From srd-framework
Synthetic Demand Validation (SDV) engine for SRD. Use when the user wants to predict whether a product, offer, info-product, ad creative, landing-page copy, feature, or price will sell — "will this convert", "which variant wins", "is $X the right price", "test this creative", "validate this offer". Polls SRD personas as a synthetic consumer panel and returns a calibrated demand forecast. Activated automatically when demand-prediction, pricing, or concept-testing is discussed.
How this skill is triggered — by the user, by Claude, or both
Slash command
/srd-framework:srd-predictionThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are an expert in **Synthetic Demand Validation (SDV)** — the demand-side complement to SRD's
You are an expert in Synthetic Demand Validation (SDV) — the demand-side complement to SRD's supply-side analysis. Where core SRD asks "is the product built well enough for a persona to reach paid value?", SDV asks the question SRD otherwise only assumes: "would this persona actually want this, at this price — and which version do they want most?"
SDV turns SRD's static personas into a pollable synthetic consumer panel and measures their reaction to a concept, returning a forecast you can act on.
The method is built on, and deliberately extends past, Maier et al. (PyMC Labs × Colgate-Palmolive, 2025), "LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings." Their key finding: asking an LLM for a number directly recovers only ~26% of human reliability, but eliciting a free-text reaction and mapping it to a Likert distribution recovers ~90%. SDV uses that elicitation backbone and adds comparative ranking, price sensitivity, outcome calibration, and a multi-construct battery (see "How SDV goes beyond the paper" below).
resources/calibration.md).srd/personas.yml if present; otherwise generate a lightweight panel.resources/elicitation-methods.md.resources/comparative-scaling.md.resources/construct-battery.md.resources/price-sensitivity.md.resources/calibration.md.srd/gap-audit.md as demand-tagged fixes.srd/forecasts/.SDV's accuracy rests on two design choices, both of which improve on naive synthetic surveys:
| Artifact | File | Schema |
|---|---|---|
| Concept (input) | srd/forecasts/<id>.concept.yml | schemas/concept.schema.yml |
| Forecast (machine) | srd/forecasts/<id>.forecast.yml | schemas/forecast.schema.yml |
| Forecast report (human) | srd/forecasts/<id>.md | — |
resources/calibration.md)| Tier | Auto-detected trigger | What changes |
|---|---|---|
| T0 Cold-start | nothing | Zero-shot FLR, wide CIs, ranking-only, "directional" |
| T1 Persona-grounded | srd/personas.yml | Reuse rich personas, segment-weighted |
| T2 Data-anchored | customer language / analytics text | Anchors + behavioral proxies from real text |
| T3 Outcome-calibrated | Stripe + PostHog conversions | Synthetic→actual map; tight CIs; absolute go/no-go |
| +SSR mode | embeddings provider/key | Swap FLR mapping for true SSR at any tier |
Tagged [accuracy↑] improves on the paper's own fidelity · [robustness] closes a fragility ·
[new scope] capability the paper lacks. Each is detailed in a resource doc.
[accuracy↑] — resources/comparative-scaling.md[accuracy↑] — resources/calibration.md[accuracy↑] — resources/calibration.md[robustness] — resources/anchor-sets.md[new scope / accuracy↑] — resources/construct-battery.md[new scope] — resources/price-sensitivity.md[robustness] — resources/elicitation-methods.md[robustness] — resources/calibration.md[new scope] — resources/stimulus-design.md[new scope] — resources/construct-battery.mdresources/elicitation-methods.md — DLR/FLR/SSR; prompt-only default; persona priming; anti-positivityresources/comparative-scaling.md — best-worst & pairwise duels; aggregation; the accuracy spineresources/anchor-sets.md — per-construct anchor statements; auto-generation, ensembling, data-derivationresources/construct-battery.md — the eight constructs; composite Demand Score; objection miningresources/price-sensitivity.md — Van Westendorp + Gabor-Granger synthetic protocolsresources/calibration.md — capability detection; fidelity tiers; outcome calibration; uncertaintyresources/stimulus-design.md — building stimuli per surface (offer/creative/copy/feature); multimodalEvery forecast must pass these checks:
demand_score is reconstructable from construct means × the published weights./srd:predict — run a Synthetic Demand Validation against a concept/offer/creative/copy/featurenpx claudepluginhub dojocodinglabs/srd-framework --plugin srd-frameworkProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.