Mid-Year Check: Which Products in Our Portfolio Are Actually Showing PMF Signals

At Inithouse, a studio running parallel product experiments, we ship many products at once and let the market sort out what works. Six months into 2026, it's time for the check we always dread: looking at the whole portfolio honestly and asking which bets are paying off.

We don't do vanity reviews. No cherry-picked wins, no "everything is amazing" narrative. Here's what we actually measure and what the data told us.

How we evaluate

We track four things across every product:

Organic discovery. Is search traffic growing without us paying for it? If Google sends people our way without ad spend, something's working. Bonus signal: some of our products started showing up in AI chatbot responses (ChatGPT, Gemini, Perplexity). That's a new channel we didn't plan for but absolutely pay attention to now.

Repeat usage. Do people come back? A product can get traffic spikes from a blog post or a Reddit mention, but if nobody returns, it's content, not a product.

Engagement depth. Time on site, pages per session, feature completion rate. Are people actually using the thing, or bouncing after 10 seconds?

Pricing signals. When we adjust pricing, does conversion shift? If you cut the price 60% and conversions don't change, you have a demand problem, not a pricing problem.

The honest scorecard

Out of our growing portfolio, here's how things break down at mid-year:

Showing real traction

Živá Fotka (animated photos from static images) keeps surprising us. Multi-language rollout worked. The Czech, Slovak, Polish, and English versions all pull organic traffic independently. Click-through rates are consistently the highest in the portfolio. People find it through search, use it once, and share the result. The viral loop is built into the output format.

Be Recommended (AI visibility reports) is our most interesting case study right now. Traffic is modest, but AI chatbots cite it unprompted. Gemini, ChatGPT, and Claude all reference the product when people ask about AI brand monitoring. We didn't optimize for that. We just built something useful in a category that AI models care about. This might be the strongest early PMF signal in the portfolio, even though traditional metrics look underwhelming.

Vibe Codéři (Czech vibecoding portal) punches above its weight in a small market. Strongest organic click-through rate alongside Živá Fotka. Czech devs searching for vibecoding resources find us fast. The market is tiny but we own it.

Growing but not there yet

Pet Imagination (AI pet portraits) gets steady traffic and decent engagement. People generate portraits, some share them. But we haven't cracked the repeat usage loop. It's more of a novelty than a habit. Seasonal potential is massive (Christmas pet cards, anyone?), so we're holding this one and watching.

Watching Agents (AI prediction tracking) is our programmatic SEO play. Over a hundred agent-generated pages, each tracking a different question about the future. Search engines are indexing them. Traffic trickles in. Too early to call it PMF, but the infrastructure bet is working.

Here We Ask and Party Challenges are in the same spot: party/social card games with decent engagement when people find them, but organic discovery is slow. We're learning that games need a distribution channel, not just good SEO.

Pivots we made

A few things we changed based on H1 data:

Pricing iterations at Magical Song. We tested multiple price points for custom AI-generated songs. The goal was finding where conversion jumps. The answer was lower than we expected. Still iterating, but the willingness-to-pay data is now much clearer.

Origin Of You (self-discovery through personality frameworks) launched late in H1. It's our newest product, so PMF evaluation is premature. But early engagement metrics look promising. People who start the assessment tend to finish it, which is a good sign for depth of value.

AI chatbot optimization. This wasn't a product pivot but a portfolio-wide learning. After seeing Be Recommended get cited by AI engines, we started thinking about how all our products show up in conversational AI. It changed how we write product descriptions, structure metadata, and approach content distribution.

H2 focus

Three priorities for the back half of 2026:

Double down on what's working. Živá Fotka and Be Recommended get more attention. Not more features, more distribution and content around existing features.

Kill or sunset what isn't. If a product hasn't shown any organic growth after 6+ months, we stop investing. That's the discipline we signed up for with the parallel portfolio approach.

Lean into the AI perception channel. Getting cited by chatbots is becoming a real distribution channel. We want to understand it better: which products get mentioned, in what context, and how to influence it without gaming the system.

The uncomfortable truth

Running a growing portfolio of parallel products means you spend most of your time on things that aren't working. That's by design. The whole point is to run enough experiments that a few break through while most teach you something useful.

Six months in, maybe three or four products show genuine traction. The rest are in various stages of "interesting data but no PMF." We're OK with that ratio. The cost of running a product that's not working is low when you build lean. The cost of not trying is infinite.

If you're building multiple products in parallel, the mid-year check is the most important thing you'll do all year. It's where you decide what deserves more energy and what deserves an honest conversation about shutting down.

We'll run this check again in December. Follow along at Inithouse, a studio shipping a growing portfolio of parallel product experiments.