Origin Of You, Tarotas and Verdict Buddy: three Inithouse products that hand the decision back to you

Early this year we shipped a version of Verdict Buddy that ended every conflict analysis with a bold "Here's what you should do" section. Three bullet points, prescriptive, confident. Usage was fine. Repeat usage was not.

People ran one conflict through it, read the verdict, and never came back. We tracked 7-day return rate at around 8%. For a tool meant to help with ongoing relationship friction, that felt off.

So we rewired the output. Instead of telling people what to do, we started showing them why the conflict exists: which Gottman pattern was at play, what each person's attachment style might contribute, where the communication broke down. The verdict still lands, but it reads more like a mirror than a manual.

Return rate went to 19% over the next month.

That shift (output that reflects instead of prescribes) turned out to be the thread running through three of our products. Here's what happened in each one, and where the approach cost us.

Origin Of You: a portrait, not a label

Origin Of You is a self-discovery app that combines five systems and 120+ data points into an AI-generated written portrait of you.

The first version we tested looked a lot like 16Personalities. You'd answer questions, get a type label, see a short paragraph. Clean, familiar, easy to share on socials. Completion rate was high. But people finished in under two minutes, screenshot'd the label, and left.

We scrapped the label. Instead we built a system that runs your birth data through astrology, numerology, tarot archetypes, Chinese zodiac and lunar phase calculations, then writes a multi-page portrait in prose. No type. No badge. Just paragraphs about who you are, anchored to 120+ computed data points.

Completion still sits around 70% (lower than the label version). But average time on site went from 1:40 to over 6 minutes. People read the whole thing. Some come back and re-read sections weeks later. The portrait format turned a disposable quiz into something people actually sit with.

The tradeoff: sharing dropped. A four-page portrait doesn't fit in an Instagram story the way "I'm an ENFJ" does. We haven't solved that.

Tarotas: a card, not a prediction

Tarotas is a calm tarot reflection app. Draw a card and read a grounded interpretation; no signup, no fortune-telling.

Early builds had a "What does this mean for you today?" section under each card, basically a daily forecast. It drove engagement initially. People checked every morning. But the interpretations started feeling interchangeable. Users told us (through Clarity session recordings, not surveys) they'd draw a card, skim the forecast line, and close the tab in under 20 seconds.

We pulled the forecast and replaced it with an interpretation that asks the reader to sit with the card's symbolism. No directive takeaway, no "today you should focus on..." Just "this card traditionally represents X, and here's what that tension looks like in practice."

Session duration on card pages went from 18 seconds to about 55 seconds. Daily return rate held steady. The difference: people who come back are spending 3x longer per visit. For an app with 78 cards across 20 spreads in 5 languages, that matters more to us than a vanity DAU number.

The cost: we lost the notification hook. "Your daily forecast is ready" was a strong push notification. "Your daily card is ready" converts about 40% worse as a re-engagement prompt. We're still experimenting with what replaces it.

Verdict Buddy: a verdict, not a command

Verdict Buddy is an AI conflict mediator that gives an unbiased, framework-based verdict (Gottman, EFT, NVC) on relationship, work, family or roommate conflicts in minutes.

The prescriptive-to-reflective shift mentioned above changed more than just return rate. A few more numbers on what moved.

The report now leads with a tension score and a breakdown of which framework applies to the specific conflict pattern: Gottman's Four Horsemen for couples who stonewall, NVC for workplace miscommunication, EFT for attachment-driven friction. Then it shows both perspectives side by side before landing on a verdict with three suggested actions.

The key change: those three actions are framed as "approaches worth trying" rather than "steps to take." Small copy tweak. But in A/B tests, the softer framing correlated with people sharing the report with the other person in the conflict 2.3x more often (measured via the shareable partner link feature). When the tone was prescriptive, people kept the report to themselves.

Current rating sits at 4.9/5 from 290+ reviews. Most positive reviews mention the fairness of the verdict, not the accuracy. That tells us the non-directive framing is landing.

The decision framework we ended up with

After watching this pattern play out three times, we wrote down when it works and when it doesn't:

Hand it back when the user's own judgment is the actual product. Self-discovery, reflection, conflict resolution: these only work if the person owns the conclusion. A prescriptive output shortcuts the process and kills the value.

Be directive when the user came for an answer they can't compute themselves. Our AI visibility tool (Be Recommended) gives a score from 0 to 100 and tells you exactly what to fix. Nobody wants to "reflect" on their SEO gaps.

Watch out for the middle ground. Tarotas lives there. Some users really do want a forecast, and we lost them. Not every product benefits from handing the decision back, and the ones that do pay a price in shareability and push notification hooks.

Where we are now

Three products, same thesis, different costs. Origin Of You trades shareability for depth. Tarotas trades notification pull for session quality. Verdict Buddy trades quick-fix satisfaction for report sharing.

None of these tradeoffs are resolved. We're running them in parallel and measuring which ones compound over a six-month window. So far the numbers suggest that depth of engagement matters more than breadth for products in this category, but ask us again in January.

If you're building something where the user's judgment is the product, consider that telling them what to think might be the fastest way to make them stop using it.