Since May, I have been working two days a week for a startup called Soultime.
Soultime is an app that helps Christians with their spiritual and emotional health. It has meditations on a range of topics including anxiety, stress, sleep, identity, relationships and emotions. The app also has calming music, scripture readings and sleep stories. And it helps you to identify and track your mood so that you stay better connected with your emotions.
It’s a bit like Calm, but for Christians.
Soultime’s mission is to help people better connect with themselves, others and God.
We’re a small team of four people, based in Vauxhall, London. We also have a small outsourced team of software engineers, based in the Ukraine, provided by a company called eKreative.
How I Got Involved with Soultime
A year ago, Mark Wagner, who founded and runs Soultime was introduced to me by a mutual friend. Mark used to run his own hedge fund and also started a ministry focused on inner healing. So he has a strong business background with good domain expertise in the area of spiritual and emotional health. But he and the team didn’t have a strong tech background. That’s what Mark needed.
When we first spoke, I was heavily involved in my education project. But as that wound down, Mark and I reconnected and I decided to join Soultime part-time. For me, it had a compelling mission focused on social impact, a great team to spend time with (as the entrepreneurial journey can be quite lonely) and it would help pay the bills!
The Journey of Soultime
Mark launched Soultime in July 2018 after 9 months of development. So when I joined in May 2019, the app was 10 months old. Since I joined 10 months ago, we’ve about doubled the core metrics – downloads, monthly active users and number of paying users (a subscription is about £60 / year).
But more importantly we’ve got to a place that will enable us to scale dramatically. This has been a function of:
- Acquisition costs (how much we pay to acquire a paying user)
- Lifetime value (how much money we’ll get from a user over their lifetime)
- Cashflows (what growth we can sustain with the cash we have and the amount we make back with immediate subscriptions revenue, paid about 45 days later from the app store)
I’ll get into all of these, but first let’s begin when my Soultime journey started last May …
Building a Better Picture
Mark and I first met over lunch at the start of May 2019. I remember being quite surprised by how far he and his small team had come, despite them not having a tech background. The most obvious gap though was a lack of understanding about how their users were using Soultime. They had some basic analytics in the app (provided through Firebase), but that only gave them indications of the number of active users. It didn’t tell them what those users were doing. And no-one had sat down with their users to see how they interacted with the app.
So my first goal was to build a better picture through a range of qualitative and quantitative methods.
Working on an App Redesign
The team were about to embark on an app redesign and had mocked up how they envisaged the new version looking. The main focus was on building a better daily flow for users.
The current version in May 2019 had a “Daily” tab that some users were using. This comprised of a bible reading, a meditation from the existing catalogue and a music track. The team planned to introduce a new shorter meditation every day and encourage users to track their mood each day.
This redesign gave me the opportunity to properly review how the app worked and help the team figure out the most impactful changes, and how to prioritise the work.
We started by reviewing their existing proposal with the engineering team to assess how much work it would involve. Part of that work was to improve the overall navigation of the app. We wanted a consistent bottom navigation and a persistent audio player. We prioritised this work as it would improve the navigation of the app. This gave the engineering team a focus while we had time to user test the rest of the team’s proposal.
We then assessed the new proposal and considered what the flow should be, especially for a new user. After some discussion and brainstorming, we decided to use the home screen of the app to show the actions we wanted our users to perform on a daily basis. Specifically, listening to a daily meditation and tracking their mood. After further brainstorming, we decided to reward users who completed their mood check with suggested meditations and bible readings based on their mood.
Qualitative Learnings with User Testing
Mocking up a Prototype
We then mocked up our best attempt at our new user flow using a tool called Marvel. This enabled us to build a prototype that users could try. Marvel is basically a series of screenshots that you can navigate between with clickable buttons. It’s much quicker to put together than building the functionality in the app. And it’s something we could do independently of the engineering team.
Finding the Right Users
We gave the prototype to users and observed how they behaved. Selecting the right users was really important. We needed Christians who had never used Soultime before so we could see how they behaved as new users. After quite a bit of thought and reaching out to a number of local churches and Christian organisations, we managed to secure a slot at a Christian conference the following week. It turned out to be perfect for what we needed …
We had a sofa in the corner of the breakout area of the conference. We put up a big sign indicating that we were running user testing sessions. And then we asked people to spare 20-30 minutes and offered them a free subscription in return. Over the day we conducted nine user tests. We ran these in pairs – one person leading and the other writing notes. We also listed seven hypotheses to test. These included whether people could onboard without issue and whether they listened to the introduction as their first action.
Learnings & More User Tests
We learnt a lot from this day of user testing. Overall, the onboarding flow and new home screen worked really well. But we discovered that the introduction was unnecessary. It would be better to just direct new users straight into their first daily meditation.
We ran a couple more user testing sessions over the following few weeks, including observing people using the existing version of the app. We noticed that users got lost navigating the app, especially while listening to an audio track. This further validated our work to improve the navigation.
Quantitative Learnings with Mixpanel
In addition to the qualitative learnings from user testing, we also implemented much better analytics in the app so we could learn more about our users’ behaviour using data.
We decided to use the market leader in analytics – Mixpanel. It is quite expensive, but as a startup they gave us a 90% discount, which made it affordable.
I spent some time carefully crafting the schema for how we should track the events and people in the app to empower us to do the analysis we needed to do, both now and in the future. This implementation took some time, but has been worthwhile.
We were immediately able to build funnels to see where users dropped off during the onboarding flow. This was game-changing as we had no visibility of this till now. We could see how many users were doing each of the core actions we cared about in the app every day. Such as daily meditation listens, mood checks, normal meditation listens and music listens. We were able to look at purchase funnels to see what caused people to pay for a subscription. And we could see which meditations were most popular and most frequently listened to.
We were also able to evaluate the impact of this new big redesign of the app. The new daily meditations proved to be very popular, with more unique people listening to these every day than any other core action in the app. The % of new signups who listened to their first audio track also improved by 60%.
And as the team came up with new questions over the next few months, we’ve often been able to get answers from Mixpanel within 10 minutes.
We also put together a Typeform survey to ask for people’s opinions on the meditations and their experience of using the app. This added further flavour to what we had already learnt from our user testing and Mixpanel data. This mostly validated that our insights were accurate. The comments that people wrote about the impact of Soultime on their lives were also really encouraging to us.
Soultime Product Direction
In addition to the changes in the app’s design and navigation, the learnings that we gleaned from the methods we employed informed our product roadmap for the next few months. We introduced a sleep timer for our music and bible readings as people often listened to these to help them get to sleep. We introduced ratings after each meditation so we could learn which meditations people preferred. And we started recording our own sleep stories so we could have more content to help people fall asleep.
We also worked hard to improve our conversion rate to paying subscribers. This was necessary for us to scale and build a sustainable business that has a long-term impact on many people’s lives. Small improvements here can make all the difference.
We tried lots of things. Some things worked and some things didn’t. One thing we did recently that worked really well was introduce a monthly subscription, in addition to annual. This about doubled the number of new subscribers we had and didn’t really cannibalise our new annual subscribers, which we were concerned about.
Creating a Full Audio Version of the Bible Using AI
Another piece of feedback we heard a few times was that people wanted to listen to the bible instead of just read it. We only had a few bible passages recorded and we estimated that it would cost at least £50k to have someone read through and record the whole bible.
But then we discovered Google Wavenet. This reads out text using a synthetic voice. Text-to-audio engines have been around for decades, but with recent advances in AI, Google’s latest Wavenet sounds pretty realistic.
Using Python to Pull It All Together
So I built a python script to take an entire textual copy of the bible (using a free version called the New Heart English Bible) and split out each book, chapter and verse. I then fed each verse through Google Wavenet using their API to generate synthetic recordings of each book of the bible. I proof listened to about 5% of the entire bible and corrected about 20 instances I noticed of incorrect readings. For example, “In His name’s sake” was pronouncing “sake” as “saké” (the Japanese beverage!). This led to humourous listening, but needed correcting! In this case, I spelled this word “sayk” and re-fed it through Wavenet, which produced an accurate pronunciation.
Finally, I used a set of tools called SoX to programmatically mix audio tracks with the bible readings so they would more closely resemble the other audio in the app.
This project took about 5 working days and cost $100 in Google Cloud fees to process the whole bible. Plus, we could easily produce any bible version, translation or even accent. Each full bible version would only cost another $100.
We haven’t yet put this in the app, due to other more pressing development priorities. But this should be released at some point. In the meantime, here’s an example reading of 1 Corinthians 13.
Building a Soultime Chatbot
Another project I worked on was a Soultime chatbot. The goal with this project was to see if we could get lower acquisition costs by engaging people first with a chatbot and then seeing if they installed the app and became a paying subscriber after.
I ported a lot of code I had already written from my previous chatbot projects (Malaria Challenge and Whack-a-mole), which gave me a head start. I then worked with the team to craft a flow that closely mimicked the process we take people through on the app’s guided meditations. Essentially, we help people identify how they are feeling and what caused them to feel that way. We then help people deal with that emotion by helping them identify an untruth they are believing or by forgiving someone.
We spent about £10 / day on Facebook ads, which gave us about 20 new users every day. One of the great things about chatbots is that you can see the entire conversation, which reflects the entire user experience. This helps you understand very quickly what works and what doesn’t, which helps you learn and iterate rapidly. We iterated on the experience about five times.
Unfortunately, not many people went on to install the app. So we didn’t achieve our original goal. But what surprised us was the level of engagement in the chatbot and how many people shared very difficult things. We’ve learnt that a chatbot experience can be a very effective therapeutic tool. So we may explore bringing this experience into the app at some point in the future.
Improved Working Practices with the Soultime Engineering Team
We also improved how we worked with the team in the Ukraine over the past few months. Historically, the team in Vauxhall worked with the Ukrainean team in a very transactional way. They would give them a detailed feature specification, the team would quote how many hours it would take to build, we would approve and the team would build it.
We moved to a more collaborative way of working. We introduced daily standups with the whole team. This gave each individual an opportunity to share what they had done the previous day and what they were working on today. And we discussed each new feature with the team to figure out the best approach, considering technical complexities and the goals we had for each given feature. We also gave the engineering team better insights into why we were building each feature and the direction we were heading in. This enabled them to take better ownership of each feature and its success.
Fundraising & Hiring
We recently closed a round of funding. I helped a little with this with some preparation on the pitch and also some conversations with investors. I also spent some time exploring bringing engineering in-house into our London office, which would improve our speed of execution and agility. But the costs would at least double from what we are currently paying eKreative. So we’d need to scale quite a bit more before we can consider this.
Improved Paid Marketing
The biggest challenge we had on growth when I joined was not being able to scale our Facebook advertising spend. During the summer, we encountered an agency that specialises in Facebook ads, through a mutual friend of mine. They have enabled us to maintain a consistently low cost of acquisition that we’ve been able to scale. We’ve increased our ad spend from £250 / day to £1k / day in the past couple of months. And we now have confidence that we can scale it further. This has been instrumental in our ability to grow and break even as a business.
Building a Soultime Business Dashboard
Challenges & App Figures
Over the past month I’ve been working on building an automated dashboard to display all of our key business metrics. This includes daily metrics on the number of subscribers, ad spend, cost per subscriber, revenue and costs per day including profit and our conversion rates. This is complicated as we offer a free 7-day trial for annual subscribers. So we need to use past conversions to predict how many of those will convert to paying subscribers for a fair comparison with new monthly subscribers on a given day.
It is further complicated by the fact that the metrics come from two different app stores (Apple and Google). There are many complications around payments being canceled etc. And our ad spend data comes from Facebook. Thankfully, we use a great tool called App Figures. This pulls in data from all of these sources, taking care of many of these complexities and presents it in their dashboard. However, most of the analysis we need isn’t provided. But thankfully they provide an API that lets us programmatically pull data out of App Figures into our own dashboard.
So I used the App Figures API to pull every data point available into a Google Sheet using Google Apps Script. We then did all the number crunching we needed to do in this spreadsheet and exported the data to a Google Datastudio dashboard. This lets us see everything we care about on a single dashboard, with controls to change the dates so we can see whatever historical data points we care about. It’s been a game changer for us!
A Healthy Business
The holy grail for any startup is to measure lifetime value (LTV) against customer acquisition cost (CAC) in real time. We’re very close to being able to do this. The hardest part is calculating churn, which you need in order to calculate LTV. The complexity lies in figuring out the churn curve in the future so we can estimate how long our users are likely to stick around for. As it’s a curve and not a straight line, the maths is quite complex. We’re close to figuring this out though. And then we can track LTV vs. CAC in real time.
A healthy business has an LTV:CAC ratio of about 3:1. Soultime’s ratio is now much better than this, which is very exciting. We also make back more than what we spend on ads within 45 days (i.e. from upfront annual subscriptions and a little from monthly subscriptions). So we should be able to scale dramatically in the weeks and months to come.
My Future Commitment to Soultime
I’ve just reduced my time that I’ll be working on Soultime from two to one day a week. This is so that I can spend more time building my startup, Would You Rather Be. As I mentioned in my last blog post, it’s been quite stressful working on two projects at the same time. So I want to create more time and space to focus on Would You Rather Be.
I’ve made some great progress on my startup since October and am hoping to raise some money in the coming weeks. I’ve also had 10,000 users complete the app in the past 6 weeks, which has validated the concept. It has also given me a lot of data points, which I can use to apply AI (essentially train a Machine Learning model) to improve the career discovery process even more. I’ll share more on these exciting updates in my next blog post!
I’ll continue to work with Soultime one day a week for now. But this will likely be a little more advisory and a little less hands-on.
It’s been a wonderful ride over the past 8 months and I’m really excited to see the impact Soultime will have in people’s lives as it scales and supports Christians all over the world with their emotional and spiritual health.