Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Research › Market and Competitor Research

Running a Desk Research Pass

How to run desk research: turn the decision into questions, search internal data and published sources, log and judge each source, reconcile the figures.
Tutorialbeginner11 min readUpdated

What desk research is for

The research design guide puts secondary research first for three reasons: it is cheaper, it narrows the primary question so the expensive work is aimed at what is genuinely unknown, and it gives the context that makes primary findings interpretable. Desk research is how that secondary stage is actually done. It can settle facts about a market, such as how many buyers exist, which way demand is moving and what rules apply, and it can show what competitors say and charge. It almost never explains why the business's own customers behave as they do, because nobody else has studied them.

The market and competitor research hub sets out the tiers of published source and the five questions to ask of any of them. This tutorial does not repeat that judgement; it applies it inside a routine, so that each source is logged, judged and used in the same way, and so that the result is a record someone else can check rather than a folder of downloads.

What desk research can doWhat it cannot do
Establish facts about a market that someone has measured with a method you can inspectExplain the reasons behind your own customers' choices
Show the direction of demand over time, from official series and search dataTell you how a new proposition will be received
Describe the rules, regulators and forces acting on a categoryReplace conversations with the people a decision turns on
Narrow the questions primary research must answerTurn a figure with no traceable method into evidence

Before you start

Have the decision written as a sentence with a verb and a date, as the research design guide asks, and know who makes it. Desk research without a decision drifts into reading everything about a topic, and a pass that cannot say what it is for cannot say when it is finished.

Set up the source log before the first search, not after. A spreadsheet is enough. The log is the most important output of the pass, and the columns are set out in the table further down this page.

Agree what counts as good enough. A reversible decision with a modest budget can rest on credible sources and an honest range; a decision that commits a large budget or sets a KPI target for the year needs figures that survive the hub's five questions and a second independent source. Decide in advance, so the bar is not lowered when the convenient figure turns up.

Find out where the business keeps its own evidence: sales and order records, the CRM, analytics, support logs, past research reports and earlier competitor reviews, and who can give you access to each.

The steps

  1. Write the questions the desk work must answer

    Turn the decision into a short, ranked list of questions that published or internal evidence could answer. "Is the workplace market worth entering?" is a topic. "How many organisations in our region fit our serviceable definition, how is spending in the category moving, and what do the main suppliers charge and include?" are questions a desk pass can work on. Mark beside each one whether you expect secondary evidence to answer it fully, partly or not at all. The questions about why buyers act as they do usually belong in the last group, and writing that down now stops the pass from pretending otherwise later.

  2. Start with what the business already holds

    Before searching outside, read what is inside. The audience research guide lists the sources most businesses already have: sales and order data, CRM records, analytics, search queries on the site, support conversations, reviews and the knowledge of sales and service staff. Past research reports and earlier competitor reviews belong here too. Internal evidence describes the business's own customers rather than the whole market, which is its strength and its limit, so note in the log which questions it answers and for whom.

  3. Plan the search by source type

    For each question, list the kinds of source likely to answer it before opening a search engine. The hub's tiers are the plan: official statistics and regulator publications; trade associations and trade press; analyst reports; vendor-published research; competitors' own public material, including annual reports where they are listed; and search data. In the UK, as the hub notes, the Competition and Markets Authority publishes market studies and investigation findings on sectors it has examined, which makes it worth checking whether your category is among them. The PESTEL framework lists sources and signals for each kind of external factor and is a useful prompt for questions about regulation, the economy and technology. For competitors, the competitor analysis guide covers what to look at, and the competitive intelligence guide lists the public sources and what each reveals. If an AI tool suggests sources, treat its suggestions as places to look, as the competitive intelligence guide advises, never as evidence.

  4. Log every source before using it

    Add a row to the source log the moment a source looks useful, before a single figure is copied out of it. Record who produced it, how the data was collected and from whom, when it was collected and published, how it defines the market or the measure, what interest the producer has in the answer, and the link or location. A source that cannot be logged, because nobody can say who made it or how, is not used. Logging first feels slow; it is far quicker than trying, weeks later, to work out where a number in a slide came from.

  5. Judge each source and record its weight

    Ask the hub's five questions of each logged source: who produced it and what did they want, how the data was collected and when, how it defines the market, whether the figure can be reconciled with a second independent source, and what it would look like if the producer's interest ran the other way. Then record a weight in the log, such as strong, usable with caveats, or context only, with a sentence on why. Official and regulator sources, as the hub notes, have documented methods and no commercial interest in the answer, but may lag the market and define categories differently. Search data is current but partial: as the hub puts it, it records expressed demand from people who search, which is not all demand. Use Search Volume as a proxy, read as the keyword research guide describes, and for movements in brand demand see the share of search tutorial.

  6. Extract figures with their definitions and dates

    Copy each figure into the log together with what it counts, the unit, the geography, the period, the date of collection and the page or table it came from. A figure without its definition is unusable, because the most common reason two sources disagree is that they are measuring different things. Copy the wording the source uses for the category, not your paraphrase of it. Where a figure is a forecast rather than a measurement, mark it as one. If a figure reached you through another publication or an AI answer, follow it to the original and log the original; the routine for fact-checking an AI answer lists the usual errors, such as an old figure presented as current or a regional one presented as national.

  7. Reconcile figures that disagree

    When two sources give different figures for what looks like the same thing, compare their definitions, periods, geographies and methods line by line before choosing between them. Often both are right about different things: one includes a channel or a segment the other leaves out, or one counts buyers and the other counts spending. Restate them on a common definition where you can, and record the reconciliation in the log. Where the gap cannot be explained, carry both forward as a range and say so. Figures that will feed a market size go into the market sizing worksheet with their source and confidence, and the market sizing guide shows how top-down and bottom-up estimates are triangulated from them. A forecast that reached a plan without this log is one of the causes the diagnosis for a challenged market size estimate traces back to the source log.

  8. Write the desk summary: settled, conflicting, unknown

    Summarise the pass in a short document with a section for each of these. Settled: what the evidence answers well, with the sources and their weight. Conflicting: where credible sources disagree, the range, and what drives the difference. Unknown: what no source answers. Organise the market findings with the PESTEL framework for the forces acting on the category and the five forces model for the pressures on price and margin, as their pages describe, so the summary feeds a situation analysis without being rewritten. The comparison of SWOT and the five forces shows how the forces' ratings then become evidenced threats in the plan's SWOT. The professional codes of bodies such as the Market Research Society, as the methods hub describes them, expect findings to be reported with their methods and limits; for desk work that means naming the sources, their dates and their gaps beside every conclusion.

  9. Hand the unknowns to the primary research brief

    The unknowns are the real output of the pass. Each becomes a candidate research question for the primary stage, written in the "what is already known" and "research questions" fields of the research brief template, with the desk summary attached. Questions about market structure that customers cannot answer, such as how buying works in a trade or which segments exist, are often better put to people who know the market from the inside, as the guide to interviewing industry experts and the trade describes. Questions about why customers choose go to interviews and surveys. Date the summary and the log, and say when they should be refreshed, because published sources are revised and markets move.

A source log

The log is a table with one row per source, filled in before any figure from that source is used. The columns below are the minimum; add a column for the question each source addresses if the pass covers several.

ColumnWhat to writeWhy it matters
SourceProducer, title and the link or locationSo anyone can open the same source
Producer's interestWhat the producer wanted from publishing itThe hub's first question; sets how sceptically it is read
MethodHow the data was collected, from whomA figure without a method is a rumour with a decimal point
DatesWhen the data was collected and when it was publishedOld figures are often presented as current
DefinitionHow the source defines the market, measure or category, in its own wordsThe usual reason two figures disagree
Figures extractedEach figure with its unit, geography and periodKeeps the number tied to what it counts
WeightStrong, usable with caveats, or context only, with a reasonRecords the judgement so it can be challenged
ReconciliationWhich other source it was checked against, and the resultShows whether the figure survived a second source
Used inThe summary line, slide or worksheet cell that relies on itLets a questioned figure be traced back at once

A log kept this way also makes the next pass faster. When the decision comes round again, the team starts by checking which sources have been revised rather than searching from scratch. Composition of a SERP and search volumes change quickly, so search rows need a capture date more than most.

A coffee roaster researches workplace coffee (illustrative)

Illustrative example — not a real result

Quellstave Coffee logs, reconciles and hands one question to the brief

Quellstave Coffee, an invented coffee roaster selling beans online, is deciding within 8 weeks whether to launch a subscription supplying offices. The desk pass starts with 3 questions: how many offices in its delivery region could buy, how much the workplace coffee market spends and whether that is growing, and what existing suppliers charge and include.

Internal data comes first: 140 existing customers ordered to a business address last year, a sign that some demand already exists. The team then logs 11 published sources before using any of them. Two give very different figures for the size of the workplace coffee market, one roughly 3 times the other. Reading the definitions side by side shows why: the larger figure counts vending machines and the drinks they sell, while the smaller one covers only coffee supplied for staff to make themselves. Quellstave does not sell vending, so the smaller figure is restated as the relevant one and the larger is logged as context only, with the reason.

Search data for office coffee subscriptions is weighted as a proxy for direction, and shows steady interest over the last 2 years. Supplier price pages for 5 competitors are logged with capture dates. The summary lists the market size as settled within a range, growth as conflicting between 2 sources, and one question as unknown: who in an office actually decides on coffee, and what makes them switch supplier. No source answers it, so it becomes the primary research question in the brief.

Common mistakes

  • Searching before writing the questions. Reading everything about a topic, with no way to tell when the pass is finished.
  • Skipping the business's own data. Paying for, or hunting down, figures the CRM and sales records already hold.
  • Using a figure before logging its source. Weeks later nobody can say where the number in the slide came from.
  • Copying numbers without definitions. Two figures that measure different things presented as a disagreement, or one presented as the other.
  • Treating an AI answer or a blog post as the source. The original publication is the source; everything else is a pointer to it.
  • Averaging figures that disagree. The gap is information about definitions, and averaging hides it.
  • Ending with a pile of reports. The output is a short summary of what is settled, conflicting and unknown, and the unknowns are what the brief needs.

Frequently asked questions

What is the difference between desk research and secondary research?

In marketing they mean much the same thing: research done by reading evidence that already exists rather than by gathering new evidence from people. Desk research is the activity and secondary data is what it reads, which includes the business's own sales, analytics and past studies as well as published statistics, regulator reports, trade sources and competitors' public material. The research design guide explains why this stage almost always comes before primary research.

Can an AI tool do the desk research for me?

It can suggest where to look, which saves time at the search-planning stage, but its answer is a lead to sources, never a source. Every figure, date and quotation it offers has to be traced to the original publication, opened and read, and logged like any other source before it is used. The fact-checking routine for AI answers sets out how to split an answer into claims and check each one.

When should desk research stop?

When the questions it was set are either answered by sources that survive scrutiny or clearly cannot be answered from published material. A pass that keeps finding new reports without changing the summary has reached the point where more reading adds nothing. The unanswered questions are the output that matters most, because they become the primary research questions in the brief.

Terms used on this page

Frameworks

Where this page is used

Also related

Sources

  1. Competition and Markets Authority — GOV.UK
  2. Market Research Society — MRS

Related