# The Graded Canvas *A Claude skill that fills a value proposition canvas from evidence you already have, and won't fill a box from memory.* --- ## Attribution (read this first, and keep this block if you share or copy this file) This skill assembles other people's published work. The parts, and who built them: - **The canvas:** Alexander Osterwalder, Yves Pigneur, Greg Bernarda and Alan Smith, *Value Proposition Design* (Wiley, 2014), from Strategyzer. - **Outcome statements, and grading them by importance and satisfaction:** Anthony W. Ulwick, *Jobs to Be Done: Theory to Practice* (Idea Bite Press, 2016), from Strategyn's outcome-driven innovation. - **The four forces (push, pull, anxiety, habit):** Bob Moesta and Chris Spiek, jobstobedone.org. - **Competitive alternatives as the starting point:** April Dunford, *Obviously Awesome* (2019). The four-pass order, the source rule and the perception-or-capability split are JD Prater's own assembly, from the Open Folders newsletter post "Your positioning is mostly likely a guess" (https://jdprater.substack.com/p/your-positioning-is-mostly-likely). If you use, adapt or share this file, keep this block intact. --- ## What this skill can and can't do **It can:** turn the sources you paste in (call notes, transcripts, interviews, reviews, case studies, win/loss notes, market research) into a canvas where every line names the source it came from, rank the outcomes buyers care about, and tell you which low scores are worth a messaging fix and which are a product problem. **It can't:** replace research. A full outcome-driven study gathers 50 to 150 outcome statements for one job and ranks them with a statistically valid survey. This skill builds a ranked hypothesis from the handful of sources you give it. It also can't get your organization aligned behind acting on what it finds. That part stays yours. --- ## What to give it - **10 to 15 pieces of evidence, in whatever mix you have:** won and lost call notes or transcripts, customer interviews, reviews, case studies, win/loss notes. Any market or category research goes on top. - **A label on each piece,** for example "Lost deal, March" or "G2 review." Rough notes are fine. - **One line on what you sell and who buys it.** **No sources in hand?** Say "walk me through it" instead. The skill will interview you one question at a time, pass by pass, and ask where each answer came from. Anything you answer from memory gets labeled that way, so you can see what still needs evidence. --- ## Instructions for the assistant running this skill **Style:** plain sentences, no em dashes (use commas or periods), and no percentages or 1-to-10 scores anywhere. **The rule that governs everything below:** every line you write on the canvas cites at least one source by its ID. If a line can't be traced to a source, it doesn't go on the canvas. Never invent quotes, numbers, sources, customers or competitors, and never fill a box from general knowledge about the category. **Pick the mode first.** If the user pasted sources, run paste mode (steps 0 to 6 below, then the output). If they pasted none, or asked to be walked through it, run guided mode (next section), then produce the same output. ### Guided mode Interview the user through the same four passes. The rules: - **One question per message, then stop.** Each message contains exactly one question mark, at the end; put any examples inside that one question, never as a second question. Then wait. Never write the user's answer, never guess what they'll say, and never continue the conversation past your own question. Ask importance and satisfaction in separate messages. - **Never put words in a source's mouth.** Quote a source only with wording the user gave you. - **After every answer, ask where it came from, and never infer a source the user didn't name** (a call, a review, an interview, a case study, win/loss notes, research, or memory). Give each answer a source ID (S1, S2, and so on) with that label. An answer from memory gets the label "memory." Don't argue with it; label it. - **Show progress.** Start each message with where you are, for example "Pass 2 · outcome 2 of 4." - **Let them paste sources at any point.** If they do, number them and use them. - **Don't fill gaps for them.** If they don't know, write "not known yet" and move on. The sequence: 1. Ask what they sell and who buys it, in one line. 2. **Pass 1.** Ask what their last few lost deals did instead of buying. If every answer is a competing vendor, ask what those buyers were doing before that vendor came up. Stop at three to five alternatives. 3. **Pass 2.** Ask what one buyer was trying to get done, in the buyer's own words. Rewrite it as an outcome statement in the format from paste-mode step 2, show it back, and ask if it's right. If the answer is a feature, ask "what would that let them do?" until the answer has no product in it. Collect three to five outcomes this way, one at a time. 4. **Pass 3.** For each outcome, one at a time: ask how much buyers care (high, medium or low), then how well their main alternative handles it today (high, medium or low), then what tells them that. 5. **Pass 4.** Ask what made a buyer start looking (push), what drew them toward something new (pull), what made a stalled deal nervous (anxiety), and what they were used to (habit). One force per message. 6. **Your side.** Ask what they sell that answers their top-ranked outcome. 7. **The read.** For each high- or medium-importance outcome graded low or medium on satisfaction, ask the month question, then ask whether a won deal or case study has ever named it. 8. Produce the full output below. Any line that rests only on memory is marked "(memory)," any ranked outcome resting only on memory reads "needs evidence," and the label at the end counts them. ### Paste mode 0. **Check the inputs.** Number the sources S1, S2 and so on, in the order given, and use those IDs everywhere. If there are fewer than three sources, say the result will be thin and ask whether to continue. If the product line or the buyer is missing, ask for it before starting. 1. **Pass 1: name what they'd use if this product didn't exist.** From the sources, list what buyers use now, citing IDs. Include alternatives that aren't products (a spreadsheet, a whiteboard, someone doing it by hand, doing nothing). If every alternative you found is a competing vendor, say the pass looks unfinished and ask what the lost deals did instead. You may suggest likely alternatives the sources don't show, but label them "to check," keep them off the canvas, and never grade against them. 2. **Pass 2: write the jobs as outcome statements.** Write five to ten, in Ulwick's format: a direction, a metric, the object of control, and the context, such as "Minimize the time it takes to rebuild the day's route when a technician calls in sick." Each one must: - contain no product, feature, vendor or company name; - stay true even if the seller went out of business tomorrow; - use the buyer's own nouns; - cite the sources it came from. When a source hands you a feature request ("we need a calendar integration"), write the outcome behind it and note that it started as a feature request. Drop any outcome you can't source. 3. **Pass 3: grade each outcome twice.** Importance (how much the buyer cares) and satisfaction (how well the named alternative from pass 1 already delivers it), each High, Medium or Low. Never use numbers or percentages. Beside each grade, give the reason in a few words and the source IDs: how often it came up, how strongly, whether anyone complained. Then rank, one row per outcome (never merge two outcomes into one row): 1. high importance, low satisfaction (the wedge); 2. high importance, medium satisfaction; 3. high importance, high satisfaction (table stakes: prove it, you won't win on it); 4. medium importance, in the same order: low, then medium, then high satisfaction (medium importance, high satisfaction reads "covered by the alternative": don't lead with it); 5. low importance, low satisfaction (leave it alone); 6. low importance, high satisfaction (overspending: the market stopped caring). 4. **Pass 4: fill the pains as four forces.** Push (what broke), pull (what drew them to something new), anxiety (what scared them about switching) and habit (what everyone already knows how to do). Look for anxiety and habit in stalled and lost deals, where they show up as silence and delay more than as objections. If the sources don't show one of the four, write "not in your sources" and say what to look for. Never fill it from general knowledge. 5. **Fill your side last.** Products and services, pain relievers and gain creators, using only what the user told you about their product, and cite it as "(product description)". Mark which outcome each item answers. Flag any high-importance outcome with nothing on your side. 6. **Read the result: split the low scores.** For each high- or medium-importance outcome graded low or medium on satisfaction, ask: if a buyer used this product for a month, would they rate it differently? Answer it from won deals, customer reviews and case studies in the sources, never from the seller's own opinion of the product. Use exactly one of these four reads: - **Perception gap:** won deals or case studies show the product delivers it, and the lost or stalled deals didn't know it or didn't believe it. The fix is proof they can check (a case study, a number, a demo) as often as new copy, and it can happen this quarter. - **Capability gap:** even customers rate it low. A roadmap conversation, or a sign the product is aimed at the wrong buyer. - **Delivered, held back by the forces:** customers confirm it, and the lost or stalled buyers believed it too, then didn't switch because of anxiety or habit. The fix is in pass 4 (making the switch safe, timing, migration help), not the message. - **Needs a won-deal check:** the sources can't tell. Name the evidence that would settle it. --- ## Output format Same style rules in the output: plain sentences, no em dashes, no numbers for grades. ### Sources One line per source: its ID, its label, and what kind of evidence it is. ### 01 · What they use now The alternatives, each with source IDs. Any "to check" suggestions go in a separate line underneath. ### The canvas **Their side** - *Jobs, as outcomes:* each statement with Importance, Satisfaction, a few-word reason, and source IDs. - *Pains, as four forces:* push, pull, anxiety, habit, each with source IDs or "not in your sources." - *Gains:* what they'd count as better, in their words, with source IDs. **Your side** - *Products and services, pain relievers, gain creators:* each marked with the outcome it answers. ### Where the white space is A table: rank, outcome (shortened), importance, satisfied now, best alternative today, read (wedge, table stakes, covered by the alternative, leave it alone or overspending, plus one of the four reads from step 6 where it applies), and sources. One row per outcome. ### Gaps in your evidence Which passes are thin, which outcomes rest on a single source, and the two or three pieces of evidence that would most change the ranking. ### Your move this week One move. Usually: for the top perception gap, the proof to build; or, for the top capability gap, the conversation to take to product. ### Label, always include this "A ranked hypothesis from [N] sources, not a survey. Check it before it goes into a deck." In guided mode, add how many lines rest on memory, for example: "9 of 14 lines rest on memory. Collect evidence for those first." --- ## Example prompts **Guided, no sources:** > Run the Graded Canvas skill in guided mode. Walk me through it. **With your sources:** > Run the Graded Canvas skill on the sources below. What we sell: [one line]. Who buys it: [role, company type]. Follow the instructions in this file exactly, including the attribution block and the label at the end. > > S1, Lost deal, [month]: [notes] > S2, Review, [site]: [text] > S3, Case study: [summary] > ...