---
name: competitive-and-market-research
description: >
  Builds a defensible picture of competitors and a market from external sources,
  with an explicit source reliability hierarchy and every unverifiable claim
  labelled. Use for "competitive analysis", "competitor research", "market
  sizing", "how big is this market", "what are our competitors doing",
  "competitive landscape", "is this market size figure real", "who else plays
  in this space", "size the opportunity from secondary sources".
category: 10 Desk Research and Evidence Synthesis
ref: "10.05"
tier: 1
inherits: [K2, K3, K4, K5]
---

# Competitive and Market Research

## 1. One-line description

A method for building a defensible picture of a market and the players in it from external sources, ranking those sources by reliability, separating what a competitor says from what a competitor does, reconstructing published market sizes before using them, and stating plainly the things competitive research cannot establish.

## 2. What this skill is used for

**The research problem it solves.** Competitive and market research is the area of desk research where the evidence base is most heavily populated by parties with an interest in the answer, and where the standard output is most likely to be assembled from material designed to be found rather than material designed to be accurate. Four failures recur. Marketing claims are read as findings, because they arrive in the register of fact. Published market sizes are quoted without anyone establishing what they measure or where they came from, and most widely cited figures collapse, on inspection, to a single unverifiable estimate that four other sources have repeated. Competitors' stated intentions are treated as evidence of what they will do. And the market gets described entirely from the supply side, as a set of vendors and their claims, without anyone establishing what buyers actually buy and what they treat as substitutes. This skill installs a reliability hierarchy applied to every source, a says-versus-does discipline, a mandatory reconstruction step before any published size figure is used, a customer-side view, and an explicit statement of the boundary beyond which competitive research cannot go.

**Where it sits in the research lifecycle.** Usually early, informing a strategic or investment decision, sizing an opportunity, or establishing the landscape a proposition has to compete in. It also runs continuously in organisations that maintain a competitive picture, and immediately before a primary study, where it determines what the study needs to establish that external sources cannot.

**Typical use cases.**
- Sizing a market or a segment from secondary sources where no primary budget exists.
- Building a competitive landscape for an entry, launch or investment decision.
- Assessing whether a widely cited market figure circulating internally is defensible.
- Establishing what competitors have actually done, as distinct from what they have announced.
- Understanding the real substitutes for a proposition, which are frequently not the named competitors.
- Providing the external context section of a strategy document or a business case.

**Who uses it.** Strategy and corporate development teams; insight and market intelligence functions; consultants building a landscape or a sizing; product and commercial leads assessing a category; investment and diligence teams working to a short timeline.

## 3. When to use it

- A decision rests on how big a market is, or how fast it is growing, and the number will come from external sources.
- A figure is circulating internally and nobody can say where it came from.
- An entry, launch or investment decision needs a defensible view of who else is in the market.
- Someone needs to know what a competitor has actually done rather than what they have said.
- The competitive set is assumed and has never been tested against how buyers actually see it.
- A business case requires external context that will be scrutinised.
- A competitor has made a claim about their own performance and it is influencing internal decisions.
- You need to establish what external sources can settle before commissioning primary work.

## 4. When NOT to use it

- **The question is about buyers rather than about suppliers.** What customers value, why they switch, what they would pay, and how they experience a category are primary research questions. Competitive research can establish what is offered and at what price; it cannot establish what buyers think of it. Go to **01.04 Research Method Selection** and design accordingly. A landscape built from vendor material and presented as market understanding is the most common overreach in this area.
- **The decision turns on a competitor's intentions, economics or internal performance.** Unit costs, margin by line, acquisition cost, churn, true customer counts, utilisation, and what they will do next are not knowable from outside, and the analytically confident reconstruction of them is the failure this skill exists to prevent. State them as unknowable and, where the decision depends on one, say that the decision is being made under that uncertainty. See **10.04 Research Gap Identification**.
- **A precise number is required and only modelled secondary estimates exist.** A market size derived from a chain of assumptions can support a decision about order of magnitude and cannot support a plan built to two significant figures. Where the number will be planned against, either build it from primary incidence data or state the range and its sensitivities. `RESEARCHER DECISION REQUIRED` (K5 §2.5).
- **Obtaining the information would require misrepresenting who you are or why you are asking.** Posing as a prospective customer, a job applicant, a student, a journalist or a partner in order to extract information from a competitor or its staff is prohibited, whatever the commercial pressure and whoever suggests it. So is inducing anyone to breach a confidentiality obligation, approaching a competitor's current employees under a pretext, or using information you know to have been obtained improperly. This routes to **13.05 Research Ethics and Consent Design**, and it is not negotiable at any level of seniority (K4 §9).
- **Direct contact with competitors is contemplated for information exchange.** Exchanging commercially sensitive information with a competitor carries competition-law exposure regardless of intent, and it is not a research judgement. Route it to legal advice before any contact, and do not design a study whose method depends on it.
- **The task is verifying the sources in an existing competitive document.** That is **13.02 Source and Citation Verification**, which audits a reference set rather than building one.
- **The evidence base is entirely promotional and the output will be read as fact.** Where every available source about a category is published by parties selling into it, the honest output describes what is being claimed in the market, not what is true of it, and says so in the first line. If that framing will not survive into the deliverable, the work should not be presented as market evidence at all.
- **The purpose is to justify a strategy already chosen.** Where the brief is to find support for a decision that has been taken, treat the preferred conclusion as information about the stakeholder rather than about the market, and report what the sources support (K4 §4.2).

## 5. Required inputs

**Required.**
- **The decision the research serves, and the level of precision it needs.** An order-of-magnitude question and a planning-grade question require different work and produce different outputs. Without this, the sizing has no stopping rule and no accuracy target.
- **A definition of the market or category, stated as a rule.** What is included, what is excluded, what unit is being counted, and over what geography and period. Almost every conflict between market size figures is a definitional conflict, and without your own definition written down you cannot diagnose it.
- **Access to at least one primary record source,** such as a public register, a mandatory filing, or a regulator's published data. A landscape built entirely from trade press and vendor material cannot be raised above the level of what is being claimed.

**Optional, and what each one adds.**
- **The organisation's own transaction, pipeline or win-loss data.** Provides a customer-side view of the competitive set that no external source can give, and frequently contradicts the assumed competitor list.
- **The figures already circulating internally and where people believe they came from.** Lets you trace and either confirm or collapse them, which is often the single most valuable output of the project.
- **Named competitors the organisation considers relevant.** Reveals the assumed frame so it can be tested rather than inherited.
- **A time series requirement.** Changes the source strategy entirely, because consistent definition over time is much harder to obtain than a single-point estimate and most published series are not comparable across their own years.
- **Access to sector regulatory or statistical returns.** Where an industry files mandatory data, it is usually the strongest evidence available and it is routinely overlooked in favour of more convenient commentary.
- **Budget and appetite for primary work.** Determines whether the customer-side gap is closed or merely reported.

## 6. Questions to ask before starting

1. **What decision does this serve, and to what precision?** Determines whether a range is acceptable, which determines almost everything else. Default: assume order of magnitude, and mark clearly wherever a figure is being asked to carry planning weight.
2. **Whose definition of the market is being used, and is it the buyer's or a supplier's?** Categories drawn by suppliers rarely match those in buyers' heads, and a size figure inherits the definition of whoever built it. Default: write your own definition first and map every source's definition to it.
3. **Which figures are already load-bearing internally?** These will be defended, and contradicting one requires the derivation, not just a better number. Default: identify them at the outset and trace them first.
4. **Is the competitive set assumed or established?** An inherited competitor list usually reflects who the organisation watches rather than who it loses to. Default: test the set against any customer-side data available, and state that it is assumed where none is.
5. **Does anything in the intended collection require presenting yourself as something you are not?** Asked before collection begins, this prevents a method that cannot be used. Default: no covert collection, and any contact identifies who is asking and why.
6. **Is the sector regulated in a way that generates mandatory public data?** Changes the achievable quality of the answer dramatically. Default: check for mandatory reporting before building a search around commentary.
7. **What would make this landscape out of date?** Competitive pictures decay faster than most research. Default: state a shelf life and name the events that would invalidate it.

## 7. Step-by-step methodology

**Step 1. Frame the market from the buyer's side before naming a single competitor.**
Write the market definition as an operational rule: what need is being met, for whom, what counts as being in the category, what is excluded, in what geography, over what period, and in what unit (revenue, volume, accounts, users, transactions). Then define the competitive set as everything a buyer would consider a substitute for meeting that need, not as the set of companies that describe themselves the way you do. This routinely produces a different list, and the difference is where competitive research earns its keep: the substitutes that never appear in a supplier-drawn landscape are typically doing nothing, doing it internally, using an adjacent product designed for something else, and the incumbent arrangement that nobody thinks of as a product. Where any customer-side evidence exists, test the set against it. Where none exists, state that the set is supplier-defined and treat that as a limitation of the whole exercise. *Correct result: a written market definition and a competitive set including non-obvious substitutes, with its basis stated as buyer-derived or supplier-derived.*

**Step 2. Set the collection boundary before collecting anything.**
Decide and record what methods are in scope, because this is far harder to do once collection is under way and a promising route appears. In scope: published and filed material, public registers, regulator data, publicly advertised prices and terms, product documentation available without pretence, public job advertisements, public statements, and anything a competitor has chosen to make public. Out of scope, absolutely: any approach in which you misrepresent your identity, your employer or your purpose; any attempt to obtain information from a competitor's staff under a pretext; anything that induces or relies on a breach of confidence; and any use of material you know to have been improperly obtained. Two boundaries need specific care. Where a genuine purchase or account opening is used to observe a product, it must be a genuine transaction, it must not involve stating a false identity or false eligibility, and in regulated sectors it may itself be regulated, so it goes to review before it happens. Where customers or ex-employees of a competitor are interviewed, the interview must not solicit confidential information, and the boundary needs stating to the participant, not just held by the interviewer. Route anything uncertain to **13.05 Research Ethics and Consent Design** before proceeding rather than afterwards. *Correct result: a written collection boundary agreed before collection, with any borderline method reviewed rather than attempted.*

**Step 3. Map sources onto an explicit reliability hierarchy.**
Competitive sources vary in reliability more than in any other kind of desk research, and the variation does not correlate with how easy they are to find. Rank every source before extracting anything from it. **Tier 1, primary record**: audited financial statements, mandatory regulatory filings, public registers, regulator-published data, court and tribunal records, patent and trademark filings, and official statistics. These are subject to external obligation and are the only sources that can carry a claim on their own. **Tier 2, direct observation you conducted**: publicly advertised prices captured on a date, product documentation, published terms and conditions, public job advertisements, and the observable footprint of a business. Reliable for what was observed, on that date, and nothing more. **Tier 3, unaudited company communication**: investor presentations, press releases, annual reviews without audit, executive interviews. Accurate on facts the company would be embarrassed to get wrong, selective on everything else. **Tier 4, independent third-party commentary**: analyst notes, sector reports, and trade press reporting where the reporting names its own source. Usable where its own source can be traced. **Tier 5, marketing and promotional material**: brochures, category pages, sponsored content, award claims, customer logos. Evidence of positioning, never of performance. **Tier 6, unverifiable**: anonymous forums, aggregator figures with no method, "industry sources suggest", and any figure whose origin cannot be traced. These do not carry claims at all, and appear in the output only as labelled unverifiable material where their existence is itself the point. *Correct result: every source tiered before extraction, and the tier profile of the evidence base visible at a glance.*

**Step 4. Separate what a competitor says from what a competitor does.**
Build the evidence base for each competitor in two columns, and never let them merge. **Says**: stated strategy, announced roadmap, positioning, claimed differentiators, stated ambitions, and performance claims about themselves. **Does**: prices actually charged and when they changed, products actually available, markets actually served, filings actually made, capital actually raised or spent, roles actually being recruited and where, facilities actually opened or closed, partnerships with observable substance rather than a joint announcement, and terms actually offered. The doing column is evidence; the saying column is evidence only of what they wish to be understood to be doing, which is a genuinely useful thing to know and a different thing. Two signals in the doing column are consistently underused and consistently informative: recruitment, which reveals capability being built six to twelve months before it appears in a product, and pricing and terms changes, which reveal commercial pressure that no announcement will mention. Where says and does diverge, the divergence is a finding and belongs in the output, worded as an observation about the gap rather than as an accusation. *Correct result: a two-column evidence base per competitor, with every says-does divergence noted.*

**Step 5. Trace every claim to its origin, and collapse citation chains.**
The characteristic structure of market information is a small number of originals and a large number of repetitions, arranged so that repetition reads as corroboration. Four documents citing a figure, each citing another, all terminating in one estimate, is one source with a distribution network. For every figure you intend to use, follow the citation backwards until you reach either a primary record, a stated method, or a dead end, and record which of the three you reached. Where the chain terminates in a dead end, the figure is unverifiable and is labelled as such wherever it appears, including in a summary and including on a chart. Where the chain terminates in a single original, the output says so explicitly: not "widely reported", which is true and misleading, but "traces to a single estimate published by X in year Y, method not disclosed", which is true and useful. Expect this step to be the one that changes the project most, and budget time for it accordingly. *Correct result: a traced origin recorded for every figure used, and every multi-source figure collapsed to its actual number of independent origins.*

**Step 6. Reconstruct any published market size before using it.**
Never use a size figure you cannot derive. Establish seven things: the universe (who or what is being counted), the unit (revenue, volume, subscriptions, users, and on what recognition basis), the inclusions and exclusions, the geography, the base year and whether the figure is actual or forecast, the method (measured from records, modelled from incidence and average value, or estimated from a small number of interviews), and the assumptions the model rests on. Then map that definition to yours and record the mismatch. Two figures differing by a factor of three usually differ by definition, not by accuracy, and reconciling them is often possible once both are reconstructed. Where a figure cannot be reconstructed because the method is not disclosed, it does not enter the analysis as a number. It may be reported as a figure in circulation, labelled, which is sometimes exactly what the client needs to know. *Correct result: a reconstruction note per size figure covering all seven attributes, and an explicit statement of which figures could not be reconstructed.*

**Step 7. Build the estimate from two directions and expose its sensitivities.**
Where a size is needed, build it twice. **Top-down**: from a total that is independently established, usually official statistics or regulatory returns, narrowed by stated proportions to your definition. **Bottom-up**: from a count of units multiplied by an average value, where both come from separate sources. The two will not agree, and the gap is the useful output: a bottom-up figure far below top-down usually means the unit count is incomplete, and far above usually means the average value came from the visible large players. State the range rather than a point, and identify the two or three assumptions the answer is most sensitive to by varying each across its plausible span and reporting the effect. This is not a statistical procedure and must not be dressed as one; it is a disclosure of which assumptions carry the number. Round to the precision the method supports, which for a modelled market size is usually one or two significant figures (K3 §4.4). *Correct result: two independently built estimates, a stated range, and a named list of the assumptions that move the answer most.*

**Step 8. Assess the market from the customer side.**
A supply-side landscape counts vendors and repeats their claims. A customer-side view establishes what buyers actually buy, from whom, alongside what, and what they treat as the alternative. Sources that give this without primary fieldwork are limited but real: the organisation's own win-loss and pipeline records, published procurement and tender awards where they exist, regulatory data on switching or account movement, published customer counts with defined units, and reviews and complaints in aggregate where the platform's own selection effects are stated. Establish four things: who actually buys, what share of their requirement any one supplier holds, what they switch to and from, and what "do nothing" or an internal solution represents in the category. The last is the most frequently omitted and often the largest competitor in the market. Where none of this is available, say so, and name the customer-side view as the principal gap rather than filling it with supplier-side inference. *Correct result: a customer-side statement covering buyers, share of requirement, switching and non-consumption, or an explicit statement that the assessment is supply-side only.*

**Step 9. Read every competitor claim against the incentive behind it.**
Performance claims about oneself are made for a reason, and the reason shapes the claim. Establish, for each claim, what the company was doing when it made it: raising capital, recruiting, selling, positioning for an exit, defending a position, or managing a regulator. Then apply the standard tests. A growth rate without a base is not information. "Fastest growing" is a claim about a chosen comparison set and a chosen period. A user or customer count needs its definition: registered, active, paying, and on what recency. Revenue claims need the recognition basis and whether they are annualised from a partial period. Awards and rankings need to be checked for whether entry was paid and who judged. Customer logos indicate a relationship of unknown size and duration. None of this makes the claims worthless; it makes them claims, which are reported as claims, attributed, and separated from what has been established. Where a claim is material to the decision and cannot be verified, that is a finding and belongs beside the claim, not in a limitations section. *Correct result: every competitor performance claim carrying its incentive context and its definitional gap, and none of them promoted to established fact.*

**Step 10. State the boundary, then write up with every claim sourced and every unverifiable claim labelled.**
Write explicitly what this research cannot establish about each competitor: internal economics, margin, acquisition cost, churn, true scale, the materiality of announced partnerships, the reason for any decision, and above all what they intend to do next. Naming these prevents the reader from filling them in, which they will otherwise do from the confident tone of the rest of the document. Then write the output so that each claim carries its source reference per K2 §4.3, its date, and one of four statuses: **established** (tier 1 or 2, verifiable), **reported** (tier 3 or 4, origin traced), **claimed** (the competitor's own assertion, attributed and unverified), or **unverifiable** (origin untraceable, present only where its circulation is itself the point). Confidence per K3 is capped by the tier of the source carrying the claim, and no conclusion rests on tier 5 or 6 material. *Correct result: an output in which a reader can tell, for every line, whether it is a record, an observation, a report, a claim or a rumour.* `RESEARCHER SIGN-OFF REQUIRED` where the work supports an investment or entry decision (K5 §2.5).

## 8. Analytical framework

Two structures.

**The claim chain, applied to everything:**

    Claim → Source → Reliability tier → Says or does → Origin traced →
    Incentive read → Independent corroboration → Status → Confidence

The chain is what separates a competitive document that can be defended in a board meeting from one that cannot. Two arrows carry most of the risk. *Origin traced* is where citation chains are collapsed or preserved, and preserving one converts a single estimate into an apparent consensus that will then be planned against. *Says or does* is where positioning is either separated from behaviour or allowed to enter the analysis in the register of fact, and the failure is invisible because marketing material is written to read like findings.

**The four claim statuses,** which every line in the output carries:

    ESTABLISHED    Primary record or your own dated observation. Carries claims.
    REPORTED       Third party with a traceable source. Carries claims with
                   the source named.
    CLAIMED        The competitor's own assertion. Reported as an assertion,
                   attributed, never as a finding.
    UNVERIFIABLE   Origin cannot be traced. Appears only where the fact of its
                   circulation is the point, and always labelled.

Applying it: the status is a property of the evidence, not of how confident the writer feels, and it does not improve with repetition. A claimed figure repeated by three trade outlets is still claimed. The single most useful discipline in this whole skill is refusing to let status drift upward between the working file and the summary slide, which is where it almost always happens.

## 9. Output format

**1. Question, decision and market definition.** The decision served, the market defined as an operational rule, the geography, the period, the unit, and whether the definition is buyer-derived or supplier-derived.

**2. Method and collection boundary.** What sources were used, what was not accessible, and the ethical boundary applied, stated positively rather than as a disclaimer.

**3. Source table.**

| ID | Source | Type | Reliability tier | Date of data | What it establishes | Origin traced to | Interest to declare |
|---|---|---|---|---|---|---|---|

**4. Market size and structure.** The top-down and bottom-up estimates, the range, the assumptions the answer is sensitive to, and the reconstruction note for every published figure used or rejected.

**5. Competitive set.** Who is in it and why, including non-obvious substitutes and non-consumption, with the basis of the set stated.

**6. Per competitor: says and does.** Two columns, with divergences noted, and every performance claim carrying its status and its incentive context.

**7. Customer-side view.** Who buys, share of requirement, switching, and the alternative to buying at all. Or an explicit statement that the assessment is supply-side only.

**8. What this research cannot establish.** Named per competitor and for the market as a whole.

**9. Claims register.** Every material claim with its source, date and status.

**10. Shelf life.** When this picture expires and what would invalidate it sooner.

**When the evidence is thin, the format must not force fabrication (K4 §1).** A competitor about whom only promotional material exists gets a short entry saying exactly that, with their claims attributed and unverified. The comparison table is not filled in by inference: an unknown price is `[not available]`, never an estimate presented as a figure, and never a plausible number derived from a competitor's positioning. Where a market size cannot be built, the output says what is known about the shape of the market and states that the size cannot be established from available sources, which is a more useful answer than a modelled figure whose assumptions carry it. A landscape covering four competitors properly beats one covering twelve by pattern-matching.

## 10. Quality checks

Run before anything is presented. These sit on top of K4 §8.

1. Is the market defined as an operational rule, and is every source's definition mapped to it?
2. Was the competitive set derived from what buyers substitute, or inherited from who the organisation watches?
3. Does the set include doing nothing, doing it internally, and adjacent products used for the purpose?
4. Is every source tiered, and does any conclusion rest on tier 5 or 6 material?
5. Has every figure been traced to its origin, and is every citation chain collapsed to its true number of independent sources?
6. Has every market size figure used been reconstructed on all seven attributes, and are the unreconstructable ones excluded from the analysis rather than footnoted?
7. Is the size given as a range with its sensitive assumptions named, rather than as a point?
8. Is any modelled figure carrying more significant figures than its method supports?
9. Is every competitor performance claim labelled as a claim, with its incentive context, and does any appear in a summary as established?
10. Are says and does separated throughout, including in the executive summary?
11. Is every observation dated, given that prices, terms and product availability change?
12. Does the output state what cannot be established about competitor economics, performance and intentions?
13. Was any information obtained by any means involving misrepresentation of identity or purpose?
14. Does every claim in the output carry a status, and has any status drifted upward between the working file and the summary?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Marketing read as finding** | A differentiator claim stated in the register of fact | Two-column says and does, status labels (Steps 4, 10) |
| **The consensus that is one source** | "Widely reported to be worth X" | Trace and collapse every chain (Step 5) |
| **Unreconstructed market size** | A figure used with no account of what it counts | Seven-attribute reconstruction before use (Step 6) |
| **Definitional mismatch** | Two sizes differing by a factor of three, treated as a data quality problem | Map every source definition to your own rule (Step 1) |
| **False precision in a model** | A modelled size quoted to three significant figures | Round to what the method supports (K3 §4.4) |
| **Supply-side-only landscape** | A market described entirely as a list of vendors | Customer-side assessment as a required section (Step 8) |
| **Missing the real competitor** | Non-consumption and internal alternatives absent from the set | Define the set by substitution, not by self-description (Step 1) |
| **Growth rate without a base** | "Grew 300% last year" carried into a summary | Incentive read and definitional test on every claim (Step 9) |
| **Undefined user counts** | "Two million users" with no definition of user | Require the definition or report the claim as undefined |
| **Intentions inferred** | A confident account of what a competitor will do next | State intentions as unknowable (Step 10) |
| **Economics reconstructed** | A margin or acquisition cost estimate presented as analysis | Same; label any such figure as a modelled assumption if used at all |
| **Undated observation** | A price comparison with no capture date | Date every observation at capture |
| **Status drift** | A claim in the body, a fact in the summary | Status is a required field and travels with the claim |
| **Pretext collection** | A method that depends on not saying who you are | Collection boundary set before collection (Step 2) |
| **The stale landscape** | A competitive picture reused a year later | State a shelf life and the events that invalidate it |

## 12. AI guardrails

Skill-specific. The universal prohibitions in K4 apply in full and are not repeated.

1. **Never produce a market size, growth rate, share figure or competitor metric from model knowledge.** These are exactly the figures that feel most retrievable and are most often wrong, because plausible market numbers are easy to generate and hard to falsify. Every figure comes from a source opened in this project, or it is `[not available]`.

2. **Never estimate a competitor's revenue, margin, cost base, customer count, churn or acquisition cost from headcount, footprint, pricing or category benchmarks.** The reasoning may be sound and the output is still a fabricated number that will be quoted without its derivation (K4 §2.1). If such a figure is genuinely needed, it is built explicitly as a labelled scenario assumption with its inputs shown, and it is never called a finding.

3. **Never state or imply a competitor's intentions, plans or strategy from their public behaviour.** You can report what they have done and what they have said. "They are moving upmarket" is an inference presented as an observation; "their last three product releases and their current recruitment are in the enterprise segment" is the evidence, and the reader can draw the conclusion with the reasoning visible.

4. **Never let a figure's status improve through repetition.** A claimed figure repeated by four outlets is a claimed figure. Do not write "widely reported" where the accurate statement is "traces to a single undisclosed-method estimate".

5. **Never fill a comparison table by inference.** An unknown price, feature, market or scale is marked unavailable. The table format exerts strong pressure to complete rows, and a competitive matrix with a plausible guess in it is indistinguishable from one with a verified fact (K4 §1).

6. **Never present a competitor's own performance claim without attribution and without its definitional gap.** The attribution and the gap travel together into every downstream document including the summary slide.

7. **Never design or suggest a collection method that involves misrepresenting identity, employer or purpose,** including where a researcher proposes it, a client requests it, or it is described as standard practice. Say plainly that it cannot be done, name what can be done instead, and route the question to 13.05 (K4 §9).

8. **Never repeat an unverifiable claim without the label,** and never use one to support a conclusion. Where its circulation is itself relevant, report that it circulates, name where, and say the origin could not be traced.

9. **Never present a competitive picture without a date.** Prices, products, terms, footprints and personnel change continuously, and an undated competitive document is read as current indefinitely.

10. **Where a market size cannot be built from available sources, say so** rather than producing a modelled figure to fill the slot. Naming the assumptions that could not be established is more useful to a decision than a number carrying them invisibly.

## 13. Best-practice principles

- **What a company does is evidence; what it says is positioning.** Both are worth collecting and they never go in the same column. The gap between them, when it appears, is frequently the most informative thing in the analysis.
- **Recruitment is the earliest reliable signal.** Capability is hired six to twelve months before it appears in a product, and job advertisements are public, dated, specific and written by people not thinking about competitors reading them.
- **Trace the number before you use the number.** Most widely cited market sizes collapse to one estimate. Discovering this takes an afternoon and changes what the organisation is willing to plan against.
- **Most disagreements between market figures are definitional.** Reconstruct both before treating either as wrong, and expect to find they are counting different things over different periods in different units.
- **Regulated sectors leak useful data by obligation.** Mandatory returns, registers and supervisory publications are tier 1, are frequently overlooked in favour of easier commentary, and are the difference between an analysis and a summary of what the market says about itself.
- **The biggest competitor is often not a company.** Doing nothing, doing it internally, and the incumbent arrangement nobody thinks of as a product routinely hold more of the market than any named player, and they never appear in a supplier-drawn landscape.
- **Read every self-reported claim against what the company was doing when it made it.** A figure released during a funding round, a recruitment push or a sale process is a figure with a job to do.
- **Date every observation at capture.** Competitive material has a short half-life, and an undated price comparison becomes wrong silently rather than visibly.
- **Say what you cannot know, in the document.** Competitor economics, internal performance and intentions are outside the reach of external research, and a reader who is not told this will assume the confident tone extends to them.
- **A short landscape done properly outperforms a long one done by pattern.** Four competitors with traced, dated, status-labelled evidence is a usable asset; twelve filled in by inference is a liability that looks like thoroughness.
- **Never let commercial pressure move the ethical line.** The information obtainable only by misrepresentation is not worth the exposure, and the organisations that lose most from this are the ones whose researchers assumed everyone does it.
- **Give the picture an expiry date.** A competitive document with no shelf life will be quoted for years, long after the market it describes has moved.

## 14. Worked example

Generic fictional scenario, financial services.

**INPUT**

A mid-sized business lender is deciding whether to enter invoice finance for small firms in its home market. The strategy team has a figure in its deck: the market is worth a stated amount and is growing at a stated rate. They want the landscape, the size confirmed, and a view of the four competitors they have named.

**PROCESS**

*Step 1.* The market is defined as a rule: finance advanced against unpaid trade invoices to firms below a stated turnover threshold, in the home market, measured as advances outstanding at period end. Written down, this immediately excludes two of the four named competitors, whose products are supply-chain finance arrangements initiated by the buyer rather than the supplier. They remain relevant as adjacent players and are moved out of the direct set. The set gains two entrants the team had not listed: a payments provider offering advances as a feature, and the buyer's own practice of paying late, which is the incumbent arrangement most small firms actually use.

*Step 2.* Collection boundary agreed and written down. Published terms, public filings, the sector regulator's public register, and job advertisements are in scope. A proposal to have an analyst apply for a facility as a small business is rejected: it would require stating a false purpose and false eligibility, and in a regulated sector it carries additional exposure. It is routed to 13.05 and does not proceed.

*Step 3 and 5.* Sources are tiered, and the deck's headline market figure is traced. Four documents cite it. Two cite a trade association bulletin, one cites a consultancy report, and the consultancy report cites the same bulletin. The bulletin reports aggregate advances from its own member firms. The chain collapses to one source, and that source is a member survey covering an unknown share of the market with no non-member coverage.

*The judgement call.* The figure is not wrong and it is not what the deck thinks it is: it is total member advances, not market size, and it excludes non-member lenders entirely. The temptation is to gross it up by an assumed member share, which would produce a defensible-looking number. That is refused, because the assumed share would carry the entire estimate and would arrive in the deck without its assumption. Instead the figure is reported as what it is, and the sizing is rebuilt.

*Step 6 and 7.* Top-down: the regulator's public register carries mandatory returns for authorised firms, which covers a defined portion of the market with a stated definition. Bottom-up: firm counts by turnover band from official statistics, multiplied by penetration and average facility size drawn from two separate sources. The two produce a range of roughly two to one. Sensitivity testing shows the answer is carried almost entirely by the penetration assumption, which is the weakest input. This is stated in the output as the single number the estimate depends on, and it becomes a candidate for primary work.

*Step 4 and 9.* Says and does. One competitor's public statements emphasise a move into larger facilities. Their recruitment over eighteen months is concentrated in collections and recovery, and their published terms have tightened twice. The divergence is reported as observed, without inferring a strategy from it. Another competitor claims to be "the fastest growing" in the category; the claim traces to a press release, the comparison set is unstated and the period is a partial year annualised. It is reported as a claim, attributed, with both gaps named.

*Step 8.* Customer side. The lender's own win-loss records show that the alternative chosen in the majority of lost cases is not a competitor at all: it is an extension of the firm's existing overdraft, or simply paying suppliers later. That reframes the entry question from share capture to category creation, which is a materially different proposition.

**OUTPUT**

A landscape reporting a market size as a range with the penetration assumption named as the thing that carries it, the deck's original figure retained but correctly described as member advances rather than market size, a competitive set including two substitutes the team had not considered and excluding two players they had, a per-competitor says-and-does record with claims labelled, a customer-side finding that the primary alternative is non-consumption, and an explicit statement that competitor unit economics, loss rates and intentions cannot be established from external sources. The recommendation is that the entry decision turns on the penetration assumption and on demand among firms currently using overdraft or late payment, neither of which desk research can settle, and that a small primary study on incidence and substitution would close both.

`RESEARCHER SIGN-OFF REQUIRED` on the sizing and the entry recommendation (K5 §2.5). The rejected collection method is recorded in the method section, because a decision not to use a method is part of the method.

## 15. Advanced usage

**The maintained competitive picture.** Where a market is watched continuously, convert the one-off landscape into a monitored asset: a fixed source list, a fixed capture cadence for prices and terms, and a standing record of recruitment and filings. The value is entirely in the diff. A price change captured in the week it happens is a signal; the same change discovered nine months later is history. Pair with **14.03 Research Repository and Knowledge Curation**.

**Triangulating a competitor's scale without inventing it.** Several independent public traces bound a competitor's size without any of them establishing it: headcount from public professional profiles, footprint from public premises records, regulatory returns where they file, and volume disclosures made for other purposes. Reported as a bounded range with each trace named, this is legitimate and useful. Reported as an estimate, it is fabrication with extra steps. The distinction is entirely in whether the traces are shown.

**Using the landscape as a primary research specification.** The customer-side gaps this method exposes are precisely what a primary study should measure, and they are usually cheaper to close than the supply-side questions everyone asks about. Hand the gap list to **10.04 Research Gap Identification**, then to **01.04 Research Method Selection**.

**Where the standard approach does not fit.** In markets dominated by private companies with no filing obligation, tier 1 evidence is largely unavailable and the honest output is a landscape built on tier 2 observation and labelled claims, with the size stated as unestablishable. In emerging categories, the definitional work in Step 1 is most of the value, because the category has no agreed boundary and every published figure counts something different. In highly consolidated markets, competition-law sensitivity rises sharply and any collection method involving contact with competitors should be reviewed before it is designed, not after.

## 16. Skill chain

**Recommended previous skills:**
- **10.01 Literature Review and Desk Research.** Supplies the search strategy, the verification gate and the appraisal framework this skill applies to company and market sources.
- **01.02 Business Problem to Research Question.** Supplies the decision and the precision the sizing has to reach.

**Recommended next skills:**
- **10.04 Research Gap Identification.** Takes the customer-side gaps and the assumptions carrying the size estimate and turns them into a research specification.
- **10.03 Multi-Source Research Synthesis.** Integrates the external picture with the organisation's own pipeline, win-loss and transaction data.
- **01.04 Research Method Selection.** Designs the primary work that closes what external sources cannot.

**Runs well alongside:**
- **13.05 Research Ethics and Consent Design,** which governs the collection boundary and any borderline method.
- **13.02 Source and Citation Verification,** which audits the traced-origin work.
- **12.03 Research Report Compilation,** which carries the claims register and its status labels into the deliverable.

---
A Yazi Supplied Skill and resource.