Connect with the Product Hunt community

I spent three weeks building a heavy export feature because three outspoken users kept asking for it again and again. After we launched it the numbers told me that than two percent of our active users actually used the new export feature.

For the big idea I chose a cheaper path. I put a button for the requested feature in the interface. When a user clicked that button a pop up appeared that said the feature would arrive soon and invited the user to give feedback.

Than forty percent of users clicked the button proving that there was real product demand before I wrote one line of backend code. That proof saved us weeks of wasted engineering effort.

I always measure real user clicks on a placeholder interface before I commit development resources to a feature.

We ve all fallen into the trap of spending weeks building a highly requested feature, only to watch it sit there untouched after launch.

People love filling out feedback forms and asking for complex options but their actual daily habits rarely match what they type in a survey.

Now before writing any code, we try putting up a fake button or a simple mockup just to see if people actually click it. How do you usually validate an idea before committing your engineering time to it?

It's tempting to keep shipping new capabilities to look active to your paying base. But growth stalled out for us until we hit pause, looked at our data, and realized our tool required too many context switches from where our user already lived.

Fixing workflow fit meant stripping things out, not adding them in.

When your core loop doesn't match your users' daily reality, how do you decide what to cut versus what to streamline?

We tried fine-tuning a small open-source model on three years of raw support tickets to catch product sentiment trends early but the model learned to repeat typos, sarcasm and angry formatting instead of helpful patterns. Our initial evaluation runs scored terribly because half the training rows contained chatter or corrupted text exports.

We wasted two days trying learning rates before realizing the problem was garbage input data, not the model architecture.

Writing a Python script to filter out short messages strip HTML tags and normalize user names fixed the training output completely. Clean data matters more, than parameter count when you are adapting a model for a specific niche workflow.

What is your go-to preprocessing step when preparing raw user text for fine-tuning?

I work a lot on my own as a solopreneur. When nobody is expecting anything from me, I find it difficult to find the right direction and stay on track with what I have to do.

I first tried to fix that with a routine. Every Sunday evening I wrote down what I wanted to do that week, and I blocked time in my calendar. It worked for a few weeks, then I dropped it.

Dashboard funnels and drop-off charts will tell you where people left, but they never really tell you why.

We tried staring at drop off steps for weeks trying to guess the issue but nothing beat just emailing people who bounced on day one.

Out of fifty quick casual emails asking what went wrong, only three or four replied but those few replies completely exposed a broken signup step we were completely blind to.

For anyone doing user research, what is your trick to actually getting a response from someone who already walked away?

Has a press release ever actually brought you a customer?

For a recent launch I sent a free press release out through a handful of distribution sites. One of them got it onto AP News, a couple hundred local news sites and Google News within a few hours. No journalist wrote about it though, it was all automatic reprints.

What did come back was a bunch of sales pitches, an affiliate network asking to list the product, and a sign-up from a company that found us through ChatGPT. I'd done the same thing for my Mac apps before and got nothing like that, so my guess is people watch those feeds for B2B tools.

Have you tried one for a launch? Did anyone actually reach out because of it?

We agreed to build three bespoke integrations for a high-paying enterprise client to secure our first six-figure contract and extend our runway. At first, it felt like a massive win the cash flow was great, and having a big name on our pitch deck gave us instant credibility with investors.

But six months in, the reality set in. Our dev team spends almost 70% of their bandwidth maintaining custom legacy code, fixing edge-case bugs, and jumping on emergency support calls for this single client. Meanwhile, our core SaaS product has barely moved forward, public feature releases are constantly delayed, and our standard paying users are starting to complain about neglected roadmap promises.

It made me realize that chasing short-term enterprise revenue can quietly hijack your product vision, drain your engineering energy, and turn your scalable SaaS company into an expensive software agency.

Have you ever built custom one-off solutions to close a big deal, and was the immediate cash worth the technical debt and lost momentum?

At times, the metrics will not move no matter how many tricks I pull out of the bag. I get zero sign-ups, zero replies, a weird support message. Last month I spent three straight days on such metrics and discovered that during this time I was just doing all kinds of productive but futile things. I organised the folder structure, redid the footer, did pricing page comparison for the fourth time.

Gradually, I started maintaining a list of activities called "boring but real". It contains actions that will make some kind of movement even when I am totally uninspired: sending one message to a past customer asking what he does manually now, fixing one bug reported via support mail, deleting one rarely used feature. All these tasks take thirty minutes at most, and I can do them even when I am exhausted.

Working with AI tools speeds up writing code but auditing their test suites can honestly take longer than writing them from scratch sometimes.

I recently found a test where the assertion was literally comparing a variable to itself just to satisfy a green checkmark. It makes you realize how easy it is to accept code that looks good on the surface without checking its actual logic.

What's the funniest or most useless test an AI has ever generated for your project?

Our automated exit survey asked only one multiple choice question. Options included missing features or pricing. Eighty percent of churned users chose the pricing option. This made us think we needed to change our price tiers. That was a mistake.

When we sent emails to a few of those churned users we learned that churned users left because the setup process was too slow not because of the cost. Churned users used the pricing answer as a cover to avoid explaining further.

We swapped the generic multiple choice form for a text box. We also gave a reward for a real explanation. The written feedback told us where the friction lay and we fixed it.

Never rely on cancellation checkboxes alone. Use qualitative follow up to discover why churned users really leave.

Our early roadmap turned into a chaotic mess of custom requests from outspoken free users who wanted special integrations. We tried to build everything. That made us delay core stability updates for our paying customers.

We made a rule: no feature request goes onto the main product roadmap unless three different enterprise clients ask for it in one month. All other requests go into a backlog.

Our core application became faster and more stable. Our churn rate fell because our engineering focus matched our real revenue source.

Do not let loud community voices pull your product roadmap into directions that do not serve your paying customer base.

Every time I ask an agent to scaffold a utility or integrate a service, it loves to pull in packages that either don't exist, are heavily deprecated, or use completely outdated syntax. You spend more time debugging ghost imports than it would have taken to write the boilerplate yourself.

For developers working with fast moving tech stacks daily how do you feed context or set rules so the model sticks to your actual dependency tree?

How do you keep your AI agents from recommending or importing phantom packages?

Our free invoice generator takes a JSON post with no account behind it and hands back a PDF. So it has limits on everything.

At most a hundred line items. At most twenty taxes on any one item. A length cap on every text field, a byte cap on the logo, and checks that run before anything expensive is decoded. Each limit was chosen by asking how large one field can reasonably be, and each one was sensible on its own.

Then we measured a request that respected every single one of them: a hundred items, each carrying twenty taxes. It answered 200, and it held a worker for about 62 seconds. The same hundred items with one tax each took about 2.4.

Nothing was broken. The limits multiplied. A hundred items times twenty taxes is two thousand line and tax pairs, and each one is printed in its line and, when the names differ, again as its own row in the totals. The item limit counts items. The per item tax limit counts what sits inside one item. Nothing counted the document.

This week's Product Hunt highlights include something that has stayed with me. The creator of Orkut is now talking about making a new social network, and the idea behind it appears to be that current feeds suck compared to what we had.

It's a little mixed for me. Nostalgia is a good sales tool, but it won't make anyone leave the app where all their friends are. Orkut failed against Facebook not because of the features, but because all your cousins switched overnight.

We fell into this exact same trap last year. We ran an A/B test on our landing page and picked the headline that gave us the highest number of free signups.

Turns out, a flashy catchy headline brings in a ton of curious people who bounce after three days while a boring, clear statement attracts the actual buyers.

Ever since that mess we stopped looking at raw conversion rates as the main success metric.

How do you usually balance getting a high volume of signups versus attracting people who are ready to pay?

I have developed a few saas. Most of which were AI based saas and they needed AI to produce output. I wanted to run lifetime deals on those saas, but there was a problem that for how long can I sustain those users? I have been through a lot of lifetime deals platform to see if those were good fit for my saas. Then there came another problem for me. All of those platforms were taking a minimum of 40% sale commission while asking me to heavily discount the listing price.
For me, those platforms were charging way more than I expected initially. Then I realized that it would be a disaster for me if I ever launched my saas on these platform. I also came to know that there were people known as power users, and these people tend to exploit your application and its features.
I also was confused whethe providing lifetime deals directly to the users of your saas would be good? But, the question here still arises - For how long can you sustain these users as your api bill will be there for you every month for the rest of your life?

So, did you ever plan for lifetime deals? Or would you ever go with lifetime deals?

For months, our churn analytics pointed to pricing or missing integrations. It wasn't until we changed how we ran exit interviews focusing entirely on their day to day routine before finding us that we saw real disconnect.

They didn't churn because we lacked depth; they churned because using us required changing how their entire communicated.

For founders doing deep qualitative research on churn: what cues do you look for to spot workflow friction early?

We used to blame our UI or missing integrations every time a paid account downgraded. But after jumping on exit calls, a harsh truth came up: the product worked fine but it forced teams to completely upend their daily routines just to use it.

Throwing more features at a workflow mismatch is just building debt.

For those who've tackled this: how did you redesign onboarding to match your users' natural habits instead of fighting them?

Early on, I measured success by my daily signup count . But after monetizing watching paid users cancel hit totally differently.

The hardest part was realizing that adding quick features to our backlog won't fix most of these cancellations. People are leaving simply because the product isn't integrating smoothly into their existing habits.

For founders who have been managing paid tiers for a while: how do you adjust your roadmap to fix fundamental workflow friction?

In the early days, every new account felt like a major milestone. But once revenue kicked in, watching an active subscriber downgrade or cancel stings far worse than a visitor bouncing from the landing page.

Digging into exit surveys, churn rarely happens because of bugs or missing features. Users leave simply because the product hasn't become a daily habit.

Does the anxiety around churn ever actually go away, or does the feedback loop just become second nature?

I spent a year trying to get better at things I was bad at. Systems, processes, consistency.

I improved slightly. Not enough to matter. And in that year, the things I was naturally good at got a little rusty because I wasn't using them.

I discover a lot of tools through Product Hunt, and the upvote count is usually the first thing I look at. But it doesn t always match how useful a tool turns out to be for me.

Last month I signed up for a #1 product of the day because of its huge upvote count. I used it twice and forgot about it. A few weeks earlier I had tried a small launch with barely 50 upvotes, a simple tool for organizing browser tabs, and I still use it every single day.

It made me wonder if I give upvotes too much weight. Maybe the comments and the maker s replies say more about quality than the ranking does.

When you decide what to try, what do you look at first: upvotes, comments, the demo, or something else? Has a low-ranked launch ever surprised you?

I built my app in months. Marketing has been the hard part. There's no "vibe marketing" yet, but a few tools made specific jobs much easier. Here's what I used, task by task:

1. Launch video ElevenLabs

• What I used: its image, video, voice, and music tools. I made one reference image of my character and generated every scene from it, so his face stayed the same. The voices use Eleven v3 dialogue with tags like [tired] and [pause].

• Tip: avoid the AI agent editor. It used about 3x the credits I expected. Generating each piece yourself is cheaper.

2. Editing, captions and motion graphics Claude

Product Hunt newsletters have taken on two different pacing schedules. The Leaderboard is on weekdays, and The Roundup comes on Sundays. I subscribe to both of them, but my reading behavior ended up being something I had not anticipated.

The Leaderboard can be read alongside my cup of coffee on weekdays and I generally remember at least one of those products by the time lunch rolls around. The Roundup is longer and so I just skim through it.