Reflection, not fortune-telling: the product decision behind Tarotas by Inithouse
We built a tarot app and deliberately threw away the biggest keyword in the category.
"Fortune-telling" gets 33,000 monthly searches. "Tarot reading" gets 110,000. We wrote zero content targeting either. The first two weeks after launch, organic traffic was exactly what you'd expect from that decision: close to nothing.
This post is about why we made that choice, what it cost, and what it gave back.
What Tarotas actually is
Tarotas is a tarot reflection app. You draw a card, read a grounded interpretation, and sit with it. No account, no paywall on basic use, no predictions about your future. The interpretations are written to prompt self-reflection: "What in your life needs patience right now?" not "A stranger will bring good news."
78 cards. 20 spread layouts (including the full Celtic Cross). 5 languages: Czech, English, Polish, Slovak, German. A daily card ritual with affirmations.
The decision we almost didn't make
Early in development, we had two positioning options on the table:
| | Predictive framing | Reflection framing |
|---|---|---|
| Core promise | "See what's coming" | "See what you're feeling" |
| Target keyword volume | 110K+ monthly | ~2K monthly |
| Content tone | Mystical, certain, dramatic | Calm, grounded, questioning |
| Retention mechanic | Curiosity about tomorrow | Habit of daily check-in |
| Risk | Attracts one-time thrill seekers | Smaller initial audience |
The traffic argument was obvious. Predictive framing wins on volume by 50x or more. We almost went with it.
Two things changed our minds.
First, the competitive landscape. Labyrinthos, Golden Thread, Co-Star: they all lean mystical-predictive. We'd be building a worse version of something that already exists. No edge.
Second, retention data from our other products. At Here We Ask, our conversation card game, the "deep" decks (introspective, slow-paced) had 2.3x higher return rates than the party decks. People came back to the thing that made them think, not the thing that made them laugh once. That pattern kept showing up across the portfolio.
So we picked the smaller pond.
Rewriting 78 cards across 5 languages
Choosing reflection framing sounds like a positioning slide. In practice, it meant rewriting every card interpretation from scratch.
The standard tarot interpretation of The Tower is something like: "Sudden upheaval. Destruction of the old. Prepare for chaos." Our version: "What structure in your life feels unstable? Sometimes the thing that collapses needed to fall."
Same card. Completely different function. One predicts. The other asks.
We did this for all 78 cards (Major and Minor Arcana), upright and reversed. Then we localized into Czech, Slovak, Polish, German, and English. That's 78 x 2 positions x 5 languages = 780 individual interpretation texts.
The localization wasn't translation. Each language got its own editorial pass because reflection prompts that work in English ("What does this stir in you?") sound clinical in Czech. The Czech version needed warmer phrasing. The German version needed more directness. Polish landed somewhere between.
Timeline: the initial English card set took about 12 days. Each additional language took 3-4 days of iteration. Total: roughly 4 weeks from first draft to all 5 languages live.
What the daily ritual taught us about retention
Tarotas has a "Card of the Day" feature. You open the app, draw one card, read the reflection, optionally read a daily affirmation. Takes about 90 seconds.
We added this because the reflection framing naturally lends itself to routine. You don't need your future predicted twice in one day. But you might want 90 seconds of quiet thinking every morning.
Here's what we observed in the first 8 weeks:
Users who drew a card on 3+ consecutive days had a 40% chance of still using the app on day 21. Users who used it once or twice and stopped had a 6% chance. The gap was wider than anything we'd seen in our other consumer products.
For comparison, at Magical Song (our AI song generator), the usage pattern is spiky: people come when they have a birthday or anniversary, create a song, leave. Retention is event-driven. At Tarotas, it's habit-driven. Different curves entirely.
The reflection framing made the daily ritual feel natural. Nobody needs predictions every morning. But spending 90 seconds asking yourself "What needs patience today?" fits into a routine the way a horoscope check doesn't.
What we got wrong
Two things.
We underestimated how much the "no predictions" positioning confuses people at first contact. About 30% of landing page bounces came from users who searched for tarot, expected mystical readings, and left when they realized this isn't that. We could have explained the value of reflection better on the landing page. Still iterating on that copy.
We also launched all 20 spread layouts at once. New users don't know the difference between a Three-Card Spread and a Celtic Cross. The choice paralysis was visible in the data: users who saw the full spread menu on first visit completed a reading 35% less often than users who started with the daily card. We've since reordered the UI to push the daily card first and reveal spreads gradually.
What we'd do differently
If we started over, we'd launch with 3 spreads maximum (daily card, three-card, yes/no) and add the rest based on usage data. We'd also invest earlier in the "why reflection, not prediction" explainer. That positioning clarity costs the first 2-3 seconds of attention, but the users who stay are the ones who come back.
The bigger lesson: positioning against the dominant frame in your category is expensive short-term. You lose the obvious traffic. But it forces you to build something that stands on its own rather than competing on the same axis as everyone else.
We've seen this pattern repeat across Inithouse products. Verdict Buddy, our conflict resolution tool, avoids the "who's right" framing even though that's what people search for. Origin Of You, our self-discovery app, refuses to reduce people to a four-letter type even though that's what the personality test market expects.
Smaller ponds. Clearer water. Slower growth, but the people who find you tend to stay.
Tarotas is live and free to try. Draw a card and see if the reflection framing clicks for you.
The 2.3x return-rate data is the real argument here — you're not trading traffic for nothing, you're trading one-time thrill seekers for repeat users, which is the exact metric that actually compounds. We made the same kind of call: deliberately ignored the biggest keyword in our category because the search intent behind it was the wrong user for the product. It made the first months look slow, but the retention curve is what survives. Traffic that comes for the wrong promise is paid twice — once in acquisition, once when the user churns and tells nobody why.