TinyPilot: Month 7

TinyPilot is facing good problems to have.

Highlights 🔗︎

• TinyPilot achieved astonishing growth in revenue, jumping from $15k in December to $42k in January.
• Most of TinyPilot’s sales came from a single positive YouTube review.
• TinyPilot is experiencing growing pains as I scramble to meet demand.

Goal Grades 🔗︎

At the start of each month, I declare what I’d like to accomplish. Here’s how I did against those goals:

Hire a freelance developer to help with TinyPilot development 🔗︎

• Result: I ran one trial hire that didn’t work out, but I’m currently trying with another.
• Grade: A

Hiring a new developer is a slow process, but it’s going according to plan. With each trial hire, I’m finding ways to help them ramp up faster. It also helps me recognize what qualities I should look for in future hires.

Receive TinyPilot reviews from two bloggers or YouTubers with a relevant audience 🔗︎

• Result: Received one YouTube review, and two others are in progress
• Grade: A-

I underestimated how long it takes for YouTubers to make new videos. One new review came out, but it’s based on outreach I did in November. Two other YouTubers agreed to review, but they’re still working on their videos.

That said, one review turned out to be almost too much to handle. More on that below.

Earn $4k in revenue from Hit the Front Page of Hacker News 🔗︎

• Result: Earned $2.6k in revenue
• Grade: C

The course hasn’t quite played out the way I’d hoped. I’m proud of the material, but far fewer people are purchasing it than I anticipated.

I tried several different techniques to market it, but none of them have had a noticeable effect on sales.

TinyPilot stats 🔗︎

Metric December 2020 January 2021 Change

Unique Visitors 3,486 11,249 +7,763 (+223%)

Total Pageviews 5,785 17,737 +11,952 (+207%)

Sales Revenue $15,358.31 $41,992.92 +$26,634.61 (+173%)

Donations $9.00 $0.00 -$9.00 (-100%)

Total Revenue $15,367.05 $41,992.92 +$26,625.87 (+173%)

This was a enormous month for TinyPilot. Sales jumped from $15k to $42k.

Every month, I think, “Wow, that was a lucky month! I won’t be able to do that again.” And then there’s even more growth the following month, so here’s hoping that luck continues.

TinyPilot’s first YouTube review 🔗︎

The majority of January’s sales is due to just one source: Craft Computing, a YouTube channel about IT hardware for professionals and home enthusiasts.

I discovered the Craft Computing channel early last year through its excellent tutorial on installing Proxmox. A few months later, the host announced that he had quit his job as an IT manager to work for himself, so I felt a kinship there.

In November, I reached out to Jeff, the channel’s host, asking if he’d be interested in reviewing TinyPilot. He agreed, so I quickly shipped one over. I didn’t hear anything for a few months and was beginning to wonder whether my Voyager was just sitting in a heap of free hardware that companies send him.

Just as I was getting ready for bed on January 12th, Jeff emailed me to say the review just went up.

It was surreal. My girlfriend and I watched it on the big TV in our living room, so I was seeing a YouTube personality I’d watched for a year holding my product and speaking directly to me from a huge screen!

Jeff loved TinyPilot and shared the many ways it impressed him. I was anxious the whole way through that he would suddenly stop and say, “Actually… I changed my mind. This product sucks. Tell everyone you know not to buy one.” But fortunately, the review was positive throughout, and his critiques were minor.

The results were strong and immediate. The next day, I received 33 orders and earned $5k in revenue, far exceeding previous sales records. The next day, the number of orders dropped, but more customers purchased high-end kits, so the total revenue remained roughly the same.

TinyPilot revenues, orders by day, January 2021

I expected things to subside on the third day, and they did a little bit, but they kept coming. My inventory is set up to absorb a rush like this, but I didn’t expect it to continue for so long.

TinyPilot’s fulfillment manager and I kept up with orders for the first week, but on day eight, we’d been run ragged. We ran out of a few parts we couldn’t replace within our two-day shipping window, so we briefly listed most products as backordered. A few days later, we were back on our feet, and the order volume got back down to manageable.

TinyPilot’s first postmortem 🔗︎

Blameless postmortems are one of the most valuable practices I learned while working at Google. When something goes majorly wrong, you get back to steady-state and then write a report analyzing what happened.

As the name implies, the document is blameless. It’s never, “We had an outage because Michael’s an idiot, and he deleted the wrong file.”

The underlying assumption of a postmortem is that everyone on the team is smart and diligent, so if something went wrong, the systems failed, not the people. When a failure looks like someone being stupid, the root cause is that processes failed to protect against expected human error.

After the Craft Computing review, my fulfillment manager and I conducted TinyPilot’s first postmortem. Below are the major issues we discovered and what changes we put in place to mitigate them.

Inventory targets were too low 🔗︎

We manage inventory with a spreadsheet. For each part we carry, we define a minimum and maximum to keep in stock. For example, Raspberry Pis had a minimum of 40 and a maximum of 80. That means that anytime our inventory of Raspberry Pis drops below 40, we order enough to get our supply back to 80.

TinyPilot’s inventory spreadsheet

The problem was that we only adjusted these targets until we run out and realize we planned poorly. Then, when we’re in a shortage, we’re so panicked that it’s hard to think rationally about what the new targets should be.

• Fix: Schedule a monthly adjustment of inventory targets based on recent order volume.
• Fix: Assume the next surge in orders will be more intense than any that have occurred previously.
• Fix: If we ever run out of a part or experience a close call, double its inventory targets.

Urgency was not obvious 🔗︎

Our standard workflow with the inventory spreadsheet is to “top up” any item that falls below our minimum. When we dipped below the reorder threshold for USB cables, we ordered new ones in our usual workflow, with standard ground shipping.

We didn’t notice that customers were purchasing so quickly that we were on track to reach zero before the new shipment would arrive. The next day, we placed an additional order with overnight shipping to bridge us to the later shipment.

• Fix: Highlight numbers in the “in stock” column when they reach dangerously low levels.
• Fix: Upgrade to Amazon Business Prime to make two-day shipping a no-brainer.
• Fix: Look for inventory management software that includes alerting based on trends.

Handling time is too short 🔗︎

When I first started TinyPilot, I promised to ship devices in one business day. It immediately became clear that a one-day turnaround was unsustainable, as it constantly forced me to context-switch and rush orders out. A few weeks in, I changed the advertised turnaround time to two days.

After the YouTube review, we had to list items as backordered after a week of scrambling, but if our shipping promise was three days, that would have been enough buffer for us to have avoided backorder. We realized that about 80% of our previous backorders or close calls were off by just a day.

The two-day window was creating too much stress. We ship most orders out next-day anyway, and customers probably don’t care much about an advertised turnaround of two vs. three days.

• Fix: Increase advertised handling time to three days.

Reordering packs of parts creates unnecessary cognitive load 🔗︎

We purchase some of our parts individually and some in packs of two or three per package. For example, we purchase our power adapters in packs of three. Our inventory spreadsheet lists absolute quantity of power adapters, not packs of power adapters.

During the rush, we meant to order 150 power adapters, but we accidentally ordered 150 3-packs, so 450 total.

The over-order was a human mistake, but it highlights how our system invites error. The person placing the order needs to do mental arithmetic that varies for each item.

Fix: Add a column in the spreadsheet for “reorder quantity” that factors in quantity-per-pack.

How can TinyPilot scale? 🔗︎

TinyPilot is growing far more quickly than I anticipated. Fortunately, my existing processes have accommodated the growth so far, but if we continue, things will begin buckling under the load. Here’s how I’m planning to scale growth across a few dimensions.

Scaling support 🔗︎

As the number of TinyPilot customers increases, I’m also receiving more support emails. Some days, I spend up to half my working hours on technical support.

I thought back to a blog post from Jen Yip, founder of Lunch Money. She manages all aspects of her business by herself, and she minimizes her support burden by implementing features in her app that help users resolve common issues.

About half of TinyPilot’s support requests are actually just gaps in the product. A common request is, “How do I turn on WiFi?” Right now, I send them a link that explains how to do it from the command line, but this should be something users can do right from the TinyPilot web interface.

As I continue developing TinyPilot’s software, I’m prioritizing features that eliminate user confusion and making debugging tools more accessible when things go wrong.

Scaling manufacturing 🔗︎

3D-printed parts are great because adjustments to the design are fast and inexpensive. The downside is that 3D printing scales poorly. The lab that makes my Voyager cases can produce only 40 per week with the material I want, and that will soon become a bottleneck.

The next step up is injection molding. Basically, the manufacturer creates a steel or aluminum mold of the case, fills it with plastic, then presses it into shape. Creating the mold is slow and expensive, but once they create the mold, they can produce 1,000 per day at low costs. I requested quotes from several vendors, but they came back in the range of $20-40k, a bit too high at this point.

I reached out to other labs to see if I could parallelize the 3D printing. The problem is that the material I like is a carbon fiber material that’s uncommon among 3D printing labs, so not many other vendors carry it. The ones that do quoted me prices that are 10-15x higher than I’m currently paying. Still, Voyager’s margins are high enough that if push comes to shove, I could turn a profit even with higher case costs.

One of the other labs told me I could achieve a similar result with an alternative material that prints faster. I’m requesting a sample from them and talking to my primary lab about alternative materials.

My girlfriend slash inventory manager handles the final manufacturing steps, such as testing circuit boards and assembling Voyagers. We may reach a point where she doesn’t have enough time to do order fulfillment, inventory management, assembly, and other tasks on top of being a full-time grad student. So, we’re reviewing which tasks we can outsource when we reach the limits of her available work time.

Scaling development 🔗︎

With all the other parts of running TinyPilot, I’ve sadly had scarce time to work on the software. In December, I began looking for part-time developers to pick up the slack. I prepared these two documents to describe the job and shared them in a few small channels:

Out of all the freelance positions I hire for, software development is always the biggest challenge. I have high standards for software, so it’s difficult to find developers who have a similar passion for quality and maintainability.

Instead of interviewing candidates, I go straight to trial hires. The first hire didn’t work out, but they told me they were pleased with the process. It also helped me find ways to make it easier for new developers to ramp up on the codebase.

Legacy projects 🔗︎

Here are some brief updates on projects that I still maintain but are not the primary focus of my development:

Hit the Front Page of Hacker News 🔗︎

Metric December 2020 January 2021 Change

Unique Visitors 2,595 1,042 -1,553 (-60%)

Total Revenue $1,431.00 $2,565.22 +$1,134.22 (+79%)

Hit the Front Page of Hacker News, my course about blogging for developer audiences, finally launched in January. There was a jump in sales around the launch, but it was significantly less than I expected. My goal for the course was to earn $20k by the end of the year. That’s looking unlikely.

When I told Cory Zue that I was disappointed in sales so far, he teased me about how much TinyPilot’s earnings have jaded me. The course earned more in January alone than I earned my whole first year as an indie developer.

I’m planning to write a retrospective later this month that’s focused entirely on what I learned from recording and marketing the course, so stay tuned for a more detailed discussion.

Is It Keto 🔗︎

Metric December 2020 January 2021 Change

Unique Visitors 49,373 80,177 +30,804 (+62%)

Total Pageviews 93,242 182,367 +89,125 (+96%)

Domain Rating (Ahrefs) 10.0 11.0 +1.0 (+10%)

AdSense Revenue $334.72 $677.36 +$342.64 (+102%)

Amazon Affiliate Revenue $149.99 $238.02 +$88.03 (+59%)

Total Revenue $484.71 $915.38 +$430.67 (+89%)

Is It Keto had a record month even though I didn’t touch the site at all. Because it’s related to dieting, there’s always a spike in interest that coincides with new year’s resolutions.

Zestful 🔗︎

Metric December 2020 January 2021 Change

Unique Visitors 507 419 -88 (-17%)

Total Pageviews 1,511 1,194 -317 (-21%)

RapidAPI Revenue $103.33 $155.50 +$52.17 (+50%)

Total Revenue $103.33 $155.50 +$52.17 (+50%)

Zestful is growing slightly, even though it’s in maintenance mode. The growth seems to be from a single long-standing customer whose usage has increased.

Wrap up 🔗︎

What got done? 🔗︎

• Hit the Front Page of Hacker News launched.
• TinyPilot Pro graduated from beta to a real relase.
• Conducted TinyPilot’s first postmortem.
• Hired two freelance developers on a trial basis.

Lessons learned 🔗︎

• YouTube reviews can be incredibly powerful, but they take time.
  • Seeing the results from the first review, this is definitely a source I want to explore further, but it will take a few months to line up new reviews.
• As you grow, review the assumptions behind your processes.
  • This is a lesson I have to learn over and over. I often establish a process for some task, and it becomes a habit. But then I stop thinking critically about the process and fail to recognize that I designed it for conditions that might no longer be true.
  • It’s important to take a step back regularly and review whether the process is still optimal.

Goals for next month 🔗︎

• Attract five bloggers or YouTubers to a TinyPilot affiliate program.
• Add two features to TinyPilot that reduce support or manufacturing costs.
• Collect feedback from 10 customers about a potential rack-mounted version of TinyPilot.