Market Research

Market Research

Applied market research for people who have to defend a number in a room. This skill is about the operational craft: constructing a market size two independent ways, reconciling the gap, cutting the market into segments that behave differently, and fielding survey instruments that do not manufacture the answer you hoped for.

When to use this skill

• Sizing a market for a board deck, investor memo, or funding request where the number will be challenged line by line
• Reconciling a TAM you inherited — an analyst report says $12B, your bottom-up build says $700M, and you need to explain the gap
• Segmenting a market before a pricing, packaging, or GTM decision
• Triangulating demand signals (search volume, inbound, win rates, analyst data, competitor headcount) into one directional read
• Designing a survey to answer a market question — willingness to pay, category awareness, switching intent — without leading the respondent
• Auditing someone else's sizing before you sign off on it

Inputs the skill expects

• The market definition in one sentence — including geography and buyer
• A top-down anchor (published market value) with its source and vintage
• Bottom-up unit economics — unit count, qualified share, annual value per unit
• The decision the number is feeding (investment size, hiring plan, pricing)
• Time horizon for SOM (1 year vs 3 years changes it by an order of magnitude)
• For surveys: population size, target margin of error, mode (panel, list, intercept)

Clarify First

Before generating, confirm these inputs. If any is unknown or vague, ASK — do not assume:

• Market definition — what is in and what is out — the single biggest driver of the number; "dental software" and "dental practice management software for multi-chair EU practices" differ by 20x
• The decision this sizing supports — a fundraise tolerates a wide TAM; a hiring plan needs a defensible SOM
• Time horizon for SOM — 12-month obtainable share and 3-year obtainable share are different artifacts
• Whether a published anchor exists and its vintage — a 2022 report in a 2026 memo needs an explicit growth bridge

Stop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.

Workflows

Workflow 1 — Build and reconcile TAM/SAM/SOM

1. Write the market definition sentence first. Everything downstream inherits it.
2. Build the top-down chain: published market value, then named filters that each cut it (geography, segment, buyer qualification), each with a retention fraction and a stated justification.
3. Build the bottom-up chain independently: unit count from a countable source, qualified share, annual value per unit, reachable share, expected win rate.
4. Run the builder. It computes both chains, reconciles them layer by layer, and flags implausible ratios and divergence.
5. Resolve every fail before the number leaves your machine. A warn needs a sentence in the memo, not a fix.

python3 research-ops/market-research/scripts/tam_sam_som_builder.py \
  --input research-ops/market-research/assets/sample_market_model.json \
  --format text

Workflow 2 — Triangulate demand signals

1. Collect every observable demand signal you have — search volume, inbound lead velocity, win rate by segment, analyst growth rates, competitor hiring, category conference attendance.
2. Score each for source independence and directional strength.
3. Run the triangulator to get a weighted demand index and, more importantly, the list of signals that contradict each other.
4. Investigate contradictions before averaging them away. A conflicting signal is usually a segmentation boundary you have not drawn yet.

python3 research-ops/market-research/scripts/demand_signal_triangulator.py \
  --input research-ops/market-research/assets/sample_demand_signals.json \
  --format text

Workflow 3 — Audit a survey instrument before fielding

1. Draft the instrument with the market question stated at the top.
2. Run the auditor. It checks each item for leading language, double-barrelled phrasing, absolutes, unbalanced or over-long scales, and missing escape options.
3. Check the sample-size verdict — it computes required n from population, target margin of error, and confidence level.
4. Fix every fail, then re-run. Field only on a clean run.

python3 research-ops/market-research/scripts/survey_instrument_auditor.py \
  --input research-ops/market-research/assets/sample_survey.json \
  --format text

Decision frameworks

Which sizing method for which situation

Plausibility thresholds

These are the ratios the builder enforces. They are heuristics, not laws — but crossing one without an explanation in the memo is how sizing loses credibility.