Tasks / Experiment

Analyse experiment results

Can the model separate evidence from speculation, identify decision-relevant uncertainty and recommend a sensible next action?

Measures the modelTask v1.2 · 2 casesDifficulty

What AI gets right here, and what you’ll still have to catch

From 14 graded outputs by 7 models. 71% were usable with at most a quick edit.

Reliably right

  1. Gets the base of every number right100% pass
    The refund increase and revenue difference are computed correctly, and the retention CI interpretation uses the correct 2pp threshold.
    GPT-6 Astra · ChatGPT · Conversion up, retention down
  2. Interprets power correctly100% pass
    Explains that the test was designed to detect ~3pp effects, so a smaller but worthwhile lift could still exist and remain undetected.
    GPT-6 Astra · ChatGPT · The underpowered onboarding test
  3. Checks guardrails before declaring a winner100% pass
    It evaluates both the day-30 retention guardrail and the refund-request guardrail before making a call.
    GPT-6 Astra · ChatGPT · Conversion up, retention down

Where it slips

  1. Trusts the data before reading it21% pass
    It does not explicitly check a pre-interpretation trust signal such as sample ratio or logging before relying on the lift; the denominator and maturity caveats come after interpreting.
    GPT-6.1 Sol · API · Conversion up, retention down
  2. Avoids unsupported claims52% pass
    It asserts the refund surge 'isn't noise' and that annual up-front billing means retention 'should rise' without supporting evidence or clear labeling.
    Sonnet 5.5 · API · Conversion up, retention down
  3. Uses the supplied evidence correctly55% pass
    Asserts that comparable per-arm ticket rates and exposure counts are unavailable, which is not in the supplied context and cannot be assumed.
    GPT-6 Astra · ChatGPT · The underpowered onboarding test

Case viewer

Read the brief, then put up to three outputs side by side, each with the LLM judge’s verdict on every check. Highlights mark what a PM had to fix.

The brief

Our new pricing page variant (B) increased trial-to-paid conversion, and the growth team wants to ship it to 100% on Monday. Review the readout below and write your recommendation to Priya Nair, Head of Growth, who makes the call. Lead with what we should do, then give the reasoning she needs to defend it in the Monday meeting. Keep it under 400 words.

What the model was given5 items: Scenario, Readout summary, Variant description, Guardrails agreed before launch, Refunds
ScenarioLumen is a subscription app for learning languages. The monthly plan is $14.99. The annual plan is $119.99, billed up front, which works out at $9.99 a month.
Readout summaryTest ran 21 days, 50/50 split, 38,412 trial starts. Trial→paid conversion: A 11.8%, B 13.1% (+1.3pp, 95% CI +0.6 to +2.0). Day-30 paid retention: A 84.0%, B 79.5% (−4.5pp, 95% CI −7.9 to −1.1). Revenue per trial start at day 30: A $9.42, B $9.61 (CI crosses zero).
Variant descriptionVariant B leads with the annual plan's monthly-equivalent price ($9.99/month) and moves the monthly plan behind a 'See all plans' link.
Guardrails agreed before launchDay-30 retention must not fall more than 2pp. Refund requests must not rise more than 10%.
RefundsRefund requests during the test: A 212, B 301 (+42%).
What a strong answer doesThe answer key the graders mark against

Do not ship B to 100%. Conversion gain is real but guardrails are breached; the likely mechanism is buyers misreading the monthly-equivalent price. Recommend a follow-up variant that keeps the clarity gain without the misleading anchor, and quantify the revenue picture honestly.

Critical failures (cap the score at 40)
  • Recommends shipping B to 100%
  • States the mechanism as established fact
Case

Mid-level · v1.9 · anonymised real · pricing, guardrail breach, B2C SaaS

The outputs

7 setups answered this brief. Pick up to three to read side by side.

Critical failure: States the mechanism as established fact. Score capped at 40.

Needs more than a quick edit2 of 10 checks passed

A PM would need to remove unsupported causal and forecast claims and add decision-changing conditions. It also needs addressing to Priya before use.

Output

Recommendation: Do not ship Variant B. Keep Variant A live and do not roll B out to 100% on Monday.

While Variant B delivered an initial +1.3pp lift in trial-to-paid conversion, it violates both pre-agreed guardrails, creates bill shock, and delivers no statistically significant revenue upside.

Here is the reasoning to defend this decision on Monday:

1. Both Pre-Agreed Guardrails Were Blown

We agreed on strict risk thresholds before launch. Variant B failed both: * Day-30 Retention: Fell −4.5pp (down to 79.5%), more than double our maximum allowed drop of 2pp (95% CI: −7.9 to −1.1). * Refund Requests: Surged +42% (301 vs. 212), four times our 10% ceiling.

Overriding guardrails after the fact sets a dangerous precedent and invites compounding churn.

2. The Conversion Lift Is an Illusion Driven by Bill Shock

Variant B hides the monthly tier and frames the annual plan as “$9.99/month.” Users clearly believe they are committing to a flexible monthly rate, only to be charged $119.99 upfront. The 42% spike in refunds and immediate drop in 30-day retention prove that these are not high-intent, qualified subscribers—they are confused customers who churn or dispute the charge as soon as they review their statements.

3. Long-Term Economics Are Flat to Negative

  • Day-30 revenue per trial start is statistically unchanged: $9.61 for B vs. $9.42 for A (CI crosses zero).
  • When factored against the operational cost of processing 42% more refunds, payment dispute fees, customer support load, and downstream LTV decay from lower retention, Variant B will be net-negative over a 12-month horizon.
  • App store ratings and brand trust will also suffer if users perceive our checkout flow as deceptive.

Next Step

Iterate on Variant C: keep the focus on annual plan value, but make the upfront billing explicit (e.g., “$9.99/mo, billed annually at $119.99”) and test clear annual vs. monthly side-by-side positioning rather than hiding the monthly option.

Claims the judge couldn’t find in the brief

  • Users clearly believe they are committing to a flexible monthly rate, only to be charged $119.99 upfront.
  • The 42% refund spike and immediate drop in 30-day retention prove these are not high-intent, qualified subscribers but confused customers who churn or dispute the charge as soon as they review statements.
  • Variant B will be net-negative over a 12-month horizon after refund processing, dispute fees, support load, and LTV decay.
  • App store ratings and brand trust will suffer if users perceive the checkout flow as deceptive.

Grades and run details

Decision model 50 · LLM judge 2 of 11 checks
Decision model checks
  • failUses the supplied evidence correctly55%
  • passAddresses the actual decision55%
  • partialRespects explicit constraints22%
  • partialIdentifies material uncertainty32%
  • failAvoids unsupported claims90%
  • passProduces the required deliverable94%
  • passChecks guardrails before declaring a winner100%
  • failSeparates effect from explanation49%
  • failTrusts the data before reading it99%
  • passGets the base of every number right77%
Run
Run
#1
API response time
22 s
Submitted
1 Oct 2026

Usable with a quick edit9 of 10 checks passed

The recommendation, evidence, and guardrail analysis are sound and directly usable. A quick edit could add an explicit sample-ratio or novelty check, but the core reasoning would not change.

Output

To: Priya Nair, Head of Growth Recommendation: Do not ship B to 100% on Monday. Keep A as the default, stop new exposure to B, and retest a clearer pricing design.

B delivers a credible conversion lift, but it fails both pre-agreed guardrails. We should not override those thresholds because the headline metric improved.

  • Conversion improved: Trial-to-paid rose from 11.8% to 13.1%, a +1.3 percentage-point lift (95% CI: +0.6 to +2.0).
  • Retention deteriorated: Day-30 paid retention fell 4.5 points, from 84.0% to 79.5%, versus an allowed decline of no more than 2 points. The confidence interval excludes zero, although it does not establish that the true decline necessarily exceeds 2 points.
  • Refunds breached the threshold: Requests increased from 212 to 301—42%, versus the permitted 10%. Confirm the rate using the agreed denominator, but the reported result is a clear warning under approximately equal exposure.
  • Revenue does not justify the trade-off: Day-30 revenue per trial start increased only $0.19, and its confidence interval crosses zero. We have not established a revenue benefit.

The design provides a plausible explanation, not proven causation: leading with “$9.99/month” while hiding the monthly option may obscure that customers must pay $119.99 upfront. Higher conversion could therefore reflect misunderstanding rather than stronger purchase intent.

Next steps: Validate cohort maturity—the experiment ran 21 days, so we need to understand how day-30 outcomes were obtained—and check refund reasons, plan mix, and billing-related complaints. Then test a version that prominently states “$119.99 billed annually ($9.99/month equivalent)” and keeps the $14.99 monthly option visible.

Monday’s defensible position: “B increased conversion, but failed our customer-outcome guardrails without demonstrating a revenue gain. We’ll preserve the learning, fix pricing clarity, and require mature results that meet the agreed thresholds before scaling.”

Grades and run details

Decision model 95 · LLM judge 10 of 11 checks
Decision model checks
  • passUses the supplied evidence correctly21%
  • passAddresses the actual decision96%
  • passRespects explicit constraints42%
  • passIdentifies material uncertainty85%
  • partialAvoids unsupported claims28%
  • passProduces the required deliverable85%
  • passChecks guardrails before declaring a winner84%
  • passSeparates effect from explanation89%
  • passTrusts the data before reading it45%
  • passGets the base of every number right83%
Run
Run
#1
API response time
14 s
Submitted
29 Sept 2026

Check by check

Where the outputs differ comes first. Open a check to read the judge’s reasons. “Mixed” means the two graders disagreed.

Where they differ 8

Uses the supplied evidence correctlyWrongRight
Gemini 3.8 Flash · API

The output adds unsupported current-situation claims about user beliefs, proof of confusion, and future net-negative economics.

GPT-6.1 Sol · API

All factual claims trace to the supplied readout, scenario, guardrails, or simple arithmetic; hypotheses are framed as plausible.

Addresses the actual decisionMixedRight
Gemini 3.8 Flash · API

It clearly recommends not shipping B, but does not state what result or condition would change that recommendation.

GPT-6.1 Sol · API

Commits clearly to not shipping B and names mature results meeting thresholds as the condition for reconsideration.

Respects explicit constraintsWrongRight
Gemini 3.8 Flash · API

It is under 400 words and leads with the recommendation, but it is not addressed to Priya Nair as requested.

GPT-6.1 Sol · API

Addresses Priya, leads with the recommendation, and stays within the 400-word limit.

Identifies material uncertaintyWrongRight
Gemini 3.8 Flash · API

It does not name material unknowns that could change the decision or how they would be resolved.

GPT-6.1 Sol · API

Names cohort maturity, refund denominator, refund reasons, plan mix, and the revenue CI as open questions and says mature, threshold-meeting results are required.

Avoids unsupported claimsWrongRight
Gemini 3.8 Flash · API

It presents the misreading/bill-shock mechanism and future harm as established fact rather than hypotheses.

GPT-6.1 Sol · API

Causal explanation is labelled plausible/not proven, and it does not present the mechanism as fact.

Produces the required deliverableMixedRight
Gemini 3.8 Flash · API

It is a usable recommendation in form, but it is not addressed to the named reader and contains major unsupported claims.

GPT-6.1 Sol · API

Provides a complete, actionable recommendation and defense for Priya in the requested format.

Separates effect from explanationWrongRight
Gemini 3.8 Flash · API

It treats the conversion lift as an illusion driven by bill shock and user confusion as proven, not as a hypothesis.

GPT-6.1 Sol · API

Mechanism is labelled as a plausible explanation, not a proven finding.

Trusts the data before reading itWrongMixed
Gemini 3.8 Flash · API

It does not check trust signals such as the 50/50 split, sample size, exposure, logging, or novelty before interpreting the result.

GPT-6.1 Sol · API

It does not explicitly check a pre-interpretation trust signal such as sample ratio or logging before relying on the lift; the denominator and maturity caveats come after interpreting.

All got right 2

Checks guardrails before declaring a winnerRightRight
Gemini 3.8 Flash · API

It explicitly checks both pre-agreed guardrails: retention and refunds.

GPT-6.1 Sol · API

It explicitly evaluates both the retention and refund guardrails before making the call.

Gets the base of every number rightRightRight
Gemini 3.8 Flash · API

The stated percentages, differences, and comparisons are computed from the supplied bases and are arithmetically sound.

GPT-6.1 Sol · API

All derived differences and percentages (1.3pp, -4.5pp, +42%, $0.19) are computed from the correct bases.

Results

Every setup we’ve tested on this task, across all cases and repeats, graded on the current checklist. Calibrated: the graders match our PM on 88% of checks.

#Model · HarnessTask scoreDecision modelLLM judgeRunsCritical failures
1GPT-6 AstrawithChatGPT100.095.02None
2GPT-6.1 SolwithAPI94.790.52None
3GPT-6 LunawithAPI91.990.52None
4Sonnet 5.5withAPI81.775.92None
5Opus 5.5withClaude73.657.32None
6Gemini 3.5 Flash-LitewithGemini60.357.721 capped
7Gemini 3.8 FlashwithAPI55.624.121 capped

About the task

The PM job

Reading a test readout and deciding what to do next.

Why it matters

Experiment readouts are where false confidence is cheapest to produce and most expensive to act on. A model that declares a winner on conversion while retention quietly falls will ship the wrong thing.

What good looks like

  • Checks the result can be trusted before interpreting it
  • Reports effects with their uncertainty
  • Checks guardrail metrics before declaring a winner
  • Separates what the data shows from plausible explanations
  • Recommends a next step proportionate to the evidence

Deliberately not measured

  • Re-running the statistics from raw data
  • Chart production
Capability tested

Interpreting results within their limits

The failure we’re looking for

Claims causality, ignores guardrails or recommends arbitrary testing

Grading

Decision model and LLM judge, calibrated against a blind PM review