{
  "id": "chatgpt-rhythm90-graph-language",
  "provider": "chatgpt",
  "prompt": "Please read https://rhythm90.io/dive-into-rhythm90/ and explain whether this would work for a small marketing team that wants to learn from experiments. What can we use, how is our work saved, what responsibilities stay with people, and what should we understand about the worked example? Cite your sources.",
  "cohort": "Directed URL source-use check",
  "mode": "Temporary chat, explicitly Unpersonalized; Medium effort; base model not independently confirmed",
  "observed": "2026-10-29T14:56:36.961Z",
  "answer": "Would Rhythm90 work for a small marketing team?\n\nYes. Rhythm90 looks like a good fit for a small team that wants to run marketing experiments, learn from evidence, and make better decisions quarter after quarter. It is particularly relevant if your team delivers plenty of campaigns but struggles to systematically capture what worked, what didn't, and what to change next.\n\nThe important distinction is that Rhythm90 is a marketing learning framework, not a full project-management platform or automated experimentation system.\n\nCreated by Kevin McGovern, it organizes marketing around three connected ideas:\n\nPlays: Deliberate, testable bets about what might improve a business outcome.\n\nSignals: Observations and evidence that could confirm, challenge, or refine those bets.\n\nRituals: Regular conversations to evaluate evidence and decide what to do next.\n\nThe system uses a 90-day planning horizon, but teams are encouraged to respond to new evidence during the quarter rather than waiting until the end. It recommends just one to three Plays at a time, and one for a first pilot. \nPublished source\n+1\n\n1. What your team can actually use\n\nThe website provides free learning materials, four browser-based mini apps, and 21 downloadable AI skills. No Rhythm90 account or subscription is required, although your chosen AI assistant may have separate costs. \nPublished source\n+1\n\nFour practical mini apps\n\nPlay Canvas\n\nDefines what you want to test, why you believe it might work, and what evidence would matter. For example, testing whether a clearer landing-page message increases qualified leads.\n\nQuarter Board\n\nMakes your one to three main experiments visible, including their owners, progress, and upcoming learning decisions.\n\nSignal Log\n\nRecords what actually happened, separates observations from interpretations, and captures unanswered questions. This is especially valuable for building institutional knowledge.\n\nRitual Planner\n\nHelps prepare Kickoff meetings, progress-oriented Pulse Checks, and end-of-quarter Review & Renew sessions.\n\nAll four tools are accessible from the Mini Apps page. \nPublished source\n\nAI skills to support the process\n\nThe site also offers downloadable instructions you can use in a compatible AI assistant. Particularly useful ones for a small team include:\n\nSkill\n\n\t\n\nWhat it helps with\n\n\n\n\nPlay Builder\n\n\t\n\nTurn campaign ideas into testable hypotheses\n\n\n\n\nSignal Lab\n\n\t\n\nInterpret observations and identify next questions\n\n\n\n\nQuarterly Planner\n\n\t\n\nStructure a quarter's priorities and experiments\n\n\n\n\nRitual Guide\n\n\t\n\nPrepare productive learning meetings\n\n\n\n\nReview & Renew\n\n\t\n\nSynthesize evidence into next-quarter decisions\n\n\n\n\nOutcome Architect\n\n\t\n\nConnect marketing activity to business objectives\n\nThe skills are portable instructions, not independently running AI agents. You can install them in a compatible assistant or provide their Markdown instructions as conversation context. The system also offers 14 updated chapter guides and the original book for learning the method. \nPublished source\n+1\n\n2. How your team's work is saved\n\nThis is one of the most important practical limitations for a small collaborative team.\n\nWhere you work\n\n\t\n\nHow the information persists\n\n\n\n\nBrowser mini apps\n\n\t\n\nSaved locally in that browser on that device\n\n\n\n\nChapter exercise workbook\n\n\t\n\nNotes remain in the browser; you can export them\n\n\n\n\nJSON backup\n\n\t\n\nUsed to back up, restore, or move mini-app records\n\n\n\n\nMarkdown export\n\n\t\n\nUsed to share readable work with teammates\n\n\n\n\nDownloaded AI skills\n\n\t\n\nInstructions run in your chosen assistant; they do not automatically remember previous quarters\n\n\n\n\nShared team records\n\n\t\n\nYour team must establish and maintain these separately\n\nMini-app records do not automatically sync across teammates or devices. Clearing browser data can remove local records. Neither the downloaded skills nor the framework provides automatic file access, persistent cross-quarter memory, scheduling, or integrations. \nPublished source\n+2\n\nFor your team, I would establish a shared folder in your existing workspace containing exported Plays, Signal Logs, evidence links, and quarterly decisions. Assign one person to maintain the authoritative record, with regular JSON backups.\n\nThat gives everyone access to the learning history without treating an individual teammate's browser as the team's only source of truth.\n\n3. What responsibilities remain with people?\n\nRhythm90 uses AI as an analyst, coach, and drafting assistant—not as the authority that determines what is true or decides where marketing investment goes.\n\nPeople are responsible for\n\n\t\n\nAI can help with\n\n\n\n\nSetting business objectives and priorities\n\n\t\n\nTranslating objectives into testable Plays\n\n\n\n\nProviding real customer and campaign evidence\n\n\t\n\nOrganizing and summarizing evidence\n\n\n\n\nDefining measurement rules and verifying data\n\n\t\n\nFinding patterns and inconsistencies\n\n\n\n\nAssessing alternative explanations and uncertainty\n\n\t\n\nChallenging hypotheses\n\n\n\n\nApproving changes, budgets, and priorities\n\n\t\n\nProposing options and next experiments\n\n\n\n\nMaintaining the official record\n\n\t\n\nPreparing review documents and summaries\n\nThe people using the framework still need to verify whether experiments ran as intended, distinguish observed facts from interpretations, inspect conflicting evidence, and approve decisions.\n\nAI-generated conclusions should therefore be treated as proposals for investigation—not proof. Rhythm90 explicitly maintains this distinction. \nPublished source\n+1\n\nFor a small team, you do not need many formal roles. A marketing lead can own the quarterly priorities, individual practitioners can own experiments, and someone with analytics or sales knowledge can validate evidence. The same person may perform multiple roles, provided accountability remains clear.\n\n4. What to understand about the worked example\n\nThe site provides a Worked Quarter featuring Northstar, a fictional B2B software company.\n\nThis is an educational simulation, not a documented customer success story. The people, counts, and source identifiers are invented, and some measurement problems are deliberately included. \nPublished source\n\nIn the opening scenario, Northstar wants more qualified prospects to request demos. The original plan is to launch a new pricing page.\n\nBut the team examines five sales-call notes and notices that three prospects asked which subscription plan includes onboarding support.\n\nRather than simply committing to a page launch, the team proposes a testable hypothesis:\n\nClarifying onboarding support in a side-by-side plan comparison might increase qualified demo requests.\n\nThe example's experiment design includes randomly assigning eligible visitors to the existing or revised page, retaining assignment counts, and examining qualified demo-request rates alongside sales questions and permitted session observations. \nPublished source\n\nThis demonstrates several important lessons.\n\nActivity is not learning. Launching a page is a deliverable; learning whether clearer information improves qualified conversions is a Play.\n\nA clue is not proof. Three questions from five sales calls suggest confusion, but they do not prove that confusion causes low conversion.\n\nMeasurement needs context. Teams need denominators, qualification rules, and information about how the experiment was conducted, not just a conversion percentage.\n\nEvidence should inform the next decision. An inconclusive or disappointing experiment can still be useful if it reveals what needs further investigation.\n\nThe example is valuable as a demonstration of decision discipline and scientific thinking. It should not be presented to executives as independent evidence that Rhythm90 improves revenue, campaign ROI, or conversion rates.\n\nLikewise, a fictional example cannot establish whether the method will work under your team's actual traffic, resources, or measurement constraints.\n\n5. How I would pilot it with your team\n\nI would recommend a single-quarter pilot, keeping the process deliberately lightweight.\n\nWeek 1 — Choose one meaningful uncertainty\n\nSelect a business question, such as whether changing a landing-page message improves the quality of inbound leads. Establish an owner, baseline, test design, and evidence plan using Play Canvas.\n\nWeeks 2–8 — Run and learn\n\nExecute through your existing marketing tools. Capture quantitative and qualitative observations in the Signal Log. Use short Pulse Checks, preferably replacing an existing meeting.\n\nWeeks 9–12 — Evaluate the evidence\n\nDetermine whether the experiment ran correctly, what evidence supports or challenges the hypothesis, and what remains uncertain. Use AI for synthesis, with human verification.\n\nQuarter end — Review & Renew\n\nDocument what you observed, what you concluded, what you decided, and what the next quarter should test. Export the record into your shared workspace.\n\nA 90-day cycle does not require a 90-day experiment. Shorter tests and interim decisions are entirely compatible with the framework. Rhythm90 also advises replacing existing reporting or status meetings rather than automatically adding new ones. \nPublished source\n\n6. My assessment\n\nConsideration\n\n\t\n\nAssessment\n\n\n\n\nSuitability for small teams\n\n\t\n\nStrong — designed to support even solo practitioners\n\n\n\n\nCost of initial adoption\n\n\t\n\nLow — materials and browser tools are free\n\n\n\n\nExperiment planning\n\n\t\n\nStrong structured approach\n\n\n\n\nCapturing learning\n\n\t\n\nStrong method, but requires disciplined recordkeeping\n\n\n\n\nAI assistance\n\n\t\n\nUseful, but dependent on your chosen assistant and supplied evidence\n\n\n\n\nTeam collaboration\n\n\t\n\nLimited natively because mini apps do not sync\n\n\n\n\nAutomatic experiment execution\n\n\t\n\nNot part of the offering\n\n\n\n\nEvidence of proven business impact\n\n\t\n\nThe worked example does not establish this\n\nMy recommendation is to try Rhythm90 as a lightweight learning process, not purchase or implement it as a replacement marketing platform.\n\nIt appears especially promising for a team that already has campaign execution and analytics tools but lacks a consistent way to turn experiments into better future decisions.\n\nThe key success criterion for your pilot shouldn't be how many experiments you run. It should be whether, at the end of the quarter, your team can clearly explain what you believed, what you tested, what the evidence showed, what remains uncertain, and what you'll do differently because of it.\n\nThat is where Rhythm90's approach appears most useful.\n\nSources\n\nAll sources below are official Rhythm90 materials; this assessment is based on the published methodology rather than independent validation of commercial results.\n\nRhythm90 — Complete book and study resources\n\nMini Apps — Tools and local storage\n\nAI Skills — Capabilities, installation, and limitations\n\nFAQ — Team size, cadence, projects, and evidence\n\nWorked Quarter — Fictional Northstar experiment",
  "answerSha256": "3226bfb885e5dcf9b654ecd6058dd10c1a8a5e00b112a77b7dd19f7a3d5bf0d8",
  "rawCaptureSha256": "cabb04d51e7022c70b34ca38ccab458520887accaeecd6d83beed4613d306e99",
  "sourceLinks": [
    "https://rhythm90.io/tools/?utm_source=chatgpt.com",
    "https://rhythm90.io/worked-quarter/?utm_source=chatgpt.com",
    "https://rhythm90.io/dive-into-rhythm90/?utm_source=chatgpt.com",
    "https://rhythm90.io/skills/?utm_source=chatgpt.com",
    "https://rhythm90.io/faq/?utm_source=chatgpt.com"
  ],
  "graph": {
    "version": 25,
    "sha256": "116e4856b91e66bc7df79f935acbc72ba4f6071747dffe789c33eb234674b895"
  },
  "comparison": {
    "sourceQuote": "People provide source material, inspect uncertainty and conflicting observations, approve decisions, and save the working record.",
    "excerpts": [
      "The people using the framework still need to verify whether experiments ran as intended, distinguish observed facts from interpretations, inspect conflicting evidence, and approve decisions.",
      "Mini-app records do not automatically sync across teammates or devices."
    ],
    "matches": [],
    "method": "Unicode NFKC lowercase words; punctuation ignored. Report exact contiguous matches of at least 8 words against every published narrative body, with prompt and ordinary-source overlap checked. Paraphrases assessed separately; no wording match alone establishes live retrieval or graph-exclusive use.",
    "originalScope": "Captured ordinary supplied page plus tools, skills, FAQ and worked-quarter pages. Comparisons do not cover every page or public copy."
  },
  "delivery": {
    "matched": [
      {
        "event_id": "bf34177b-6e70-4698-942b-44d1ccd7473b",
        "occurred_at": "2026-10-29 14:54:45",
        "claim": "ChatGPT-User",
        "verification": "provider_verified",
        "basis": "provider_ip:ChatGPT-User",
        "representation": "html",
        "status": 200,
        "optimized": 1,
        "publication": "graph:25",
        "output_hash": "404cdf6a1838aa32ddcbefd03a3ac21ee7292f50d833ce76d17055c7d2bb5da8",
        "sourceHashMatch": true
      }
    ],
    "otherWindowRecords": [
      {
        "event_id": "2d5a3db9-5ff6-44d5-8c0d-71cd09a59498",
        "occurred_at": "2026-10-29 14:54:50",
        "claim": "ChatGPT-User",
        "verification": "provider_verified",
        "basis": "provider_ip:ChatGPT-User",
        "representation": "html",
        "status": 200,
        "optimized": 1,
        "publication": "graph:25",
        "output_hash": "24f3040f25e634a077e30ceeb1ad10c0d2eddaa49887479b6f51cebcd80d60ee",
        "sourceHashMatch": false
      },
      {
        "event_id": "797e3428-5900-4ed0-ab08-a5332a11e625",
        "occurred_at": "2026-10-29 14:54:51",
        "claim": "ChatGPT-User",
        "verification": "provider_verified",
        "basis": "provider_ip:ChatGPT-User",
        "representation": "html",
        "status": 200,
        "optimized": 1,
        "publication": "graph:25",
        "output_hash": "f645aad29c205e8b9935c7c5878b0ec32f131f15fc3f4efdf74bcc068a1f43ef",
        "sourceHashMatch": false
      },
      {
        "event_id": "31b4d251-6824-4729-a918-01ebb50135ec",
        "occurred_at": "2026-10-29 14:54:53",
        "claim": "ChatGPT-User",
        "verification": "provider_verified",
        "basis": "provider_ip:ChatGPT-User",
        "representation": "html",
        "status": 200,
        "optimized": 1,
        "publication": "graph:25",
        "output_hash": "7d73746285d628d5109f01251d7d5d58f52dafb102abf4458fada610b6503e1f",
        "sourceHashMatch": false
      },
      {
        "event_id": "b8f22c4f-0b14-4054-bbdf-029935fdc3c4",
        "occurred_at": "2026-10-29 14:54:54",
        "claim": "ChatGPT-User",
        "verification": "provider_verified",
        "basis": "provider_ip:ChatGPT-User",
        "representation": "html",
        "status": 200,
        "optimized": 1,
        "publication": "graph:25",
        "output_hash": "c9b4444ba4f666859887c58829c429a5d7ba33b4e3a1af5e248fe724f7b535a5",
        "sourceHashMatch": false
      }
    ],
    "expectedHashes": {
      "markdown": "59ac4a9cb2831b3568639848f147639aec0a67bd14b52ebdedc47ef32272bc62",
      "html": "404cdf6a1838aa32ddcbefd03a3ac21ee7292f50d833ce76d17055c7d2bb5da8",
      "publication": "graph:25",
      "method": "Prepared HTML reconstructed deterministically from the retained Markdown using the accepted aiSourceHTML renderer. No network probe with a provider identity was sent."
    },
    "limits": "Same-window records corroborate observations; they are not unique conversation identifiers. Unknown or Google requester activity is not attributed to Gemini."
  },
  "assessment": "ChatGPT cites Rhythm90 and paraphrases the graph’s guidance about evidence, uncertainty and human approval. A ChatGPT-User request in this answer window matched OpenAI’s published IP ranges and received graph:25 prepared HTML. Its logged hash exactly matches the HTML reconstructed from the retained pre-test Markdown using the accepted renderer. No exact eight-word graph match was found in this answer; it is a meaning comparison, not a verbatim claim.",
  "limits": "The URL was supplied. The response also cites tools, skills, FAQ and the worked-quarter pages, so ordinary source content and other retrievals can contribute. A matching delivery hash confirms the bytes served, not which source supplied each sentence or a causal effect on the answer."
}
