By Max Sherman ยท December 4, 2023
MinutesGenerator has revenue!
Grade: F
I didn't end up doing this, but I did find more clarity on my ideal customer profile, namely: administrative professionals. This can include virtual assistants, but isn't limited to them. I'll talk about future plans for marketing later. Ultimately, I decided I can find a more efficient way to reach audiences.
Grade: A-
The real value of doing this is to decrease the time to delivering value, resulting in a better customer experience, resulting in less churn and more revenue.
I removed the inital text prompt and the transcribe button - just upload your file and things will start automatically. You literally have to do one thing and it's really easy. I spent a lot of time optimizing the transcription flow after upload. I rented a GPU server so I didn't have to wait for my API provider's cold start boot times. I started streaming back results to the client from the server as soon as they were ready, instead of waiting for everything to be ready before showing. I created a progress bar that communicated where in the process the job was, and gives an ETA for being done. Other stuff too, but ultimately the take away is that people hate waiting, and the faster my service gets, the better experience people have. That reduces churn, and boosts the potential for sales. Waiting sucks and speed is a feature.
This goal is an A- not an A because I am almost certain every successful transcription in the current system is well under 10 minutes. However, I don't yet have analytics on how long each step in the process takes, so this is based on usage I have seen personally. I keep watch very diligently, so I am confident we're there, but speed is so critical that it has to be measured and tracked.
The north star is to create the best software in the world for generating meeting minutes, and that software should produce instant results.
All the optimization I did in November was such low hanging fruit that I didn't need to measure to see performance improvements. The improvements were obvious. Going forward, I have optimization ideas that are more experimental, and I want to track perf metrics more rigorously to make sure everything gets faster and stays fast.
Grade: A
I made four sales this month! All with zero marketing - customers came through organic bing search. I was worried about this goal at the beginning of the month because it's more of a lag measure than a lead measure, for instance I can control sending X cold emails in a month, but it's harder to control the outcome of that activity.
I had a tiny bit of confidence that sales would come because I could see people signing up for MinutesGenerator, but then hitting issues with the product and then leaving the site. The biggest issue I noticed was that it took a long time to receive any value from the site. A user would sign up, and then be faced with three things to do: upload a recording of their meeting, write a text description of any names or proper nouns in the meeting, and then click a button to start transcribing.
People would do these steps, which is a testatment to how much they do not want to write their own meeting minutes. After clicking the button, they would see a loading spinner and transcription would take on average 45 minutes to finish. Super slow! There was no feedback things were working, most users assumed the product didn't work and would leave the site (many times this assumption was correct and no error was surfaced).
I realized that nobody wants to pay for a shitty product that doesn't work (duh). I also realized that I can make the experience more compelling by doing two things:
First, speed it up. This is what the last section is about so we already know about it.
The second strategy is to show results to the user as early as possible. If the customer sees the first few seconds of audio transcribed, they get more confident things will work. I think once they get invested, a better outcome is more likely.
I changed the product so that it produces a partial transcript, and if the audio is longer than the free trial allows, an upsell pricing table appears asking you to upgrade your plan to continue transcribing. The partial transcript shows the user this will work, which boosts confidence, which boosts likelihood of them paying. That's the theory anyway, and it seems to be working so far!
Grade: A
I averaged 20.9 hours worked per week, slightly above my goal. This goal combined with live streaming my work has been the most transformative for my productivity. It's nice to have something measurable I can track.
I get a steady stream of real user feedback now and it's been helpful to see where the product breaks. My mental model is that MinutesGenerator is a machine that other people interact with. I input money and time into the machine, and users input files, clicks, scrolls, etc. into the machine. The machine outputs meeting minutes and also sometimes the machine outputs money. Most of the time something gets messed up, the machine breaks and outputs nothing and the user leaves unsatisified.
I've started looking at every visitor to the site as beginning a sequence of hundreds of small steps. Every step has to go perfectly in order to make money. Any of these steps can break, for example the user uploads a file and it takes too long, or they click a button and there's a UI bug, or they pay money and our stripe webhook handler isn't set up correctly, or they read something on the landing page and got confused, etc etc etc.
It doesn't matter if a failure happens on the first step or the 100th step - if something fails, it's game over and you make zero money and probably pissed someone off.
Because of this, I look very closely at how every user's experience is. If they encounter any failure at all, I immediately prioritize a fix so it can't happen again. I have done this for one month and there are still problems, but I can see the improvements reflected in sales numbers as well as successful meeting minute generations.
It's tempting to think that you can create a machine that works well without observing user behavior, but I think it's extremely difficult. For example, here are some differences between you and a new user - you may have a different OS, browser, internet speed, screen dimensions, mobile device, familiarity with online tools, file formats used with testing, expectations for what meeting minutes look like, expectations for meeting minutes file format, reading level (or else bias that your own copy writing is easy to understand), and on and on and on and on.
One arm of my strategy is building a best in class product that users have an overwhelmingly positive reaction to. I want churn low and to retain users forever.
The other side of the story is the process of making known to strangers. To do this, I'm going to try and form some affiliate partnerships with creators who have audiences that have administrative professionals. I spent some time this month building a list (and a system to build lists) to do some email outreach. I'm currently warming up some email accounts to start doing cold outreach, and I'm hoping that I can land a few partners and drive hundreds or thousands of sign ups.
For example, if I can get a few creators to produce videos that get total 10,000 views, and this has a 1% conversion rate, then this would be 100 sign ups, and would 25x my current sales numbers. Part of this plan is going to include raising prices, and offering exclusive affiliate discounts. If I can generate $15/month subscriptions, then this would make MinutesGenerator profitable with an $18,000 annual recurring revenue.
In general, the plan is to build out, optimize, and automate customer acquisition channels. Right now organic search is my only channel. Becoming more well known will help this channel's efficiency improve via higher search result ranking.