Decision Community

See what was tested, how far to trust it, and what happened in the real world.

START HERE

Understand the product—and its limits—in three minutes

Seven directions recovered from public experiments, three recorded failure boundaries, and two MCP recipes you can reproduce. The misses are shown beside the wins.

Verified exampleCopy

Social proof beat appeals to duty

A neighbor-visibility message moved turnout from 29.7% to 37.8%.

Gerber, Green & Larimer randomized trial, about 344k voters

This recovers a public experiment direction, not your campaign lift.

See evidence and method
Verified examplePricing

Free mattered more than a one-cent cut

Moving to free flipped the majority from premium to the free option.

Shampanier, Mazar & Ariely real-transaction experiment

It shows the psychology of free; unit economics still need testing.

See evidence and method
Verified exampleOnboarding

One default beat more persuasion

Automatic enrollment moved participation from 37% to 86%.

Madrian & Shea 401(k) participation study

Use beneficial defaults with clear, easy opt-out—not dark patterns.

See evidence and method
Verified exampleOffer

Six choices converted 10× better than 24

The jam study saw 3% purchase with 24 options versus 30% with six.

Iyengar & Lepper choice-overload experiment

The right assortment can differ between discovery and checkout.

See evidence and method
Verified exampleConversion

Asking when turned intent into action

Writing a date and time raised flu-shot completion to 37.3%.

Milkman et al. implementation-intention field experiment

Retest the idea with your actual booking or calendar flow.

See evidence and method
Verified exampleCTA

A reason helped even a small ask

Copier-line compliance moved from 60% to 93–94% with a reason.

Langer, Blank & Chanowitz persuasion experiment

Large asks need a substantive reason, not a formulaic one.

See evidence and method
Verified examplePromotion

A match helped; a bigger match tied

Donation matching helped, but 1:1, 2:1, and 3:1 were near-ties.

Karlan & List field experiment with about 50k prior donors

Near-ties should be withheld, not turned into a fake winner.

See evidence and method
Failure boundaryFailure boundary

Premium-product customers churned more

In telco data, fiber customers churned more, reversing the intuitive prior.

Documented reversal in the public Telco churn dataset

Synthetic panels must not finalize churn or market-entry calls.

See evidence and method
Failure boundaryFailure boundary

New visitors converted better than returning visitors

One ecommerce dataset ran opposite to the common intuition.

Documented reversal in the Online Shoppers dataset

Close segment-priority decisions with observed traffic or a pilot.

See evidence and method
Failure boundaryFailure boundary

The cheapest option was not the winner

In intercity travel, speed and convenience beat the cheapest fare.

ModeCanada observed-choice data

Price-only models fail when non-price value dominates.

See evidence and method
ReproduceMCP recipe

Which copy should I A/B first?

Add an audience, two variants, and a metric for a directional memo and test design.

worldclone_decide · copy_ab

Copy output is a hypothesis, never a confirmed winner before a real A/B.

See evidence and method
ReproduceMCP recipe

Which first customer segment should I narrow to?

Compare two segments for the same product and choose the first pilot.

worldclone_forecast_segments · segment_priority

Priority is directional and needs an observed channel pilot.

See evidence and method

Recent published signals

No community posts yet.

Start with the evidence and MCP recipes above, or share the first decision memo.