was stable at 5.0, then jumped to (and remained stable at) 7.0 upon doing a change, would you be showing 0% (the new gradient) or +40% (the mean difference)?
If your mood was completely stable over any period you would see 0% for that period. But it doesn’t really make sense to think in absolute terms about these numbers. Most entries will have a score of either -1 or +1. The trend is calculated by iterating through all the entries with a calculation like this:
let runningScore = 0;
let smoothingFactor = 0.98; // closer to 1.0 = smoother line
for entry in entries{
runningScore = (runningScore * smoothingFactor) + entry.rating; //# -1, 0, 1; or sometimes -2 or 2 for "high points" and "low points"
entry.trendlineValue = runningScore;
}
Positive entries will push the trend line up, negative ratings will pull it down.
If you just looked at the raw data you’d see a chaotic sequence of discrete 1 or -1 (or 0) values. The same regression would fit that data as fits the smoothed data though, it would just feel less intuitive.
The concept of ‘changes’ is used to carve up the data.
While the trend line will be affected by all previous entries, the “change period” calculations are made in isolation. It is then up to the user to compare values, for example: “when I broke up with this person my mood went down by 23% but when I quit drinking it went up by 11%”.
I would never expect somebody’s mood to remain stable over any period of time - the times when you report your mood will be spread through a sea of momentary successes, failures, annoyances, unexplained anxieties, pleasurable experiences - moments that we usually forget. This is the key idea of the app - to collect these moments and look at them objectively instead of letting our squishy brains smooth out the data in idiosyncratic ways. Because we’re bad a remembering our emotions, especially when we’re currently experiencing a stronger feeling.
Then in all listings of QS apps your app I will name “changes” as the “michaelforrest app”. See my data flow graph project. Seriously, your use of a very common term and nothing else puts you in the same category as dozens of no effort throw away calorie counter apps. See for yourself: https://nutritionj.biomedcentral.com/articles/10.1186/s12937-018-0366-6#MOESM1
I think the easiest way to think of this is as an ongoing accumulation of happiness or unhappiness. If you were happy every day you’d see a rising line, yes. If you were unhappy every day you’d see a falling line. If you reported neutral every day you’d see a horizontal line at y=0. This visualisation means I can show more nuance and make better use of the vertical axis than if I tried to simply plot absolute ratings over time. Because the data points are pretty discrete, you’d just see a kind of “sample and hold” wave which wouldn’t tell you very much. In that case it would be easier just to show three numbers - the counts of each rating type (14 happy entries, 23 neutral entries, 12 unhappy entries) - but then you wouldn’t be able to see how things are changing over time.
Obviously this idea of “accumulating happiness” is quite subjective and open to myriad philosophical objections. All I can tell you is that in practice, it fits with my intuition about the data and reveals things I might otherwise have missed.
Unfortunately it’s hard for me to know if this works for other people in general because my privacy policies have always meant I can’t really see other people’s data. With Changes I ask for permission to see into people’s data a little bit so hopefully, in time, I can learn a bit more about how this works for others.
