AMA [W23] - My journey from $0 to $5k MRR in 6 months, and why I decided to pivot

[frishy]

@veyhenlo “Dont listen to your users”

Thats a wrong learning, but I understand why you look it that way.

Rather your learning should be:

Fix yourself on a user “segment” and keep pivoting on the tech/solution.

What users will tell you will not be their problem, it will be manifestation of some deeper problem, and with iterations you go deeper into finding that problem.

But, you started with a solution, and then tried to find a big enough problem where it will fit - this is classic trap of Tech solutions seeking a fitting user segment.

However, if you do start out with a Tech solution, then you “fix the tech” and go deeper in the tech, and you keep pivoting on user segment to find a segment big enough for the solution - this is not the best approach but it does work too.

So, the takeaway is:

Start with a user segment, identify their problem, and then build tech for it. Any other way will just increase your probability of failure manifolds.

BUT

If you do start with a Tech solution in need of a user segment, then atleast fix on the Tech, and keep pivoting on user segment - don’t change both the user segment and the solution at the same time - and that was the real problem.

 

[veyhenlo]

@frishy Thank you for this! You make a great point.

I want to say we didn't fall for the solution in search of a problem cliche, but perhaps I oversimplified. Remember, we originally started the spreadsheet because of our own inventory management needs.

Later on as we were looking to niche down and start monetizing, we were talking with users who were already using our spreadsheet (arguably to solve a problem they have) + doing discovery around asking people about the tasks they carry out in spreadsheets.

Does that make sense?

 

[frishy]

@veyhenlo Yeah, makes sense, I understand that. I didn't mean to downplay the grill you been though, and I don't think there is any shame in failing at what is consider cliche either - its just that we are human and fallible, emotions carry us as our natural instincts, and it takes time to learn to keep emotions aside in business, and that process makes us fail at even most cliched things.

I have done same mistake multiple times, failed miserably, even at most cliched and stupidest things multiple times, only after that my intuition stated to become brutally rational, and now if I have to make a solution for inventory management, I will focus on that only, weather it requires spreadsheet or an app or manual labor, ML or no ML, or whatever, none of that matters in my perspective, what matters is the main metric that matters to my target segment when they manage inventory, if that is being moved or not, everything else is secondary detail, I just stay ready to completely pivot on the details in a heartbeat.

I just wanted to share that perspective, it helps being brutally rational, that was one change in myself that changed the entire trajectory of my life from a slog to a cruise.

 

[veyhenlo]

@frishy You are right that I said we set off to fix our own inventory management problem and then I didn't mention inventory management ever again

At the time we looked elsewhere because we were a bit jaded with the grocery delivery business (slim margins + small number of players = small market). Maybe I should have included that part!

I was also excited to see YC companies in the two most recent batches go after our original problem, it's a very real problem.

 

[jonib]

@veyhenlo This is the best post I've read for a long time, thanks so much for sharing!

Similar to your story, I've been building ML algorithms since pre-gpt days to automate data collection and analysis, but for environmentalists. It's a very niche but growing market and we have a lot of domain knowledge. Our tech shares many similar roadmap to yours. I wonder if you are open to have a chat and if there's room for collaboration in some way?

 

[veyhenlo]

@jonib Wow, thank you! Will make sure to share more often

DM me or email me at gonzalo[at]double.bot, happy to chat

 

[oyinkan]

@veyhenlo such a good post.

One note - I come from sales and the idea that “personalisation doesn’t change the outcome” is very controversial. It’s the basis on which companies like Clay have built a client base. Additionally so much outbound agencies etc.

What industry/type of product were you doing a campaign for?

 

[veyhenlo]

@oyinkan Appreciate the kind words! I think we'll share more often

Our campaigns were for our enrichment product itself, a tool to build in-app surveys for mobile apps and SaaS tools, and a credit cards & expense management platform.

It's interesting because the latter two had issues with targeting using traditional methods (i.e using Apollo and SalesNav), that alone probably warrants a post of its own in a Sales related subreddit.

Also regarding personalization, I'm a bit out of the loop now but a lot of the companies that were offering similar solutions to ours have also trimmed down their promises. I think Clay is great because it allows you to pull data from various sources in a single platform (something that was very hard before), and use that to choose the best template to send out.

What's your experience with personalization? Are you doing anything beyond a couple of custom fields on a template? I'd love to be proven wrong!

 

[oyinkan]

@veyhenlo Very interesting. In general, both linkedin and email are overloaded in SaaS.

I have not seen personalisation working at scale either. The kind of things that you want to personalise on in general are the things that you don’t want to be automated. If its easy to automate, everyone will do it and it will be less successful.

Some things that I currently do for most important prospects are transcribe podcasts/videos and get an insight. Or looking at an annual statement and creating a POV related to a problem they might have and we can solve.

There should still be an opportunity to build something there btw, just don’t know how defensible it would be. Unless you can capture a data source that noone else can, everything can be copied.

 

[veyhenlo]

@oyinkan

transcribe podcasts/videos and get an insight.

This is genius! The other day I also saw someone use their prospect's LinkedIn profile picture to generate an AI picture of them in a way that catches attention (bit more cliche and corny than your example).

Honestly that's what I found so exciting about working in sales tech. Non-stop competition to find an edge.

 

[marvin54]

@veyhenlo Great to see the progress. We spoke way back when you were trying to make a sales tool and your own reflection is spot on. That's a space you have to know very well to succeed in. If you already know software development, that's where you should be building and it is paying off. Good luck!

 

[veyhenlo]

@marvin54 That's really cool, small world! What are you working on?

 

[marvin54]

@veyhenlo higherlevels.com

 

[calpastor]

@veyhenlo Great post and insights - your focus on hypothesis based testing is impressive. Obviously copilots for software engineering is not a novel idea in 2024 with many incumbents - my question is:

With the deep improvements in LLMs and the biggest companies also releasing software engineering copilots that are improving every day, how do you plan to face the threat of larger competition over the long term? Will you ultimately pivot again?

 

[veyhenlo]

@calpastor Thank you for reading, and appreciate the kind words

That's a great question. We're wrapping up fund raising right now and will have more to share soon. The way I see it, this copilot is only the first step in our master plan to create a model to automate SWE from ticket to PR. In order to create this model, data is the ultimate bottleneck and having a copilot is the best way to get high quality data.

Also, nothing that we've seen from the big co's copilots tell me that this is a serious priority for them (or maybe they are that slow?). And I foresee the other big LLM labs forging partnerships with copilot companies the way OpenAI has already done.

 

[martinsg]

@veyhenlo This is awesome! Nice read and good story

 

[veyhenlo]

@martinsg Thank you for reading

 

[thelonemusician28]

@veyhenlo Thanks for the post!

If someone wanted to look into creating an AI copilot for design (I’m a web and UX designer), what first steps would you recommend? Is it identifying the areas for myself where a copilot would be helpful? Is it researching the tech that would go into it?

That’s my specific question, but wouldn’t mind zooming out and learning how you broach learning about a new concept or subject.

 

[veyhenlo]

@thelonemusician28 Thank you for taking the time to read!

That's an interesting idea. The simplest approach would be, start by looking at what's out there given that it's not entirely a 100% new idea. For example within YC CodeParrot and Magic Patterns might match the description of a Design Copilot. There is a Figma plugin with 70k+ users that also matches that description.

Get a sense of what people like and dislike about the tools in the market (read reviews and forums). Try them out yourself and add your professional opinion to the mix, from here formulate a hypothesis for how these can be improved (if you genuinely see a way), and build an MVP of it which you then should share with the world.

It should be simple, yet demoable. The goal should be to craft a tweet that can go viral + waitlist.

If you get a good number of people in your waitlist, you can start to build it and engage them + convert them to paying users (speed here matters a lot)

 

[thelonemusician28]

@veyhenlo Thanks a lot for taking the time to respond! I appreciate the insight into how to start to research the competitive landscape. I’m trying to improve my personal “heuristic” or process when it comes to vetting ideas I might get and this is a great help.

Good luck with your company!