User Intent Prediction #10: Universal Patterns.
I treated my users like unique snowflakes. They were all just melting ice.
I thought every diet was different.
- Keto users: Intense, science-focused.
- Fasting users: Patient, disciplined.
- PCOS users: Desperate for a medical solution.
So I built separate models for each cohort. “You can’t use a Keto model on a PCOS user! They’re totally different!”
I spent 2 weeks training 4 different models. Each one “specialized” for its niche.
I was wrong. Desperation looks the same in every language.
The Experiment
I took a model trained only on the PCOS cohort (12,000 users). I tested it on the Rice Diet cohort (8,000 users).
These are opposite demographics.
- PCOS: Women with hormonal issues, ages 25-40, medical focus.
- Rice: Budget-conscious users, ages 18-50, looking for a simple detox.
Different age groups. Different motivations. Different content.
The Result:
- PCOS Accuracy: 96.5%
- Rice Accuracy: 96.1%
The model lost less than 1% performance. It worked perfectly on users it had never seen, in a funnel it didn’t know.
The Universal Language of Hesitation
Why did it work? Because behavioral psychology is universal.
| Behavior | PCOS User | Rice User |
|---|---|---|
| Hesitation | Stares at “Hormone Test” for 15s | Stares at “Price” for 12s |
| Momentum | Clicks through 5 screens in 10s | Clicks through 7 screens in 12s |
| Rage-Clicking | Taps “Back” 4 times | Taps “Back” 3 times |
The content of the funnel changed. The human reaction to the funnel stayed exactly the same.
- Hesitation means the same thing for a Keto user as it does for a Rice user.
- Momentum looks the same.
- Rage-clicking looks the same.
A confused user in a PCOS funnel behaves exactly like a confused user in a Rice funnel. An excited user in a Keto funnel behaves exactly like an excited user in a Fasting funnel.
The Lesson
Humans are more alike than they are different.
I wasted weeks building “specialized” models for every niche. I treated my users like unique snowflakes. But when it comes to buying, we are all the same.
We hesitate when we’re unsure. We rush when we’re excited. We quit when we’re bored.
The signal is human, not demographic.
This concludes the “Systemic Failures” of October. Next month, we stop playing with models and start trying to deploy this thing. Spoiler: It was a nightmare.