市场调研

HarborMug: US audience-validation decision brief

Recommend funding a bounded US audience-validation study before committing to expansion. Prioritize commuters, retain desk-first users as a comparison group, and exclude international expansion from this decision. Mia should prepare the study; Alex decides funding. No dollar research budget can be approved from these materials because recruitment and fielding costs were not supplied.

Data note: All business inputs and competitor observations are synthetic, unverified planning assumptions. No customers were contacted, no fieldwork occurred, and no business improvement was measured. Arithmetic is deterministic; empirical confidence is low.

Addressable market and funding implication

The planning window is one year, geography is the US, and the category assumes one 500ml mug per annual buyer at $24. Figures represent product revenue in USD/year, excluding taxes, shipping and refunds. TAM covers the full defined category; SAM is the assumed 40% that HarborMug can serve; SOM is assumed demand capped by planned supply.

Layer	Top-down calculation	Bottom-up calculation	Annual result, each method
TAM	E1 category spend: $48m	E2: 2m buyers × $24	$48m
SAM	$48m × 40%	2m × 40% × $24	$19.2m
SOM before capacity cap	$19.2m × 1% reachable share	800,000 serviceable buyers × 1% adoption × $24	$192,000
SOM after capacity cap	min($192,000, 8,000 × $24)	min(8,000 demand units, 8,000 capacity units) × $24	$192,000

Sources: supplied E1, E2 and E4, preserved in source evidence. The original sizing tool computes percentage-based SOM but does not implement a separate capacity input. The explicit cap above closes that gap. These are scenarios, not sales forecasts or profit estimates.

TAM divergence is |top-down − bottom-up| / top-down = 0%; SAM and SOM also match. Both views share category scope and price, and the buyer assumption was deliberately aligned with spend. This is arithmetic consistency, not independent validation. Do not average them or treat their agreement as stronger evidence. The consumer script allows 40% TAM divergence, but the decision brief requires ±20% planning tolerance; use the stricter 20% reconciliation threshold for this funding decision. Neither threshold is a statistical confidence interval.

The 1% obtainable-share assumption implies 8,000 buyers and consumes 100% of planned annual capacity. It has no demonstrated acquisition or conversion basis. The 8,000-unit limit is a new planning assumption, independent of current weekly orders; current sales do not establish expansion capacity or demand. A 60/40 split in research contacts does not establish national segment shares, so no national commuter or desk revenue allocation is defensible yet.

Adoption of the 800,000 serviceable buyers	Demand units	Units after supply cap	Product revenue
0.5%	4,000	4,000	$96,000
1.0%	8,000	8,000	$192,000
1.5%	12,000	8,000	$192,000

These are illustrative sensitivities, not probability bounds. More demand above 1% cannot increase revenue under the stated capacity. Validate category buyer counts, purchase frequency, the serviceable fraction and reachable demand before using the base case to fund supply expansion.

Ranked target segments

The original Kotler scorer weights measurable 15%, substantial 25%, accessible 25%, differentiable 20%, and actionable 15%. Substantiality and accessibility must each reach 40/100. With both gates passed, a weighted score of at least 65 yields TARGET; otherwise WATCH. Scores are E5 analyst judgments, not survey findings; the consumer profile leaves substantiality unchanged.

Priority	Segment	Measurable	Substantial	Accessible	Differentiable	Actionable	Weighted score	Verdict
1	Commuter	85	80	80	75	85	80.5	TARGET
2	Desk-first	80	70	65	70	80	71.8	TARGET
Exclude	International	70	75	20	60	20	49.2	DROP: accessibility

Commuters have a portable-drinkware need and a consented-list channel; study portability, spill concerns and existing alternatives. Desk-first users have a workspace-drinkware need and the same channel, with weaker assumed accessibility; retain them to test whether needs and response actually differ. These needs are hypotheses to test, not established product advantages. International fails the channel gate and is outside domestic fulfillment. Its composite score cannot override that gate. The listed contact counts measure the research frame, not segment profitability or national market size.

E6 supplies fictional $22 and $30 alternatives. HarborMug's $24 item price sits between them, but there is no like-for-like feature or delivered-price evidence. Five Forces assumptions are high rivalry, high substitution (including cups already owned), moderate entry threat, moderate supplier power and high buyer power. This makes replacement motivation and delivered-price acceptance central questions; it does not establish competitive advantage. No real brand or website was inspected.

Survey design and recruitment requirement

Use randomized online invitations without replacement within the two mutually exclusive primary-use strata. The eligible consented frame is 10,000 US contacts: 6,000 commuters and 4,000 desk-first users (E3). These frame proportions are not US population weights. The intended estimands are frame-level proportions and within-group proportions.

At 95% confidence, z = 1.96 and conservative p = 0.5:

• Unadjusted n₀ = z²p(1−p)/e² = 384.16 at e = 0.05; round up to 385.
• Finite-population n = n₀ / [1 + (n₀−1)/N] = 369.984 for N = 10,000; round up to 370.
• At e = 0.08, each group needs 150.0625 before rounding, or 151 completes. The original planner conservatively applies no group-level FPC.

Group	Frame	Proportional allocation of 370	Original planner's group floor	Funded completes	Invitations at 25% usable completion
Commuter	6,000	222	151	222	888
Desk-first	4,000	148	151	151	604
Total	10,000	370	—	373	1,492

Retain the original planner's conservative 373-complete design. Group-specific FPC would reduce the individual floors to 147 and 145, but this is not needed to make the main plan feasible. Desk users receive three additional completes to protect the original floor. Invitation counts are ceil(completes / 0.25); they are expected recruitment requirements, not guaranteed yields. They assume 25% of invited contacts become eligible, usable completed interviews. If the supplied response assumption refers only to survey starts, more invitations will be needed.

The slightly disproportionate allocation requires overall estimates of 0.6 × commuter estimate + 0.4 × desk estimate, rather than an unweighted pooled percentage. Under random-response assumptions, the finite-population stratified variance gives worst-case overall margin ±4.98 percentage points; commuter ±6.46 points and desk ±7.82 points. This satisfies the targets. These are pointwise intervals for proportions, not power guarantees for differences, small cross-tabs or multiple comparisons.

Randomize the invitation order within each frame stratum and release controlled batches. Monitor eligibility, delivery failures, partials and completes separately. Use remaining randomly ordered contacts for top-ups if yield falls short; never exceed the eligible frame or count repeated invitations as new people. Log recruitment dispositions and actual completed-interview yield by stratum. Reconcile changed primary-use classifications before claiming the planned precision.

The finite-population correction and nominal precision apply to this frame under the stated sampling assumptions. Nonresponse may invalidate representativeness even when quotas are achieved. Compare available frame characteristics between respondents and nonrespondents, document coverage and response differences, and disclose weighting and denominators. This consented contact list cannot validate national buyer counts or national 1% adoption by itself. Smaller item-level response counts require recalculated intervals; do not claim the full-sample margin for skipped questions.

Completed questionnaire and analysis specification

The complete customer-facing instrument, invitation copy, routing and pretest instructions are in questionnaire.md, ready for cognitive pretesting. English is used because no different target language was supplied.

Its sequence is consent → US eligibility → exclusive primary-use classification → purchases and usage → existing substitutes and unmet needs → selection priorities → category purchase intent → product concept with full single-unit shipping cost → product intent → likely alternative → reasons. The concept states $24 plus $5 shipping ($29 before tax), care restrictions and delivery timing. It makes no unsupported health or environmental claims.

The primary outputs will be Q9 stated purchase likelihood by group and Q10 preferred alternative, alongside Q3 purchase frequency and Q6 unmet needs. Ask behavior and needs before showing the brand concept to reduce priming. Randomize unordered options, keep ordered scales intact, and preserve neutral and Not sure responses. The instrument does not equate stated intent with realized adoption or attempt to optimize price from a single price point.

Before launch, conduct 6–8 proposed cognitive interviews across both groups and test mobile display, eligibility exits and option randomization. Keep pretest answers out of the main analytic sample. No pretest or survey responses exist yet. After fielding, report group denominators, missing answers, weighted frame estimates, intervals and exclusions. Do not infer that groups differ solely because one ranks higher; test the difference with its own uncertainty.

Boundary checks: smaller frame and conflicting buyer count

The supplied boundary changes the buyer estimate to 1m and the frame to 100 eligible contacts split 90/10. These changes expose decision-relevant failures rather than a new recommended base case.

Layer	Top-down, unchanged	Bottom-up, 1m buyers
TAM	$48m	$24m = 1m × $24
SAM	$19.2m	$9.6m = $24m × 40%
SOM, capped at 8,000 units	$192,000 (8,000 units)	$96,000 (4,000 units)

TAM divergence is 50%, exceeding both the script's 40% tolerance and the decision's 20% tolerance. SAM and SOM also differ by 50%. Reconciliation requires checking scope, annual buyer counts and spend: $48m at $24 implies 2m annual buyers, whereas 1m buyers at $24 implies $24m. Do not average to $36m, silently change assumptions, or claim either view is validated. Hold expansion until evidence resolves which premise is credible.

For the 100-person frame, the original survey planner reports 80 overall completes but recommends 302 after applying two uncorrected 151-person floors. That result exceeds the frame and is infeasible. Group FPC yields floors of 57 out of 90 commuters and 10 out of 10 desk users. Those floors total 67 and do not alone ensure overall ±5%. A feasible complete-response design would be 72 commuters plus all 10 desk users (82 total); the 90/10 weighted worst-case overall margin is about ±4.67 points, with the desk result a census of its frame group.

At 25% yield, inviting the entire 100-person frame produces only 25 expected completes (22.5 commuters and 2.5 desk users). Even the individual FPC floors imply 228 commuter and 40 desk invitations, beyond the available 90 and 10. Do not issue impossible invitation quotas or label the study powered. Seek much higher participation, expand a legitimate eligible frame, or reduce precision and treat the work as exploratory. A full census invitation alone does not make nonrespondents observed.

Assumptions, evidence and decision gates

Evidence	Planning assumption	Confidence and action before expansion
E1	$48m annual US category spend; one 500ml mug per buyer	Low; verify category definition and independently sourced spend
E2	2m annual buyers, $24 spend, 40% serviceable	Low; verify buyer count and frequency; define which needs and channels justify 40%
E3	10,000 exclusive consented contacts, 60/40 split, 25% usable yield	Low; verify eligible frame and classifications; monitor actual response and nonresponse
E4	1% reachable adoption, 8,000 annual supply units	Low; validate acquisition and purchase behavior separately from supply readiness
E5	Five segment scores and known channels	Low; validate distinct needs, reachability and economics; do not treat scores as measurements
E6	Two fictional prices and Five Forces judgments	Low; no independent competitor evidence; assess comparable delivered offers before commitment
Merchant terms	$24 item, $5 single-unit shipping, domestic delivery, hand washing	Supplied assumptions; include them in concept testing because they affect purchase decisions

Alex should authorize only the validation stage once Mia has costed 373 usable completes, 1,492 initial invitations and the separate pretest. No cost-per-complete or research budget ceiling is supplied, so this brief establishes quantities rather than a fabricated spend request.

For the next expansion review, require: independently checked sizing inputs reconciled within the ±20% decision tolerance; both survey group quotas or an explicitly revised precision statement; documented nonresponse and frame limitations; evidence that needs distinguish the two groups; and an acquisition/economic case plus confirmed capacity for the proposed volume. Survey intent is an input to that case, not a conversion forecast. Profitability, acceptable acquisition cost and investment-return thresholds are not supplied, so a numeric expansion go/no-go threshold cannot responsibly be invented.

The immediate decision is therefore to validate commuters first while preserving the desk comparison, withhold international expansion, and defer expansion funding until the demand and economic gaps are resolved.

Supporting deliverables

Market sizing, survey planner, segment ranking, boundary tool results, calculation checks and recomputable calculations preserve the arithmetic. Bundled source files preserve the brief, evidence and three model inputs. Method follows the supplied market-sizing canon, survey-methodology reference (Cochran FPC, total survey error, neutral wording and pretesting), and segmentation/competitive-intelligence reference (Kotler gates and Five Forces); these are methodological references, not external validation of the business inputs.