Research methodology
How Douglas Insights Builds Its Numbers
Every Douglas Insights market figure is built from the demand side, tied to official data wherever it exists, informed by primary interviews, and checked against independent views of the market from several methods before it is published. Each report states how well its number is checked.
Principles
Seven rules every study follows
Build from demand, check from supply
Markets are sized as volume multiplied by price, country by country and segment by segment. Company revenue, trade and production data then test the total. Checks validate the build; they never replace it.
Official data before assumptions
National dispensing and prescription data, reimbursement and claims records, health and household surveys, customs and production statistics and official price lists come first. An assumption is used only where no such data exists, and every assumption is named in the model.
Primary evidence fills the gaps
Interviews with manufacturers, distributors, buyers, specialists and regulators set and test the assumptions that public data alone cannot: real transaction prices, share estimates, capacity and adoption behaviour.
Nothing typed, nothing invented
Every number in a model is a sourced fact, a named assumption or a value derived from the two. Source files are archived, and figures taken from data files keep a reference to the exact table and cell they came from.
Several independent views must agree
Before publication each model is checked against at least three independent views of the market from at least two different methods. Disagreement is investigated, not averaged. A gap that remains is disclosed.
The whole market, organised and unorganised
Where unbranded, locally made or informally sold products form part of a market, that share is measured from documented sources and shown separately. Where regulation rules it out, the report says so.
Revised in the open
Reports carry an edition, a change log and a next review date. Corrections are published on the corrections page.
The pipeline
Nine stages from scope to publication
-
1
Scope and definitions
Product boundaries, segments, channels, end uses, geographies and units are fixed in writing, including what counts as the organised and the unorganised market.
Output: scope note and segmentation tree -
2
Source sweep
A standard checklist is worked for every market: dispensing and point-of-sale data, surveys and registries, official prices, company disclosures, trade and production statistics, claims and reimbursement data, structure evidence, related markets and published estimates. A missing source family is recorded, not skipped.
Output: source register -
3
Anchors
Independent totals from official or audited data, for example national dispensing value, claims spend, or apparent consumption from production and trade, are set before the model is built.
Output: anchor set from at least two methods -
4
Primary interviews
Structured interviews with manufacturers, distributors, buyers and specialists fill what public data cannot: transaction prices, share estimates, capacity and adoption behaviour. Findings test the assumptions in the model.
Output: interview log with coded findings -
5
Model build
Volume and price are modelled by country, product and year from the sourced drivers. Countries sum to regions and regions to the global total by construction.
Output: working model with every input traced -
6
Company evidence
Reported company revenue in the market forms a floor; smaller suppliers form a documented long tail; the unorganised share forms its own line. Private company figures are estimated by two independent routes.
Output: supply-side build -
7
Validation and triangulation
The demand build is compared with the supply build, the anchors and published estimates. Each check is scored as agreeing, near, weak or in conflict.
Output: validation table -
8
Quality gates and approval
Automated gates test sums, overlaps between segments, sourcing, anchors, exchange rates, validation coverage, market structure, consistency with related markets and the integrity of the data workbook. A model that fails a gate does not publish. Every version is approved before release.
Output: gate log and approval -
9
Publication and maintenance
The report, its fact sheet and the licensed data pack are generated from the same approved model version, so every published number matches.
Output: report, fact sheet, data pack
Data sources
What feeds a study
Manufacturers, distributors, buyers, clinicians, engineers and regulators, selected per market.
National prescription and pharmacy dispensing statistics, reimbursement datasets and retail panels where public.
National health and household surveys, disease and patient registries, programme procurement records.
Tariffs, reimbursement lists, price ceilings, reference price guides and tender awards.
Annual reports, segment reporting, investor presentations and regulatory filings.
Customs flows by HS code, national production statistics and association output data.
Public insurer and government payer datasets.
Regulation documents and informal-sector statistics.
Parent, adjacent, upstream and downstream markets in the Douglas Insights model library.
Third-party market totals, used only as a low-weight check, never as a build input.
Typical evidence mix
Weight of each input
- Primary interviews · 40%
- Company filings · 20%
- Customs and trade data · 15%
- Associations and registries · 15%
- Price evidence · 10%
The mix varies by market. Thinly documented markets lean harder on interviews; regulated markets lean on dispensing, claims, registry data and filings; industrial markets lean on production, trade and company data.
Model construction
Anatomy of a Douglas Insights market model
Segments sum to the total
Products, channels, end uses and structure splits each add up to the market total. If they do not, the model does not publish.
Structure is explicit
Every region carries an organised and unorganised split, each with a documented source or a documented reason it is zero.
Cross-checked before it closes
- Demand build versus reported company revenue
- Model total versus official anchors, including dispensing, claims, production and trade
- Implied volumes versus capacity and trade flows
- Model prices versus official and transaction prices
- Assumptions versus primary interview findings
Forecasting
Growth is decomposed, never asserted
Forecasts run ten years from the base year and are driven by named drivers. Each report separates volume growth from price movement and publishes base, slower and faster scenarios, so readers can see where the growth comes from and test their own view in the data pack.
Validation
Checked from several independent directions
A model closes only when independent views of the market agree within tolerance. Each report is tested against at least three checks from at least two of these methods.
Each check is scored: inside the range agrees; slightly outside is near; further outside is weak; far outside is a conflict. A conflict blocks publication until it is explained and approved, and it is disclosed. Licensed buyers receive the validation table with the data pack.
The check sits inside the tolerance range for the market.
Slightly outside the range, noted in the validation table.
Further outside the range, investigated and explained.
Far outside the range. Publication is blocked until it is resolved or disclosed and approved.
The methods
Four independent readings
Dispensing, claims, surveys and patient data.
Reported company revenue, capacity and producer output.
Customs flows and apparent consumption.
Third-party totals, carrying the lowest weight of the four.
Market structure
Counting the market companies do not report
In many markets part of the demand is met by unbranded, locally made or informally sold products that no company reports. Douglas Insights measures that share from documented sources, such as informal-sector statistics and regulatory evidence, and shows it as its own line in the market and in the company tables.
Company coverage is measured against the organised market only, so a fragmented market is neither inflated nor penalised for revenue no company could report.
Confidence
Every report tells you how much to trust it
Report confidence label
Most of the market value is backed by reported company revenue or official anchors, and checks from at least three methods agree.
A substantial part of the value is backed by sourced evidence, and checks from at least two methods largely agree.
Evidence covers a smaller part of the value, or one check disagrees and the gap is explained.
Any unresolved disagreement caps the label at Low. The label reflects the strength of the evidence, not the importance of the market.
Input grades
Corroborated by two or more independent sources, such as official statistics, national datasets, audited filings or primary interviews.
A reliable documented source, or a private company estimate confirmed by two independent routes.
A single estimate or a third-party published figure, flagged and never used alone for a headline number.
The report page shows the confidence label and a one-line summary of the checks behind it.
Editions and corrections
Published once, maintained after
Each report states its base year, edition number and next review date on the page. Markets are monitored between editions, and when conditions move materially the study is revised with a change log describing what moved and why. Errors, when found, are corrected and disclosed on the corrections page rather than silently overwritten.
- Base year 2025, forecasts to 2035
- Next review date printed on every report
- Edition history with change log
- Public corrections policy
- Every version is approved before release, and the report page, fact sheet and data pack come from the same version
Limitations
What this method cannot do
No methodology removes uncertainty. Private company figures are estimates, however well checked. Some markets have little official data, and their reports carry a Low or Medium label for that reason; we state that rather than overstate the evidence. Fast-moving markets can outrun an edition, which is why review dates exist.
We would rather publish an honest range than a precise fiction.
Common questions
Methodology: frequently asked questions
What methodology does Douglas Insights use for market research?
Douglas Insights builds each market from the demand side, as volume multiplied by price by country and segment, anchors it to official data such as dispensing, claims, trade and production statistics, and checks it against at least three independent views of the market from at least two methods before publication.
How many interviews go into a report?
It varies by market and is stated in each report. A typical single-market study draws on structured interviews across manufacturers, distributors and buyers, which set and test the assumptions public data cannot.
What data sources does Douglas Insights use?
Primary interviews, national dispensing and reimbursement data, health surveys and registries, official prices, company filings, customs and production statistics, public claims datasets and structure evidence. Published third-party estimates are used only as a low-weight check.
How does Douglas Insights forecast market growth?
Forecasts separate volume growth from price effect, use named drivers over a ten-year horizon, and publish base, slower and faster scenarios.
What does the confidence label mean?
It states how much of the market value is backed by sourced company revenue or official anchors and how well independent checks agree. High, Medium and Low are defined on this page; any unresolved disagreement caps the label at Low.
Does Douglas Insights count the unorganised market?
Yes. Where unbranded or informally sold products form part of a market, their share is measured from documented sources and shown as its own line. Company coverage is measured against the organised market only.
How often are reports updated?
Each report prints a next review date and is revised when the market moves materially. Revisions ship as new editions with a change log, and corrections are disclosed on the corrections page.
What do licensed buyers receive?
The report and a data pack with market, product, region, country, channel, end-use and structure tables, scenarios, company figures, a validation table and definitions, all from the same approved model version.
See the method applied
Open any report and read its fact sheet: the headline figures, the confidence label and the checks behind it.