TierWise

Simulate how 500 psychologically-modeled users decide what to pay — then get an AI-generated guide to rewrite your pricing for the minds you're losing.

Comment

Inspiration

We've all been there. You find a product you like, you see the pricing page, and you click free. Not because you can't afford it. Not because you don't want it. Something about the way it was presented just didn't convince you.

Now multiply that moment by millions of users every single day.

94.5% of ChatGPT's 900 million users pay nothing. The average SaaS product converts only 2-5% of free users to paid. Companies like Cursor — one of the fastest growing AI startups in history — had to publicly apologize after a pricing change went catastrophically wrong. Replit watched their profit margins collapse in months from a single pricing mistake.

We kept asking the same question: why are billion-dollar companies still guessing at this? And the answer was always the same. There is no tool that simulates how real humans — with different incomes, different personalities, different psychological relationships with money — actually make the decision to pay. Until now.

What TierWise Does

TierWise is the first behavioral pricing simulator.

You upload your pricing page. TierWise generates 500 simulated people — each with a unique income level, personality type, and psychological profile grounded in behavioral economics research. Each person independently decides which pricing tier to choose based on what they can afford and how your pricing page speaks to their specific decision-making style.

You watch it happen in real time. 500 colored dots on a canvas, each one a different mind, each one making a different decision. Then TierWise tells you exactly what happened — which types of people converted, which didn't, and precisely what to change about your pricing page to reach the ones you are losing.

The output is not a chart or a data table. It is a plain English guide that tells a founder or marketing team exactly how to rewrite each pricing tier — the specific words, the framing type, the information to show and hide — matched to the psychological profile of each customer segment.

How We Built It

The Core Idea: People Are Predictably Different

Behavioral economics — the field that studies how humans actually make decisions rather than how we assume they do — has established that people respond to the same information in fundamentally different ways depending on their psychological makeup.

Some people respond to loss. Tell them what they will lose by not upgrading and they convert immediately. Others follow the crowd — show them that 23,000 people already chose Pro and they follow. Others need an authority figure to recommend a plan before they trust it. Others are overwhelmed by complexity and leave the moment they see a 15-feature comparison table.

These are not personality quirks. They are documented, measurable, consistent patterns backed by decades of peer-reviewed research — including work that has won the Nobel Prize in Economics.

TierWise takes these patterns and turns them into a simulation engine.

The Five Types of Customers

Every agent in TierWise is classified into one of five archetypes — each representing a distinct psychological profile that exists in any real customer population:

Archetype Share How They Decide
Anxious Planner 22% Responds to urgency and loss framing
Social Follower 28% Goes where the crowd goes
Spontaneous Mover 20% Needs simplicity, bounces off complexity
Authority Truster 18% Follows the recommended option
Indifferent Drifter 12% Hard to move, stays free by default

The Decision Process

Each agent goes through four steps when evaluating a pricing page:

Step 1 — Can they afford it? Every agent has a financial profile based on their income level and existing subscription behavior. If a tier is genuinely out of reach, no framing change fixes that. TierWise respects this constraint.

Step 2 — Does the page speak to their psychology? The agent's behavioral profile is scored against the framing signals present in the pricing page. A loss-framed message activates loss-averse agents. A social proof badge activates social followers. An authority recommendation activates authority trusters. The match between page framing and agent psychology determines receptivity.

Step 3 — Can they handle the complexity? Agents with low cognitive bandwidth incur a dropout penalty when the pricing page is complex. This is why adding more features to a pricing page does not increase conversion — for a significant portion of any audience, it actively reduces it.

Step 4 — The decision The three factors combine into a conversion probability. The agent makes a decision. The result is recorded. The dot pulses on the canvas.

This process runs 500 times simultaneously — each with a different financial and psychological profile — producing a realistic population response to your specific pricing structure.

The Technology

The backend is built in Python using FastAPI. The simulation engine runs entirely client-side in JavaScript for speed — 500 agents generate and decide in under two seconds. The behavioral insight report and cognitive framing guide are generated by Anthropic's Claude AI, which reads the full simulation output and produces plain English recommendations a non-technical founder can act on immediately.

Connecting to the Hackathon Prompt

The prompt asked us to create a system that reshapes information to match a user's unique cognitive style and sensory preferences.

Most approaches to this prompt reshape information for one person at a time. TierWise takes a more powerful position: before you can reshape information for your customers, you need to understand the full distribution of cognitive styles in your audience. You need to know that 28% of your users are Social Followers who need crowd signals, that 20% are Spontaneous Movers who will leave if you overwhelm them with features, that 22% are Anxious Planners who need to feel the cost of not upgrading.

TierWise simulates that entire distribution simultaneously and produces a guide that reshapes how each pricing tier is presented — matched to the cognitive profile of each segment. Information reshaping at population scale, grounded in behavioral science.

The Validation Study

We recognized early that the most important question about TierWise is not "does it look impressive" but "does it actually work."

To answer this we designed the Digital Twin Study. We survey real people using an 11-question behavioral instrument that captures their income level and psychological profile. Each participant views a fictional pricing page and makes a real subscription decision. We then build a digital twin for each participant — an AI agent constructed directly from their survey responses — and run all twins through the same simulation. We compare twin decisions against real human decisions and publish the accuracy rate.

Stanford researchers running a similar study found AI agents replicated real human decisions at approximately 85% accuracy — matching the rate at which humans replicate their own answers across a two week gap. We are targeting that benchmark and publishing the methodology publicly so the results can be independently verified.

Challenges We Faced

Scope under pressure. TierWise was built in under 24 hours. The hardest design decision was not what to build but what to cut. We had to keep the simulation engine rigorous while keeping the output legible to someone who has never heard of behavioral economics.

Translation problem. The academic frameworks are sophisticated. The business owner using TierWise needs to act on the output in minutes. Every insight had to be translated from behavioral science language into plain revenue language without losing the rigor underneath. "Loss framing receptivity score" becomes "this person responds to urgency." That translation is harder than it sounds.

Making the simulation trustworthy. A simulation that produces impressive-looking numbers but doesn't reflect real human behavior is worse than no simulation at all. Grounding every agent attribute in published research — and designing the validation study to verify accuracy empirically — was essential to building something we could stand behind.

What We Learned

The gap between academic rigor and practical usability is enormous. You can build something grounded in Nobel Prize-winning research and still produce an output that a real founder finds useless because it speaks the wrong language. The interface between behavioral science and business decision-making is where TierWise had to earn its value.

We also learned that the pricing problem is far larger than we initially understood. Every number we found in the research was worse than we expected. The scale of revenue being lost to free tiers across the AI industry is measured in billions. The tools to address it systematically simply do not exist yet.

That gap is what TierWise is built to close.

What's Next

  • Complete the Digital Twin Study with 25+ participants and publish the accuracy rate
  • Expand the validation corpus with real customer post-launch data
  • Build multi-structure comparison — test two pricing approaches side by side against the same population
  • Submit the behavioral agent methodology as a research paper

Built for the prompt: "Create a system that reshapes information to match a user's unique cognitive style and sensory preferences." We simulated 500 cognitively diverse minds processing the same pricing page — and built a guide to reach all of them. The era of blind pricing is over.

Built With

Share this project:

Updates