Two years between a note and a seed round

In January 2019 my Apres notes included this sentence: “We’re about to launch the product next quarter and are just taking meetings now to open our seed round.”

The seed round was announced in January 2021. Two years, almost to the month. This is what happened in between, as far as my notes and the public record show it.

The bar I wrote down

The same notes have a short section titled “When to raise a seed.” It is a rule of thumb I had saved for business software sold through a high-touch process: at least one pilot of a software product (the note rules out consulting work) worth $50,000 or more, already underway, plus some evidence of a sales pipeline, such as one or two signed letters of intent.

A few headings later, under “Where is the proof that it works?”, is what we had:

• Customer interviews and tests with our target market, ending with about 150 to 200 signups from companies and entrepreneurs.
• A first paying customer before we started building.
• Smaller clients and larger pilots that took shape during the build.
• “We did as much work manually as we built out the product.”

Read side by side, the two sections don’t line up. The bar was counted in dollars and signed letters. Our proof was counted in signups and customers, with no pilot value written down. I had the rule and the evidence on the same page and didn’t put a number from one next to the other.

What happened in between

We launched in June 2019, as a platform that creates AI training data using AI. We put it on Product Hunt the same year.

Then the company changed direction. My plan for January 2020 opened with “Explainability V1.” That spring Apres went through Techstars in Lisbon. When Techstars ran a remote demo day, TechCrunch named Apres among its top ten startups. Matt Turck included Apres in his 2020 Data and AI Landscape.

The round we eventually closed was for a different product than the one in the 2019 note. The note sold training data. The round was for explainable AI.

The round

In January 2021 we announced a $1.5 million seed round, co-led by Faber and Adara Ventures, with Techstars and local angels. The release described Apres as an explainable AI solution for heavily regulated industries like finance and insurance. My quote was the sentence I still believe: “Companies should not have to choose between AI performance and transparency.”

What I would tell myself in January 2019

Put your evidence in the units of the bar. If the bar is a $50,000 software pilot and two letters of intent, write down how many of those you have, even if the answer is zero. A list of good-sounding proof in other units feels like progress and tells an investor nothing they asked for.

Manual work is evidence. The line about a first paying customer before we started building was the strongest in that section. Doing the work by hand first is how you find out what to build, and a customer paying for the hand-made version is proof of demand.

The round may fund a different product. The product in the 2019 note and the product we raised on in 2021 shared a belief about trust. Almost everything else had changed. That is fine. Write down what you learned that made you change, so you can say it plainly when someone asks.

Check the numbers in your own announcement. The first draft of our seed release said 87% of models never reach production and successful ones carry “an 85% instance of error.” Neither number had a source in the draft. I have written separately, in “A number without its source,” about how numbers like that drift. They are easiest to catch before the press does.

In April 2021 I gave a talk for AWS Startups about when to join an accelerator, and when not to. Most of what I knew about raising, I learned in the two years this note didn’t predict.