At Inithouse, we run a lab that ships a growing portfolio of AI products in parallel. Not one product at a time. Not a pivot-heavy path from idea to idea. A deliberate strategy: build multiple MVPs, measure what sticks, double down on what works.
Here is how we actually do it, and what we learned shipping products like Magical Song (studio-quality custom songs from your story), Be Recommended (AI Visibility Reports for brands), and Ziva Fotka (AI photo-to-video tool, multi-domain across 5 languages).
1. Start with one product, then fan out
We did not start with a portfolio. We started with a single MVP, validated the build pipeline, then replicated it. The key lesson: don't scale the number of products until you can ship one in under 3 weeks. If your first product takes 3 months, your tenth will too.
2. Standardize the tech stack
Every product in our portfolio runs on the same foundation: React SPA, Supabase backend, shared analytics layer. When we build Pet Imagination (AI pet portrait generator with 9 styles), the deploy pipeline is identical to what we use for Verdict Buddy (AI conflict resolver using established psychology frameworks). Same CI, same monitoring, same hosting config. A new product spins up in hours, not days.
3. Share infrastructure across products
We built a shared analytics and reporting layer that covers all products at once. One dashboard shows signups, conversions, and engagement across the entire portfolio. We track Google Search Console, GA4, and Clarity for every product from a single config file. When something breaks on one product, we usually catch it because the same pattern broke elsewhere first.
4. Automate the repetitive work
Our reporting, SEO audits, and content publishing run on automated schedules. We measured the time we spent on manual weekly reports and replaced most of it with scheduled jobs that pull data, flag anomalies, and post summaries. The team reads a digest, not a spreadsheet.
5. Measure early, redirect fast
We observed that most products show clear signals within the first 30 days: either users come back, or they don't. We track returning user rates, funnel drop-offs, and activation events from day one. Products that show early retention get more attention. Products that stay flat get a narrower focus while we learn more.
6. Document everything
With a growing portfolio, context switching is the real enemy. We keep a single config file for every product (domain, analytics IDs, database refs, notes) and update it after every change. When someone on the team picks up a product they have not touched in two weeks, they read the config and the last report, not a Slack thread from three days ago.
Common pitfalls we hit
Too many products too early. We tried fanning out before the pipeline was reliable. Result: half-shipped MVPs with broken analytics. The fix was simple: finish the pipeline first, then replicate.
Ignoring maintenance. A growing portfolio means a growing number of small fires. We now batch maintenance days where we run through every product and fix whatever surfaced that week.
Not tracking the portfolio as a whole. Individual product metrics are useful, but portfolio-level patterns are where the insights hide. We noticed that products with shared referral traffic outperformed isolated ones. That changed how we think about cross-linking and content.
The stack
For anyone curious about the specifics:
- Frontend: React (Lovable for rapid builds)
- Backend: Supabase (auth, database, edge functions, storage)
- Analytics: GA4 + Google Search Console + Microsoft Clarity per product
- Automation: Scheduled tasks for reporting, audits, content
- Config: Single YAML file per product with all IDs, domains, and notes
What we would do differently
We would invest in shared component libraries earlier. We would set up cross-product A/B testing from day one. And we would resist the temptation to ship "just one more product" before the existing ones have solid funnels.
At Inithouse, we keep shipping a growing portfolio of AI products because the data tells us which bets to keep making. If you are building multiple products, standardize early, automate everything you repeat more than twice, and measure before you decide.
Top comments (0)