C'est la vie. Learn from it. Get to work. Be creative.
One of the things I noted for the next product is this one:
>First thing to do before entering a market is to create a community around the problem space. This will serve market research, product development, and customer development. Channel and distribution. It also is useful to start doing lay of the land, ecosystem charts, and a repository like awesome-x, or tools-of-the-trade for that domain
One of the things that works for us is that we built our product for ourselves after spending years in that space, delivering ML products for large enterprise clients and carrying out actual projects and getting paid for it. It's not a startup idea and we were not looking to make things pretty with CSS or looking for a product idea, we built it for ourselves not to die, or go insane, so our roadmap is clear and it itroduced massive savings for our consulting operation. We've also been avoiding investors because we have been profitable since day one. That is a position of privilege, and depending on your situation, your experience may dramatically vary with regards to product launch/audience/ and whether the product you're building is a startup idea and you'll go work at a FAANG if it fails, or something you very desperately need.
To your point, and in any case, talk with prospects and potential customers as soon as possible. One of the good things we did when we started our internal platform, was to develop it as if it were a public facing product, which means we put up a website, talked with external people, onboarded external users on a case-by-case basis, had a landing page and request access field, a Slack workspace, even posting it on HN and Reddit from time to time.
I also use email. Hundreds of emails, and hundreds of conversations to a list I manually gathered and filtered and curated through several websites. And follow-ups, for people who simply forgot to reply.
This had allowed us not only to solve problems our internal team had while carrying out machine learning projects, but also to solve a more generalized form of these problems and implement things that made sense, or validate issues in our backlog.
We have issue templates for features with a section called "Instances". We need a list of several instances of something that sucks in order to implement the feature or the improvement. Even our internal users must list instances of something preventing them from getting work done before we go "OK, this thing here sucks. Let's fix it.". In other words, you might benefit from developing your product that way as well, getting users early on.
I have learned things from people who did not want to use our platform, and that's where the "consulting skillset" comes into play: digging at the underlying problem, not the symptom of the problem or the client's idea of a solution. That, and ensuring the comment came from someone who's worked in the field and belonging to my target audience, not a hobbyist.
Take a look at this objection: https://old.reddit.com/r/MachineLearning/comments/kolobf/p_c...
>Sounds like a great tool. But for me to seriously consider it would have to be OSS.
If you take it at face value, you'll likely think "OK, we seriously need to consider open sourcing". That's why it's dangerous to listen to solutions instead of discovering problems. It turns out from that conversation that the "open source" was a solution to an underlying problem.
Digging deeper, this is the Redditor's response when I pointed out a counter-argument that they do use closed-source solutions:
>A major distinction between why we are ok with using something such as GKE versus a non-OSS version of your tool is that the ML space is quite immature and crowded these days, thus there is no guarantee that companies providing services/tools will be around in the long run.
They were afraid we might die, and their solution was to rely on "OSS" to manage that risk, but it's not the only solution.
I wrote about that dynamic here: https://news.ycombinator.com/item?id=28378650
In any case, what are you up to? What are you working on?
The above is full of assumptions (B2B, revenue from the sale/subscription as opposed to ads or something, viable from the get go without venture funding, etc). Not sure if my assumptions are correct.