Marketing Research (Chris West (auth.))

1. 1State the time limit in the problem definition, next to geographical scope and expected reward. The source lists 'the amount of time that is available to solve the problem' as one of three factors that together set the scale and cost of the research. Write it as a decision date, not a vague 'soon'. The date is when the decision gets made, not when the report is due.
2. 2Check the time against the reward. The source says there is no point specifying a major exercise unless the likely gain is commensurate with the cost. When time is short, the reward has to justify the speed premium or you shrink the scope.
3. 3Start with ready-made sources. The source says users 'generally want their data quick and cheap', which customised surveys struggle to deliver. Published research, databases and multi-client studies supply 'instant information'. The internet, intranets and databases give quick, cost-effective access to ready-made market information. Mine these first for a 'rough and ready portrait' of the market.
4. 4Check what your organisation already holds before commissioning anything. Search internal records, the marketing information system and the knowledge base, so the same task is not repeated. Collect salespeople's reports and compare them. The source says that once pooled, they sharpen the picture of the market, though they carry distortions and biases.
5. 5Use informal sources to fill gaps quickly: conversations with colleagues, salespeople's reports and trade show contacts. The source accepts these as often among the most useful information, even though they are neither formal nor scientific. Label them as informal when you report them.
6. 6If primary data is unavoidable, choose the fastest method that fits the question. Electronic collection shortens the path to results. CATI distributes names to the next available interviewer, re-presents interrupted interviews, and closes quota categories once they are filled. CAPI and CASI give electronic data entry, verification and analysis, and the set-up cost no longer limits them to large surveys. Internet questionnaires, focus groups and depth interviews are further options. Match the method to the sample, since CAPI does not suit all locations and CASI needs supervision.
7. 7Add capacity rather than cutting scope where you can. Research services such as field, data processing and analysis are available separately, and 'virtual research companies' buy them in. Freelance consultants take sub-contract work in overload situations. Brief these resources early, because the brief and the supplier selection both consume calendar time.
8. 8Protect planning time. The source says time invested in planning 'will pay considerable dividends'. Pilot the questionnaire, which the book lists as a step, and run any new electronic system alongside the manual one rather than converting in a rush.

The passages give no dated project, so this is a constructed illustration built only from the source's mechanics. A marketing manager must answer 'How much of it can I expect to sell?' (one of the source's key marketing questions) before a planning meeting. She writes the decision date into the problem definition alongside the territories and the expected reward. She first pulls the published reports and multi-client studies available on the market and queries databases over the internet. Next she collects the salespeople's reports and compares them to see where they agree. This gives her a rough market portrait quickly and cheaply. The gap is customer attitudes, which the published sources do not cover. A telephone survey run through a CATI system fits the window: names are distributed automatically, interrupted interviews are re-presented, quotas close when filled, and the data is ready for instant analysis. She flags the salespeople's input as biased and the secondary data as unverified, and reports both caveats with the numbers.

• If the decision date is too close for a customised survey, then build the answer from published research, multi-client studies, databases, internal records and salespeople's reports, and state the limits of that evidence.
• If the potential reward is small relative to the time and cost a full survey needs, then do not specify a major exercise; settle for the rough and ready portrait.
• If primary data is essential and time is short, then choose an electronic method (CATI, CAPI, CASI or internet) over paper, because entry, verification and analysis are built in. Check the method suits the respondents and the location.
• If the workload exceeds your team's capacity in the window, then buy in field, processing or analysis services, or use freelance consultants, instead of dropping quality control.
• If you are tempted to switch to a new electronic information system under deadline, then run it alongside the manual system first, as the source advises.

• Treating time as an afterthought: the problem definition lists scope and reward but no decision date. The tell is a research plan scoped before anyone has asked when the answer is needed.
• Commissioning a customised survey when ready-made data would answer the question. The tell is a fieldwork schedule that finishes after the decision meeting, for a question already covered by published studies.
• Ignoring what the organisation already holds. The tell is the same data gathered separately in different parts of the business, or salespeople's knowledge never consulted.
• Taking speed shortcuts without checking quality: relying on internet or informal data with no verification or stated bias. The tell is a report that cites downloaded figures or sales anecdotes with no caveat.
• Cutting planning and piloting to save days. The tell is a questionnaire or system change that fails in the field and costs more time than the planning would have.

• ▸Marketing question: write the decision the research must inform, in one sentence.
• ▸Decision date: record the date the answer must be in hand, not the report deadline.
• ▸Expected reward: note the likely gain from solving the problem, to test against the cost of speed.
• ▸Geographical scope: list the territories or countries to be covered.
• ▸Ready-made sources checked: record the published reports, multi-client studies, databases and internet sources reviewed, with the date.
• ▸Internal sources checked: record the internal records, MIS entries and salespeople's reports consulted, and the known biases.
• ▸Gap remaining: state what the existing sources do not answer.
• ▸Fastest primary method: name the chosen method (CATI, CAPI, CASI, internet or informal) and why it fits the time window.
• ▸Extra capacity: list the bought-in services or freelance support, with the booking date.
• ▸Quality caveats to report: state what was unverified, informal or biased, so the report discloses it.

1. 1Name the drivers for your own market. The source gives three: the pace of technical development, competitive pressure, and a continual uplift in customer expectations. Write one line per driver describing what is changing in your market. Add the other causes of change, which the source says are diverse and mostly beyond any single supplier's control.
2. 2List every move that takes you into unfamiliar territory: new strategies, new product launches, new target markets, new service packages. The source says the quest for growth or defence pushes companies into these moves, and the problem of losing touch is worst there. Mark each move as familiar or unfamiliar.
3. 3Audit the knowledge base against the four information types: market measurements (size, structure, trends), customer research (practices, intentions, attitudes, expectations, satisfaction, purchasing procedures, media exposure), channel research, and competitive intelligence. For each, record what you hold and when it was last updated. Knowledge is a wasting asset, so an undated item counts as suspect.
4. 4Rank the gaps by decision risk. Gaps that sit under an unfamiliar move, or under a driver that is changing fast, come first. Each gap is a question the next decision depends on, such as 'What do I need to do to beat the competition?'
5. 5Scale the response to the problem before choosing a method. Fix the geographical scope, the potential reward from solving the problem, and the time available. These set the scale and cost of the research. Do not commission a major exercise unless the likely gain is commensurate with its cost.
6. 6Start with cheap sources. Use informal inputs (salespeople's reports, colleagues, trade-show contacts) and desk research. For desk research, use 'sources on sources' such as Marketsearch, Findex and Predicasts, and low-cost published reports that pre-digest and harmonise data. Move to primary research only for gaps that remain.
7. 7Assign one owner for specifying needs. In a large organisation this is the research or business intelligence manager. In a smaller one it is a user such as the marketing, sales or product manager. Have the owner consult the other users and any external agency, so coverage is complete and users buy into the findings.
8. 8Attach a refresh to each unfamiliar move and each fast-changing driver. Update the knowledge base as the business develops, and re-run the audit when a new strategy, product, market or service package is planned.

The Nielsen history in the source shows measurement tools changing as the market's information needs changed. Art Nielsen began with economic and engineering analyses for manufacturers of industrial products. He then launched the Nielsen Drug Index and the Nielsen Food Index. These were retail audits that let manufacturers track their sales and market shares. In 1936 he started a radio audience measurement service using 'audiometers', which recorded the stations a sample of radio sets was tuned to. Radio research then led on to television ratings. The reading for a practitioner is that each new market or medium needed a new way of knowing it. A knowledge base built for one environment would not have covered the next. Apply the same test to yourself. If you enter a new channel or medium, ask whether your current measurement instruments can see it. If they cannot, the gap goes at the top of your audit.

• If you are about to launch a new product, enter a new market, or start a new service package, then treat your existing knowledge as stale for that move and commission fresh information before committing.
• If the likely reward from solving the problem is small relative to the cost of a survey, then use informal sources and desk research rather than primary research.
• If a knowledge item cannot be dated or traced to a source, then treat it as a gap, because knowledge is a wasting asset.
• If several functions will use the findings, then consult them (and any agency) when specifying needs. Findings from a process they did not join are easy to dismiss.
• If the change is driven by factors outside your control (competitors, technology, customer expectations), then monitor continuously. You cannot manage the cause, so you can only keep your knowledge current.

• Assuming the market is stable and reusing last year's plan data. The tell: market size or customer attitude figures with no date or source, or no one who can say when they were last checked.
• Treating formal survey research as the only valid knowledge. The tell: salespeople's reports and trade-show contacts are never logged, even though the source says useful information often comes from them.
• Over-scoping research. The tell: a large primary exercise is commissioned without stating the geographical scope, the reward or the time available, so the cost is out of proportion to the gain.
• Tracking only one driver, such as competitors, and ignoring technical change or rising customer expectations. The tell: the knowledge audit has detailed competitive intelligence but thin customer research on expectations and satisfaction.
• Specifying information needs in isolation. The tell: users who were not consulted reject the unpopular findings, claiming they had no involvement in the preparatory stages.

• ▸Market/segment: name the market this audit covers.
• ▸Technical development: describe what is changing and the evidence you have.
• ▸Competitive pressure: describe the moves competitors are making and how you know.
• ▸Customer expectations: describe how expectations are rising, and the latest customer evidence.
• ▸Other change factors outside our control: list the diverse factors beyond any single supplier.
• ▸Unfamiliar moves planned: list new strategies, products, markets or service packages.
• ▸Knowledge held by type: for market measurement, customer, channel and competitive intelligence, give source and last-updated date.
• ▸Gap and decision at risk: state what you do not know and which decision depends on it.
• ▸Response: informal source, desk research or primary research. Note scope, reward, time available and cost versus gain.
• ▸Owner and consulted users/agency: name who specifies the need and who must buy in. Add the refresh date.

1. 1Diagnose the cause of resistance before you choose a fix. The passages name five: respondent fatigue (over-researched sectors such as telecommunications and IT), cynicism after high-profile research exposure (political opinion polls, focus-group reports), publicity about data protection legislation, voicemail filtering of calls, and the two questions respondents silently ask: 'Do I really want to be bothered?' and 'Am I the right person to ask?'. Write down which of these applies to your target group.
2. 2Answer both silent questions in the introduction. State the reason for the survey, which respondents are eligible, and that each contribution is a valuable input to the client's marketing programme that will ultimately bring some benefit to consumers. Do not leave eligibility implicit.
3. 3Separate three kinds of non-response and treat each differently. Unwillingness: use good interviewers and an introduction stressing the worthwhile enquiry. Genuine inability: catch it in screening questions and politely drop the respondent. Perceived inability: assure the respondent that cooperation is worthwhile and the experience will be painless or even enjoyable.
4. 4Sequence the questionnaire to protect completion. Open with questions that are interesting and easy. Put difficult or highly personal questions last, when the respondent is confident and relaxed. Keep the flow logical, keep every topic visibly relevant to the stated objective, and minimise refer-backs to earlier answers.
5. 5Remove anxiety triggers. Avoid questions that require mental calculation without a calculator. Numeracy levels mean a significant proportion of respondents will make mistakes or refuse. Cut length, repetition, irrelevance and confusion, since each builds resistance.
6. 6For industrial respondents, plan around voicemail. Expect refusals to rise as telephone access gets harder, and build in a second route to the respondent rather than relying on repeated calls.
7. 7For long commitments such as diaries or panels, buy cooperation deliberately. Offer compensation (cash or competition prizes), engage respondents in the survey's objectives and outcome, and run a 'club' with newsletters and regular incentives. Limit diary periods, because the diary itself can bias respondents' behaviour.
8. 8Budget for non-response and test its effect. Refusal rates raise cost and can compromise sample accuracy, so set an expected response rate when choosing the research mix and check for non-response effects before you generalise.

The passages give the telecommunications and IT sectors as examples of respondent fatigue: they have been over-researched, so refusals climb. Suppose you must interview buyers in such a sector by telephone. Your first problem is that voicemail filters make it easy for them to refuse the call. Your second is that the people who do answer ask whether they want to be bothered and whether they are the right person. You respond by opening with a statement of the survey's purpose, who is eligible, and the value of their input to the client's marketing programme. You put screening questions up front so anyone genuinely unable to answer is politely dropped, and you start the interview with easy, interesting questions. You place personal or difficult items at the end and avoid any request for mental arithmetic. Because refusals in this sector also threaten sample accuracy, you record the response rate and note where refusals cluster rather than reporting the achieved sample as if it were representative.

• If respondents refuse because they are tired of being surveyed (over-researched sectors), then stress the worthwhile purpose and the benefit to them, and shorten the interview, rather than adding more contacts.
• If a respondent cannot answer (genuine inability), then screen out early and drop them politely; do not push for an answer.
• If a respondent only thinks they cannot answer, then reassure them that cooperation is worthwhile and the experience will be painless.
• If a question is difficult, personal or needs calculation, then place it at the end of the questionnaire, or remove the calculation.
• If the task demands sustained effort over time (diaries, panels), then pay or reward participants and keep them engaged, and cap the duration to limit bias.

• Treating all non-response as one problem: the same pushy script is used on unwilling, unable and doubtful respondents. The tell is that refusals persist and dropouts occur mid-interview.
• Opening with hard or personal questions: respondents decide the exercise is a waste of time or too difficult. The tell is breakoffs in the first few questions.
• Disorganised flow and heavy refer-backs: topics jump around and the interviewer hunts for earlier answers. The tell is respondents becoming bored, resentful and uncooperative.
• Ignoring voicemail and access barriers in industrial research: the plan assumes calls get through. The tell is rising refusal and unreachable rates and a thin, skewed sample.
• Letting the diary or incentive change behaviour: respondents who track their own purchases become untypical of the market. The tell is panel habits drifting away from market trends over long diary periods.

• ▸Target respondent group: who you must reach and their sector.
• ▸Fatigue/cynicism/data-protection risk: note whether the sector is over-researched or sensitised to polls and privacy publicity.
• ▸Access barrier: voicemail, gatekeepers or other contact obstacles, and the second route you will use.
• ▸Introduction statement: draft wording covering purpose, eligibility and benefit to the client's programme and to consumers.
• ▸Screening questions: the questions that identify genuine inability and trigger a polite drop.
• ▸Reassurance for perceived inability: the line that tells doubtful respondents their cooperation is worthwhile and painless.
• ▸Question order check: confirm easy and interesting questions first, personal or difficult ones last, no mental calculation, minimal refer-backs.
• ▸Incentive/engagement plan: compensation or club mechanics, and the maximum duration for diaries or panels.
• ▸Expected response rate and non-response test: the planned rate and how you will check the effect of non-response.

1. 1Set the accuracy level before you collect anything. The source lists 'Accuracy levels required' and 'Expected response rates' as research-mix decisions. Write down what a decision needs: a rough portrait, or sample-survey-grade figures. Every later check is judged against that level.
2. 2Tag each data item as formal (statistics, reports) or informal (appearances, feelings, suspicions held in staff heads). Informal data is 'considerably more prone to biases and inaccuracies induced by the particular standpoint of the observer'. It can support a decision only once it is proven accurate, so mark it unverified until it is checked.
3. 3Discount by source motive. Salespeople are biased towards their product and company. Customers who know their information will reach a supplier may plant inaccurate information for tactical purposes. Pricing is the most exposed area, because buyers say competitors offered lower prices, bigger discounts or better payment terms. Log these claims as 'reported', not 'confirmed'.
4. 4Verify with a second, independent route. Business intelligence reaches survey-level accuracy through 'multiple research approaches, highly targeted interviewing, and checking procedures'. A fact from a single respondent is acceptable 'provided it can be shown to be accurate'. Accept a single-source fact only after a different source or method confirms it.
5. 5Audit the questionnaire for built-in bias. Multiple-choice lists must include every option. If many respondents choose 'others', name those answers or the results are diluted. For purchase criteria, expect respondents to pick what they feel should matter rather than what they actually weigh. Check that semantic words are read the same way by everyone: 'honest/dishonest' is safe, 'frequency' is not. Where a rating scale is used, convert it to numbers (1 = poor, 10 = excellent) to reduce interpretation problems.
6. 6Check coverage of the decision-making unit. Industrial purchases involve several people, some of them unacknowledged, and management and operators seek different benefits. List every role that influences the purchase and confirm the sample reaches each one, or state which are missing.
7. 7Pilot the instrument before full fieldwork (the contents list includes a 'Pilot survey' section). Then test the effects of non-response: compare responders with non-responders on the variables that matter before treating the results as representative.
8. 8Verify any secondary or internet-sourced material before it enters the analysis (the contents list has 'Verification of information downloaded from the internet'). Apply the controls and data-entry checks the source lists under 'Analysis of Findings', and review the final results for quality before they go into planning.

The source describes a forklift truck purchase. The plant manager, warehouse manager and financial director are visible in the decision, but the truck operators may not be acknowledged, even though they exert a strong 'behind-the-scenes' influence. Management cares about load-carrying capability, speed, fuel economy and purchase price, while operators may care more about comfort and safety. A survey that interviews only the three visible managers produces clean, consistent answers that are systematically incomplete: the benefits operators seek never appear. The fix is a coverage check. List all roles in the buying unit, then add operator interviews or flag the gap in the report. The same source warns that customers may claim a competitor offered lower prices to gain leverage in negotiation. Any such price statement from a salesperson's call report is therefore logged as a claim, then checked against a second route, such as another respondent or published information, before it feeds a pricing decision.

• If a data item comes from a salesperson or from a customer in a negotiation, then treat it as unverified, especially on price, until a second independent source confirms it.
• If only one respondent supplies a fact, then accept it only when you can show it is accurate through a separate approach; otherwise label it a lead, not a finding.
• If more than a small share of answers fall in 'others', then name and code those answers before analysis; otherwise the option list was incomplete and the results are diluted.
• If a question asks respondents to rank purchase criteria, then expect over-statement of what 'should' matter and cross-check against actual purchase behaviour such as brands bought.
• If the buying decision involves several roles, then do not generalise from one role's answers; either cover each role or restrict the claim to the roles you reached.

• Treating informal knowledge as fact because it is vivid or comes from an experienced insider. The tell: no second source and no record of how the claim was checked.
• Taking customer price claims at face value. The tell: reports quote 'competitors are cheaper' with no named source or corroboration.
• Option lists that omit real answers. The tell: a large unexplained 'others' share, or respondents who skip the question.
• Vague scale words read differently by each respondent. The tell: terms like 'frequency' appear undefined, and answers cluster inconsistently across similar respondents.
• Sampling only the visible decision makers. The tell: findings describe management priorities and never mention users or operators.

• ▸Data item: state the fact or figure exactly as it will appear in the report.
• ▸Source and type: name the source and mark it formal or informal.
• ▸Source motive: note any reason the source might bias or plant the information (sales advocacy, negotiating position).
• ▸Accuracy level required: record the precision the decision needs, and why.
• ▸First check: describe the first method used to confirm the item and its result.
• ▸Independent second check: name a different source or method and record whether it agrees.
• ▸Question-design flags: note missing options, large 'others', ambiguous scale words, or 'should matter' bias.
• ▸Roles covered: list the buying-unit roles reached and those missing.
• ▸Status and decision: mark verified, partly verified or unverified, and state whether it may be used in the report.

1. 1Separate the three layers for every item in your market file. Label each as data (raw facts or figures), information (data digested into a pattern) or knowledge (information interpreted into an understanding you can act on). The book treats these as distinct stages, so a folder of figures is not yet a knowledge base.
2. 2Write the decisions first. List the marketing decisions coming up, then the information each one needs. Use the mousetrap test from the Preface: before launch, the better-prepared company has analysed customer requirements, tested the concept, estimated potential demand, researched satisfaction with existing products and obtained intelligence on competitors. Mark each of those five areas as covered or not covered.
3. 3Run an information audit and name the information gap. Inventory what you hold internally (company records, documents, salespeople's reports) and what you can get from published external sources. Then state what the decisions need that neither supplies. Only the gap justifies new research.
4. 4Grade the accuracy of each internal and informal input before it enters the knowledge base. Casual conversations, trade-show contacts and salespeople's reports can be the most useful items you have, but treat them with caution and check them before accepting them. Salespeople can be biased towards their product and company, and customers know their statements will be passed on.
5. 5Treat pricing intelligence as the highest-risk category. When a customer says a competitor offered a lower price, a bigger discount or better payment terms, record it as a claim made in a negotiation, not as a competitor fact. Do not move it into the knowledge base until a second, independent source corroborates it.
6. 6Verify anything pulled from the internet or a database before it is used. The book gives verification of downloaded information its own section. Record the source and the date, and note whether a second source agrees.
7. 7Choose primary research only for the residual gap, and set the accuracy level you need. Small representative samples can describe a whole market, so you do not need a census. Specify the accuracy required, then test for the effects of non-response before you trust the result.
8. 8Interpret before you present. Add value to the findings by reading them against your company knowledge, as the in-house researcher does. Finish each finding with a statement of the opportunity, threat or likely consequence of a specific action, and file that statement as knowledge.

A sales rep reports that a key customer says a competitor has offered lower prices and better payment terms. Taken at face value, this is a single data point, and it comes from a source the book flags twice. Customers know what happens to what they tell salespeople, and they sometimes plant inaccurate information to exert price pressure. Pricing is where this is most pronounced. The rep also has a natural bias towards defending the company's product. So you log it as a customer claim made during a negotiation, not as competitor information. You then look for corroboration from other sources: other customers, trade-show contacts, published information, and the competitor profile you already hold. If two or more independent sources show the same pattern, you have information, meaning a pattern across observations. You then interpret it against company knowledge to reach an actionable statement about what a price move would likely cause. If nothing corroborates it, it stays a flagged claim and does not drive a price cut.

• If an input comes from a salesperson or a customer and concerns price, discounts or payment terms, then treat it as an unverified claim until a second independent source confirms it.
• If the information audit shows the decision's needs are already met by internal records or published sources, then do not commission new survey research. Commission it only for the stated information gap.
• If a finding cannot be turned into a statement of opportunity, threat or likely consequence of a specific action, then it is still information, not knowledge. Send it back for interpretation.
• If material was downloaded from the internet or a database, then verify it and record its source before it enters the knowledge base.
• If your budget is minimal, then compile a rough portrait from internal records, informal sources, databases and the internet. Do not assume competitors know nothing about their markets.

• Filing raw figures and reports as 'knowledge'. The tell: the file is large, yet no one can state what a given item means for a decision.
• Treating informal intelligence as fact, or discarding it as unscientific. The tell: a single sales anecdote drives a decision with no check, or a useful trade-show conversation never gets recorded.
• Accepting customer price claims at face value. The tell: competitor-price 'intelligence' appears only in negotiation contexts and always favours the buyer.
• Commissioning research before auditing what you already hold. The tell: a new survey duplicates facts present in internal records or published sources, and nobody can name the information gap it fills.
• Passing results to decision-makers without interpreting them against company knowledge. The tell: the report restates the data tables and ends without an implication or consequence for any action.

• ▸Decision supported: name the specific marketing decision this item informs.
• ▸Item and layer: write the item and mark it as data, information (a pattern) or knowledge (an interpreted, actionable understanding).
• ▸Source and type: say whether it comes from internal records, a salesperson, a customer, published external material, the internet or a database, or commissioned research.
• ▸Bias or motive check: say who supplied it and whether they gain from you believing it. Pricing claims from customers are always flagged.
• ▸Corroboration: list the independent sources that agree or disagree, or write 'none yet'.
• ▸Verification status: mark verified, pending or unverified. Downloaded material must be verified before use.
• ▸Mousetrap coverage: tick which area it covers: customer requirements, concept test, potential demand, customer satisfaction or competitor intelligence.
• ▸Interpretation against company knowledge: state what this means given what the company already knows.
• ▸Consequence statement: complete 'If we do [action], the likely result is [outcome], representing [opportunity/threat].'

1. 1Treat the report as part of the research, not a postscript. Draft the report skeleton while fieldwork or analysis is still running. Compiling a report exposes data collection and analysis problems while there is still time to correct them, and exposes missing items of information that can still be collected.
2. 2Run the comprehensiveness test first, because it is the easiest of the three objectives to establish. List the original research objectives in one column and the findings that answer each in another. Any objective with no finding is a gap.
3. 3Explain every gap in the report itself. Say whether the research approach did not permit full coverage (more likely with business intelligence approaches than with survey research) or the data originally specified proved not to be relevant. Never let the reader discover a gap on their own.
4. 4Test accuracy before you write for readability. Check that figures reconcile and that the conclusions stay within what the sample can support. Disclose known limits that affect interpretation, such as non-response, exclusion of groups, order effects and recall limits. Do not cover weak research with artistic reporting.
5. 5Build for assimilation. Assume management does not have time to pore over detailed tabulations. Put the answer to the brief up front and keep the detailed tables behind it for those who want them.
6. 6Choose the delivery channel to fit the client's use of the data. The default is a written report supported by an oral presentation. Provide results on disk or by e-mail if the client wants to integrate them into internal information systems. Use electronic delivery when speed matters.
7. 7Write the report as a durable testimonial. In most surveys it is the only lasting evidence the survey happened. It will sit on a shelf after the discussion ends, so it must make sense to someone who was not in the room.
8. 8Match the presenter to the presentation. Numerate and statistical researchers are not always the best people to explain results to users. Use division of labour: let the analyst answer the technical questions and the stronger communicator carry the narrative. In multi-country work, the main contractor owns reviewing findings with sub-contractors, reporting findings and client liaison.

Take a multinational that commissions a European study using a common budget-saving design: the top four countries (France, Germany, Italy and the UK) surveyed in full, plus a sample of small countries such as the Netherlands, Switzerland or Sweden, with results extrapolated to the remaining markets. The comprehensiveness check puts the original objective, 'size and attractiveness of every European market', against the findings. The four large markets and the sampled small markets are covered directly. The remaining markets are covered only by extrapolation, so each one is a gap that must be stated, with its basis. The report says which countries were measured and which were extrapolated, and it does not present the extrapolated figures as survey results. Before writing, the researcher checks accuracy by confirming that the sampled countries are representative of the market types that remain. The headline answer goes first and the detailed tabulations sit behind it. The main contractor reviews the findings with the local sub-contractors before delivery and then handles client liaison. The same logic applies to speed: television audience clients need the previous night's viewing figures on their desks each morning, so an online format replaces printed reports. The format is chosen by how the client uses the data.

• If an original objective has no matching finding, then add a stated explanation of why (approach limits or irrelevant data) rather than leaving the gap silent.
• If compiling the report reveals a data or analysis problem, then fix it now while corrective action is still possible, rather than footnoting it.
• If the audience is management with little time, then lead with conclusions and keep detailed tabulations behind them.
• If the client needs results quickly or wants to feed them into internal systems, then deliver electronically (disk, e-mail or online) alongside or instead of printed copy.
• If the best analyst is not a strong communicator, then split the roles: the analyst supplies the substance and a stronger presenter handles the oral delivery.

• Good research undermined by poor reporting. The tell is that the findings are correct but the client cannot say what they mean for the decision.
• Poor research hidden by artistic reporting. The tell is polished visuals with no stated sample limits, non-response or caveats.
• Silent gaps. The tell is that the client later asks about a topic in the original objectives that the report never mentions, or that was dropped without explanation.
• Data dump for time-poor readers. The tell is that the report is mainly detailed tabulations and has no summary the client can absorb quickly.
• Wrong presenter. The tell is that a statistician reads the numbers aloud to decision makers and the discussion stalls on method rather than implications.

• ▸Original objectives: copy each objective from the research brief, one per line.
• ▸Matching finding: for each objective, name the finding or section that answers it, or write 'GAP'.
• ▸Gap explanation: for each gap, state whether the approach did not permit coverage or the data proved irrelevant.
• ▸Accuracy checks done: list the reconciliations and checks completed, and any problems found and corrected.
• ▸Disclosed limits: note the sample limits you disclose (non-response, excluded groups, order or recall effects, extrapolation).
• ▸Headline answer: write the one- or two-sentence answer to the brief that opens the report.
• ▸Delivery format: choose written report, oral presentation, disk, e-mail or online, and give the reason.
• ▸Presenter: name who presents and who answers technical questions.
• ▸Durability test: state whether someone who was not at the meeting could understand the report from the shelf (yes/no, and what is missing).

1. 1Write the decision down before you commission anything. The source frames research as 'a firm foundation for marketing decisions' and treats defining the marketing problem as the first step (Ch. 4.1). If you cannot state the decision and the options in one sentence, you are not ready to buy data.
2. 2Map the knowledge base first. The source describes knowledge as sitting in employees, databases, reports, plans, budgets and memoranda, and as a 'wasting asset' needing constant updating. List what you already hold (internal data) and what you must get (external data), then record the gap between them (the 'information gap', Ch. 6.4).
3. 3Specify the information required against the decision, not against curiosity. Tie each item to a marketing task and its information requirement (Ch. 4.3, Box 4.7). Drop any item that would not change which option you pick.
4. 4Exhaust cheaper sources before fielding primary research. Internal records, published secondary sources, databases and the internet let even minimal budgets build a 'rough and ready portrait' of the market. Informal sources (colleagues, salespeople's reports, trade-show contacts) can supply useful information too.
5. 5Discount each source for its bias before it enters the decision. Salespeople are 'inevitably biased in favour of his product', and customers sometimes plant inaccurate information, most often on pricing (competitors' lower prices, bigger discounts, better payment terms). Check this information against another source before accepting it.
6. 6Match the research mix to the decision's stakes. Set budget, time, type of data, accuracy level, expected response rate and geographical scope (Ch. 13) in proportion to what a wrong decision would cost.
7. 7Have the findings interpreted against company knowledge, then presented to the decision makers. The in-house researcher's role is to 'add value to the final results by interpreting them in the light of his company knowledge'. Assess the quality of the results (Ch. 20.7) before you integrate them into the market plan.
8. 8Close the loop by recording the options considered, the evidence behind the choice, and the knowledge base updates. This keeps the knowledge base current rather than wasting.

The preface describes the benchmark: the company with the better mousetrap is now much more likely to have analysed customer requirements, tested the concept, estimated potential demand, researched customer satisfaction with existing products and obtained intelligence on competitors before launch. Apply it to a product launch decision. Decision: launch, modify or hold. Step one is internal records and secondary sources for the rough market portrait and the information gap. Step two is research on customer requirements and concept testing, which de-risks targeting. Step three is potential demand estimation, which sizes the prize. Step four is a competitor information profile (Box 4.6). A salesperson reports that a rival is undercutting on price. Treat that as an unverified claim and check it against a second source before it shapes your price. Each of the five stages removes a specific uncertainty, so by launch the decision is informed, targeted and de-risked rather than a bet. The source's contrast: smaller firms once assumed competitors were equally ignorant, and doing so now 'would be courting disaster'.

• If a research item would not change which option you choose, then cut it from the brief.
• If the answer sits in internal records, published sources or a database, then use it before commissioning primary research.
• If information comes from salespeople or customers on pricing, then verify it against an independent source before using it.
• If the decision is high-stakes or hard to reverse, then set a higher accuracy level and budget in the research mix; if low-stakes, a rough secondary-source portrait is enough.
• If findings arrive without interpretation against company knowledge, then have them interpreted before they reach the decision makers.

• Researching without a defined decision: the report is interesting but no one can say which option it favours. The tell is a brief that lists topics, not choices.
• Equating formal with good: dismissing informal intelligence, or trusting a survey simply because it is scientific. The tell is that nobody can say which source is reliable for which question.
• Accepting salesperson or customer claims at face value, especially on price. The tell is a pricing move justified by 'competitors are offering lower' with no second source.
• Letting the knowledge base decay: reusing old plans and data in a changed environment. The tell is a market assumption that no one can date.
• Assuming competitors are as ignorant as you are, or skipping research on budget grounds. The tell is a launch with no concept test, demand estimate or competitor intelligence behind it.

• ▸Decision to be made: state in one sentence, with the options (e.g. launch / modify / hold).
• ▸Marketing task and information requirement: link the decision to the task it serves (Box 4.7 logic).
• ▸Existing knowledge (internal data): list records, reports and people that already hold relevant knowledge, with dates.
• ▸Information gap: state what you need to know that you do not.
• ▸Sources to try, cheapest first: internal, secondary, database/internet, informal, then primary research.
• ▸Source bias check: for each source, note likely bias and the second source that will verify it.
• ▸Research mix limits: budget, deadline, accuracy level, geographical scope.
• ▸Interpretation owner: name who interprets findings against company knowledge and presents to decision makers.
• ▸Knowledge base update: record what changed and where it is stored.

1. 1Start from the claim the evidence supports, not the one it doesn't. The source finds that companies using market research are more likely to make above-average profits and less likely to incur losses. It states that no causal relationship was demonstrated. Read research use as a marker of a management style that produces good results, and write your business case on that basis.
2. 2Name the outcome you want to move before you commission anything. Pick from the four in the construct: market share, profitability, a new product launch, or reduced risk of loss. Research exists only to solve a problem, and 'if there is no problem there is no need for research'.
3. 3Classify the trigger. The source gives three periodic triggers: change in the marketplace, actual performance failing to meet expectations, or a requirement to take a new product or marketing initiative. Also decide whether you need continuous data to monitor progress or a one-off answer. Once agreed, continuous data becomes a regular part of the marketing programme.
4. 4Define the problem so that it states content, coverage, timing and scale. Check first whether the problem needs market data at all, and whether the data can come from inside the company rather than an external survey.
5. 5Build the routine monitoring baseline for the outcome. The source's routine information list covers sales by product category, customer category, channel and price level; margins; inventory; promotional spend; awareness; total demand and demand by segment; market share and share by segment and channel; competitor products, prices, key accounts and promotions; and forecasts. Profit and share outcomes cannot be judged without it.
6. 6Pre-test and post-test every significant marketing proposal. The source says companies with a research culture are significantly more likely to do this. Test before launch to reduce the risk of loss. Test after launch to learn whether the initiative produced the intended result.
7. 7Add competitive intelligence when share is the goal. Offers from competitors have become less differentiated and demand growth is slow. Companies therefore need share gains, or need to defend against rivals who are gaining. Use retail audits and customer satisfaction surveys for comparative data. Go further and predict competitors' likely next moves.
8. 8Treat informal inputs (distributors, key customers, trade shows, sales reports) as leads. The source warns that all informally obtained information is open to challenge. Use them to shape the formal question, then verify before committing money.

Take the one quantitative finding in the source and use it to decide where you stand. The survey found that 60 per cent of companies with sales under £2.5 million had no research budget, falling to 15 per cent for companies with sales above £20 million. Fast moving consumer goods suppliers were twice as likely to use research as other types of company. Companies that used research were more likely to make above-average profits and less likely to incur losses, but no causality was shown. Suppose you run a small non-FMCG supplier with no research budget, and you justify that by saying you already know your market. The source says this is the usual reason firms skip research, and that the cost 'cannot be justified' is the standard objection. The correct move is not to buy a large survey. Start with one live decision, such as a new product launch. Pull the routine data you already hold (sales by product and customer category, margins, competitor prices and share movements), run a pre-test of the proposal, and plan a post-test. Judge success against the outcome you named at the start, such as share in one segment or avoided launch loss. You are building the research habit the source associates with better profit performance, without claiming it guarantees that performance.

• If you cannot state the problem in terms of content, coverage, timing and scale, then do not commission research yet. Define the problem first.
• If the problem can be answered from internal records or existing secondary sources, then use those before paying for primary research.
• If you are about to launch or relaunch a product or a major marketing programme, then pre-test it, and schedule a post-test so the result feeds the next decision.
• If your goal is to gain or defend market share against near-identical competitor offers, then invest in competitive intelligence, not just customer data.
• If someone cites the research-profit correlation as proof that research causes profit, then restate it as association only, because the source says causation was not demonstrated.

• Claiming research caused the profits. The tell is a business case that says 'research users earn more, so research will raise our profit'. The source supports only an inference about management style.
• Skipping research because 'we already know our markets'. The tell is that the only evidence is anecdote, and no pre- or post-test results exist for the last launch.
• Relying on informal conversations alone. The tell is that decisions rest on distributor chat or trade-show remarks that nobody has checked, which the source says are open to challenge.
• Commissioning a survey with no defined problem. The tell is a brief that cannot say what decision the data will change, or the scale and timing required.
• Tracking company data but ignoring competitors and the market. The tell is that sales and margin reports exist but there is no data on market share movement, competitor prices or demand by segment, so you cannot tell whether a result is a gain or merely market growth.

• ▸Target outcome: market share, profitability, new product launch success, or reduced loss risk.
• ▸Trigger: market change, performance shortfall, or new initiative. Say which one and describe it.
• ▸Problem statement: one or two sentences giving content, coverage, timing and scale.
• ▸Data already held internally: list the routine information you can pull now (sales, margins, share, competitor data).
• ▸Information gap: what the decision needs that you do not yet have.
• ▸Pre-test plan: how the proposal will be tested before launch, and the result that would stop or change it.
• ▸Post-test plan: when and how you will measure the result against the target outcome.
• ▸Competitor intelligence needed: current activity, and likely future moves, of named rivals.
• ▸Evidence caveat: what the data can and cannot prove, such as association versus cause.

1. 1Write the decision first, not the research topic. State in one sentence what the marketing manager will do differently depending on the findings. The source's test is that clients want 'a timely and accurate solution to their problem', so the problem is the decision, not the survey.
2. 2Name the marketing problem and the marketing task it sits inside (for example, launching a product). The book links marketing tasks to market information requirements (ch. 4.3), so list the information that task needs before choosing any technique.
3. 3Break the problem into discrete information needs, each written as a question that can be answered. The passages' launch example gives the pattern: customer requirements, concept acceptance, potential demand, customer satisfaction with existing products, competitor intelligence. Each need becomes one line in the specification.
4. 4Run an information audit before commissioning anything. For each need, check what already exists internally (sales reports, colleagues, trade-show contacts) or in published and secondary sources. What remains is the information gap, and only the gap justifies new primary research.
5. 5Add the parameters that bound the project: geographical scope, accuracy level required, budget, and the date the decision must be made. These match the 'research mix' factors the book lists (budget, time, accuracy, geographic scope). Put them in the definition, because they limit which techniques are feasible.
6. 6Set the reward test. For each information need, state what it is worth and what happens if it stays unknown. If you cannot say, drop the need or demote it to 'ancillary' (ch. 4.4). Do not let it enlarge the scope.
7. 7Assign responsibility. Name who owns the problem definition, who signs off the specification, and who must share internal knowledge. The passages warn that 'knowledge is power' and holders may withhold information from decision makers.
8. 8Convert the definition into the written brief and confirm it in a briefing meeting. The brief is the problem definition in transferable form. Revise the information yield in writing if scope changes mid-project (ch. 20.4), rather than letting it drift.

The preface describes the company with 'the better mousetrap': before launch it analysed customer requirements, tested the concept, estimated potential demand, researched customer satisfaction with existing products and gathered intelligence on competitors. A poorly defined brief would say 'research the market for our new mousetrap'. A well-defined one starts from the decision, which is whether and how to launch. It then lists those five needs as separate questions. Before commissioning a survey, the team audits what it already knows from internal records, salespeople's reports and published sources, and marks the remaining information gaps. It then attaches the deadline (the launch date), the budget, the countries or regions covered and the accuracy required for the demand estimate. Needs that would not change the launch decision, such as nice-to-have attitude data, are labelled ancillary and cut or deferred. The result is a specification where each question has an owner, a source and a purpose, and the findings are 'right, can be shown to be right, and are on time'.

• If you cannot state a decision that the findings would change, then stop and redefine the problem before spending any research budget.
• If an information need is already answered by internal records, colleagues, salespeople's reports or secondary sources, then exclude it from primary research and record it as covered in the audit.
• If a need would not alter the decision or has no stated value, then classify it as ancillary and cut or defer it, rather than letting it inflate the scope.
• If the deadline or budget is fixed, then size accuracy and geographic scope to fit them in the definition. A quick, rough answer, such as a few well-placed telephone calls, is acceptable if it is sufficient for the decision.
• If scope changes after the brief is issued, then reissue the revised information requirements in writing and re-confirm timing and budget with the researcher.

• Defining a technique instead of a problem ('we need a survey'). The tell: the brief names a method and sample but no decision it will inform.
• Specifying information that already exists in the organisation. The tell: findings repeat what salespeople or colleagues already knew, because no information audit was done.
• Open-ended scope with every desirable question included. The tell: a long questionnaire, an overrun budget, and no ranking of needs by their value to the decision.
• Leaving timing and rewards out of the definition. The tell: findings arrive after the decision date, or nobody can say what the research was worth.
• Unclear ownership of the definition, with internal information withheld. The tell: key decision makers are surprised that market information already existed, or each stakeholder gives the researcher a different problem.

• ▸Marketing problem and task: one sentence naming the problem and the marketing task it belongs to (e.g. product launch).
• ▸Decision to be made: what you will do differently depending on the findings, and who decides.
• ▸Information needs: each need as a separate answerable question (e.g. customer requirements, potential demand, competitor intelligence).
• ▸Existing sources and information gap: for each need, what internal or secondary information exists and what is still missing.
• ▸Value of each need: what is gained, or risked, if the need is answered or left unknown. Mark ancillary items.
• ▸Scope limits: geographical coverage, market or segment, and accuracy required.
• ▸Timing: the date the decision must be made and the date findings are needed.
• ▸Budget: the amount available and any constraint on technique.
• ▸Owner and sign-off: who owns the definition, who approves it, and who must supply internal knowledge.

1. 1Write the decision the research must support in one sentence, and the information that decision needs (Chapter 4: define the problem, then specify the information required). Every later choice is tested against this sentence. If a design element does not serve it, drop the element.
2. 2Set the five structural choices explicitly and in order: purpose, approach, type of market, type of survey, and how the information is obtained (the Chapter 5 structure). Write one line of justification for each. A blank justification means the choice was made by habit.
3. 3Run the ten constraint filters from the research-mix chapter against the draft: budget, time, type of data required, availability of secondary data, volume and complexity of data, sources of data, expected response rates, accuracy level required, geographical scope, and research resources available. Mark each as a binding or non-binding constraint.
4. 4Exhaust internal and secondary sources before commissioning primary work. Run an information audit to find the gap (Chapter 6). Internet and database sources now let even minimal budgets compile a 'rough and ready portrait' of a market. Commission primary research only for the gap that remains.
5. 5Match the instrument to what the respondent will really do. Ask whether the market is consumer or industrial, because stated purchasing intentions track behaviour more closely in industrial and professional markets. Ask whether the delivery channel suits the respondents: an e-mail questionnaire full of images is slow to download and takes storage space on older PCs and slow modems.
6. 6Build in a way to see past what people say. Where the decision depends on what drives behaviour, collect attribute ratings and an overall satisfaction rating, then derive importance by regression instead of relying on stated importance alone.
7. 7Check the design for the pace of the market and the user. Research users want data 'quick and cheap', which customised surveys struggle to deliver. Consider published research, databases or multi-client studies if they answer the question adequately.
8. 8Pilot the design (a pilot survey is a listed step) and plan a test for non-response effects before fieldwork. Record which constraint you accepted, so the report can state the limits of the findings.

Customer satisfaction research shows design appropriateness in practice. A supplier of venture capital wants to know why customers choose it. A design that only asks customers to rank service aspects by importance is mismatched to the question. Surveys repeatedly show that users rate 'a deal on favourable terms to themselves' as the most important aspect, yet they still use suppliers whose terms they regard as adverse. The better-matched design has the three elements of a customer satisfaction survey: a list of the attributes customers use to judge suppliers, ratings of each supplier's performance on every attribute, and a rating of overall satisfaction. Importance weightings are added, and explanations for poor ratings, loyalty and willingness to recommend can be asked as well. Regression of attribute performance on overall satisfaction then yields derived importance. Setting stated importance against derived importance exposes the gap between what customers say matters and what influences their behaviour. Awareness questions can be added, because a supplier with low awareness outside its customer base is unlikely to gain share however satisfied its customers are.

• If the market has few customers and competitors, or is static, and the company is small, then use an informal system (collect information when needed, or hold a small manual database) instead of investing in a major marketing information system.
• If the research aims to forecast short-term demand or predict new-product success in an industrial or professional market, then treat stated purchasing intentions as a strong input. In consumer markets, use them as one input to a forecasting model, not as the forecast.
• If the decision depends on what drives customer behaviour, then do not rely on stated importance alone. Collect performance and overall satisfaction ratings so derived importance can be calculated and compared.
• If budget is minimal or speed matters most, then start with secondary sources, the internet and published or multi-client research. Commission a customised survey only for the gap those sources leave.
• If the respondent base uses older PCs and slow connections, then avoid heavy e-mail or internet questionnaires, with images in particular. Choose a lighter or different delivery method.

• Choosing technique before defining information needs. The tell: you cannot state in one sentence the decision the data will inform, and the questionnaire collects 'interesting' items.
• Taking stated importance at face value. The tell: customers rate a factor as top priority, yet the same customers buy from suppliers who score badly on it.
• Over-building the information system for the context. The tell: a small firm in a concentrated, static market runs a costly formal system that few people use, when an informal arrangement would serve.
• Ignoring respondent-side practicalities of the data collection method. The tell: low response from an e-mail survey, with complaints about download time or storage, concentrated among older-equipment owners, so the sample skews.
• Commissioning primary research that existing sources already answer. The tell: no information audit exists, and findings duplicate published or internal data that could have been had faster and cheaper.

• ▸Decision to be supported: one sentence stating the decision and who makes it.
• ▸Information required: the specific items the decision needs, and the existing-data gap found by the information audit.
• ▸Purpose / approach: the chosen research purpose and approach, with one-line justification.
• ▸Market type and survey type: consumer or industrial, the survey type chosen, and why it suits that market.
• ▸Sources and secondary availability: what internal and secondary data cover, and what primary data must fill.
• ▸Constraints (budget, time, resources): the limit on each, and which one binds the design most.
• ▸Response and accuracy: expected response rate, the accuracy level required, and the plan for testing non-response effects.
• ▸Geographical scope and delivery channel: where respondents are, and whether the method works for their equipment and access.
• ▸Validity check: how the design will reveal behaviour as well as stated opinion (e.g. derived versus stated importance).
• ▸Accepted trade-offs: the constraints you accepted and the limits this places on the findings.

1. 1Write the universe in one sentence before choosing any sample. Name who determines purchase, not only who buys. The passages say the primary target is customers, whose requirements, practices and attitudes set marketing strategy. They also say surveys reach distributors (retail and wholesale), designers and others who influence what is purchased. Decide which of these groups are in the universe and which are out, and record why.
2. 2Check whether a sample survey is the right instrument at all. The source says sample surveys are sometimes impractical, too slow or too costly, and that business or competitive intelligence can be an equally valid source. If a representative sample cannot be reached, change the approach rather than run a weak survey.
3. 3Set sample size from three named factors: the structure of the market (how concentrated or fragmented it is), the level of detail required, and the accuracy required. Write down the sub-groups you must report on separately. Detail requirements, not total headcount, usually force the size up. Accept that these three factors also set the cost.
4. 4Choose a structure that mirrors the universe. The source calls for 'representative sampling', and its examples include a general sample representative of the whole population versus a defined group such as motorists or mothers with young children. Match the structure to the question: use a defined sample for a defined segment, and a general representative sample for market-wide claims. Check the structure against known market classifications (socio-economic, geodemographic, industrial) so that you can see the gaps.
5. 5Where the market is small or concentrated, test whether a census or near-census is feasible. The source notes that point-of-sale data are moving retail research toward censuses, and that high accuracy comes from covering large samples of respondents or a high proportion of outlets. Where a census is partial, record who is missing, because not all retailers cooperate.
6. 6Plan non-response before fieldwork. Set an expected response rate for the chosen method, and decide how you will handle refusals. Schedule a test for the effects of non-response, which the source lists as its own step for internet surveys. Compare responders with the known universe to see whether non-response is skewing results.
7. 7Screen for topics where accuracy will fail whatever the sample. The source names sensitive areas: sex, personal hygiene and personal finances for individuals, and product formulations, production methods and finances for companies. If respondents will refuse or give false answers, the sample cannot fix it. Check that the approach overcomes the problem, and if not, abandon the exercise.
8. 8Report sample error and state inference limits. Separate descriptive statistics (what the sample says) from inferential statistics (what you claim about the market). State the universe, structure, size, response rate and known exclusions next to every headline number.

The source's retail research comparison shows sampling quality being checked against an external reference. A retail audit draws on a sample of retail outlets. A consumer panel draws on a sample of customers. Both are described as achieving high accuracy by covering large samples of respondents or a high proportion of outlets. The source also says what passes through retail outlets should agree with purchases made by customers. That agreement is a validity test: if the two samples disagree, one of them is mis-structured or has gaps. Electronic point-of-sale collection is moving audits toward censuses, which removes sampling error, but not all retailers cooperate. So even a near-census carries exclusion, and the missing retailers must be identified and described. A practitioner would run both sources, reconcile the totals, name the non-cooperating outlet types, and say in the report which part of the market the figures cover. The passages give no figures for this case, so none are invented here.

• If a representative sample is impractical, too slow or too costly, then use business or competitive intelligence rather than a thin survey, and label the findings as non-survey evidence.
• If the client needs detailed breakdowns by segment, then size the sample for the smallest segment you must report, since detail required drives sample size alongside market structure and accuracy.
• If the question is about the whole market, then use a general sample representative of the population; if it is about a defined group such as motorists, then use a defined sample of that group.
• If the topic is sensitive and respondents are likely to refuse or give false answers, then confirm the approach can overcome it, and abandon the research if it cannot.
• If a near-census is available but some firms do not cooperate, then treat it as a sample with exclusion, and name who is missing before generalising.

• Defining the universe as customers only when distributors, designers or other influencers shape purchase. The tell: the report cannot explain purchase decisions made upstream of the respondents.
• Fixing sample size on budget alone, ignoring market structure, detail required and accuracy required. The tell: sub-group figures rest on very few respondents but are presented as findings.
• Treating a near-census as complete. The tell: no mention of retailers or firms that declined to cooperate, and totals that do not reconcile with another source.
• Ignoring non-response and refusal. The tell: no response rate is reported, and no comparison of responders with the universe.
• Trusting a sample on a sensitive topic where answers will be withheld or false. The tell: suspiciously uniform or implausible answers on finances, formulations or production methods, with no check that the approach can cope.

• ▸Universe definition: state who is in (customers, distributors, designers, other influencers) and who is out, with the reason.
• ▸Research approach check: survey, or business or competitive intelligence? Note why a sample survey is or is not practical, fast and affordable enough.
• ▸Sample structure: general representative sample or defined sample (name the group), and the classification used to check it (socio-economic, geodemographic, industrial).
• ▸Sample size drivers: market structure, level of detail required (list the sub-groups to report), and accuracy required.
• ▸Census feasibility: could a census or near-census be used? List who is missing or non-cooperating.
• ▸Expected response rate and refusal plan: state the expected rate for the method and how refusals will be handled.
• ▸Non-response test: describe how responders will be compared with the universe, and the result.
• ▸Sensitive-topic screen: list topics at risk of refusal or false answers and decide go, adapt or abandon.
• ▸Inference statement: sample error noted, what is claimed about the market, and known exclusions to disclose in the report.

1. 1Agree the information needs before drafting a single question. Every question must map to a decision the management user will make; cut any question that does not. The passages tie questionnaire quality to the survey's objectives and to keeping the instrument from becoming too long.
2. 2Choose the question format deliberately. Fixed-category scales (e.g. 'extremely important' to 'not important at all') force answers into a predetermined straightjacket but analyse easily. Open-ended questions invite variation in interviewer interpretation, responses that range from monosyllabic to verbose, and difficult analysis at volume. Pick per question, not per questionnaire.
3. 3Write each question in the fewest words that convey the exact meaning. Space and time are limited on a questionnaire, so you cannot pad as you would in speech. Replace vague or colloquial words with precise ones: 'sort' becomes 'brand', 'like' becomes 'prefer', 'where' becomes 'in which shop', 'get' becomes 'purchase'.
4. 4Screen every word against the Box 9.3 pitfalls: ambiguous words, words that mean different things to different respondents, offensive words, words that fail to capture the full range of options, jargon and slang, and wording outside the respondent's normal experience. Check for social-group variation, e.g. 'dinner' can be the midday or the evening meal.
5. 5Handle sensitive terms (race, sex, bodily functions, religion, income) with care. If a topic is unavoidable, choose the form least likely to offend. Asking whether someone is 'a Jew' rather than 'Jewish' is the passage's example of a form that can give offence.
6. 6Build the instrument so it works without the author present. Interviewers, not the designer, apply it, and training and briefing cannot be relied on to produce consistent responses. The questionnaire itself is the first line of defence.
7. 7For self-completion formats (mail, fax, internet), raise clarity, unambiguity and simplicity further, because nobody is there to explain unclear points. Test against the disincentives: too long, too complex, obviously ambiguous, questions respondents cannot or will not answer, boring questions, and a survey that looks futile or a waste of time.
8. 8Have management approve the questionnaire before full-scale launch. Ask them to comment on missing or superfluous questions, terminology, whether respondents can actually answer, and whether the wording is likely to be ambiguous. They are not experts in design, so direct their review to these four points.

The passages supply a word-substitution test you can run on a draft consumer-purchase questionnaire. A draft asks: 'Which sort do you like, and where do you get it?' Each word fails. 'Sort' is colloquial where 'brand' is meant. 'Like' is vague where 'prefer' is meant. 'Where' does not say whether you want a town or a store, so it becomes 'in which shop'. 'Get' could mean buy, receive or borrow, so it becomes 'purchase'. The rewrite is: 'Which brand do you prefer, and in which shop do you purchase it?' Next, check the draft for words that vary by respondent, such as 'dinner', which one respondent reads as the midday meal and another as the evening meal. Specify the meal or time of day instead. Then send the draft to the managers who will use the results. They will not redesign it, but they can flag a missing option, a term their customers do not use, or a question respondents could not answer. Only after that review does the full-scale launch go ahead.

• If a question will be applied by interviewers rather than by you, then do not count on training or briefing to harmonise interpretation; fix the ambiguity in the wording itself.
• If the questionnaire is self-completion (mail, fax, internet), then shorten it, simplify it and make every question self-explanatory, because no one can clarify a point for the respondent.
• If a word is colloquial or vague ('sort', 'like', 'get', 'where'), then replace it with the precise term ('brand', 'prefer', 'purchase', 'in which shop').
• If you need easy analysis across large numbers of respondents, then use fixed-category scales; if you accept variation in interviewer interpretation and harder analysis in exchange for respondents' own words, use open-ended questions.
• If a sensitive subject (race, sex, religion, income, bodily functions) cannot be avoided, then keep it but choose the least offensive phrasing and have management review the wording before launch.

• Padding questions with extra words as in ordinary speech. The tell: long, wandering questions that take interviewers or respondents longer to get through, when meaning should be conveyed in a minimum number of words.
• Relying on interviewer briefing to patch an ambiguous instrument. The tell: different interviewers explain the same question differently, and answers are inconsistent between respondents.
• Using words that mean different things to different groups, such as 'dinner'. The tell: answers cluster by social group in ways unrelated to the behaviour you are measuring.
• Skipping management sign-off, or asking management for design expertise they lack. The tell: missing or superfluous questions, unfamiliar terminology, and questions respondents cannot answer are discovered after launch.
• Treating self-completion like an interviewer-led survey. The tell: low participation, because the questionnaire is too long, too complex, obviously ambiguous, boring, or seems a waste of time. A further risk: respondents who read through the whole questionnaire first can work out the survey's objectives and bias their answers.

• ▸Survey objective and decision served: state the decision management will make with the results; delete questions that do not serve it.
• ▸Question format per item: mark each question as fixed-category or open-ended, and note the analysis or interpretation trade-off accepted.
• ▸Word screen (Box 9.3): list each flagged word as ambiguous, variable-meaning, offensive, range-limiting, jargon/slang or outside respondent experience, with its replacement.
• ▸Colloquial substitutions: record each vague word and its precise form (e.g. 'sort' to 'brand').
• ▸Sensitive terms: list any race, sex, bodily function, religion or income wording and the chosen non-offensive form.
• ▸Administration mode: interviewer-applied or self-completion (mail, fax, internet); if self-completion, note how each unclear point is made self-explanatory.
• ▸Participation disincentive check: tick too long, too complex, ambiguous, unanswerable, boring, perceived as futile, and note the fix for each.
• ▸Management review: record reviewer, date, and comments on missing or superfluous questions, terminology, respondent ability to answer and ambiguity.
• ▸Launch approval: confirm management has agreed content before full-scale launch, with name and date.

1. 1Start by pricing the decision, not the survey. Write down what the data will be used for and what rides on it. The passages show the range: a 'nice to know' item at one end, and at the other TV advertising revenue, executive performance bonuses, and roughly £350 million of London bus revenue apportioned by passenger survey. The higher the stakes, the more accuracy the budget has to buy.
2. 2Test the 'we already know our market and the cost can't be justified' objection. The passages name it as the usual reason firms spend nothing. Write down what you believe about the market, then list which beliefs would change a decision if they proved wrong. Spend only on those.
3. 3Set a budget tier before choosing techniques. With a minimal budget, build a 'rough and ready portrait' from the internet, intranets and databases. With a mid-range budget, add cheap internal and informal sources. Reserve large-scale surveys and advanced techniques for decisions where the data's value is high.
4. 4Mine free internal sources first. Collect salespeople's reports (acknowledged as a major, rarely pooled information store), contacts from trade shows, and conversations with colleagues. Pool and compare them, because the passages say the picture sharpens when data from all salespeople is rationalised. Treat the result as biased but cheap.
5. 5Check for work already paid for. Look in the existing marketing information system and secondary sources (written records, data disks, CD-ROMs, internal files, syndicated feeds) so the same task isn't repeated in separate parts of the organisation.
6. 6Re-test technique costs against current prices. Computer-aided interviewing (CATI, CAPI, CASI) once had set-up costs so high that it paid only on large-scale surveys. It can now be used even when interview numbers are low, so an old 'too expensive' verdict may no longer hold.
7. 7Stage the spend on systems. If you build a knowledge or information system, run a manual version first and prove it works and is accepted before investing in hardware and software. A part-time, update-on-demand approach can be the more cost-efficient option when the data does not need constant consultation.
8. 8Fix the accuracy–cost trade-off in writing. State the accuracy level needed, the scale and techniques the budget supports, and what you are knowingly giving up. Use this to compare agency proposals and to explain cost variations.

The passages give a stakes-versus-budget contrast. A UK survey found that 60 per cent of companies with sales under £2.5 million had no research budget, against 15 per cent for companies with sales above £20 million. Meanwhile some £350 million of London bus revenue, not collected on the buses, is apportioned between operators on the basis of a passenger survey. The passage's own conclusion is that 'with stakes this high, the research must be accurate and acceptable to all the parties concerned.' Applied to budgeting: the bus survey's value justifies a large, accurate programme. A small firm with no research budget can still compile a rough and ready portrait through the internet, databases, salespeople's reports and trade-show contacts. The mistake is to apply the small-firm approach to a decision with bus-survey stakes, or the reverse. Match spend to the value placed on the data, and accept the lower accuracy at the lower tier.

• If the data will drive revenue, bonuses or other financial transactions, then budget for high accuracy and sampling that all parties will accept; do not use a rough portrait.
• If the decision is low-stakes or the budget is minimal, then use the internet, databases, intranets and internal sources to build a rough picture, and say explicitly that it is rough.
• If the stated reason for not researching is 'cost can't be justified', then ask what a wrong belief would cost in the decision before accepting it.
• If you last priced a technique (such as computer-aided interviewing) when set-up costs forced large samples, then re-price it, because it now works at low interview numbers.
• If you are tempted to buy a knowledge-management system, then run a manual version first and invest in hardware and software only after it is proven and accepted.

• Treating research as an unaffordable luxury: no budget line exists and 'we already know our market' is the stated reason. The tell is decisions made with no pre- or post-testing.
• Assuming formal, scientific surveys are always best and spending on them when a casual source would do. The tell is a costly survey answering a question the sales team could have answered.
• Buying low-cost data for a high-stakes decision. The tell is that stakeholders reject the numbers as inaccurate or unacceptable.
• Relying on an outdated cost assumption about techniques. The tell is a computer-aided or low-volume approach dismissed as 'only for large surveys'.
• Spending on systems before proving the process. The tell is hardware or software bought while no staff share data or use a manual version.

• ▸Decision to be supported: one line naming the decision the data feeds.
• ▸Value at stake: what depends on the data (revenue, bonuses, launch, none) and whether others must accept the numbers.
• ▸Total budget available: the amount, or 'none', including staff time.
• ▸Free/internal sources checked: salespeople's reports, colleagues, trade-show contacts, existing records, internal files.
• ▸Low-cost external sources checked: internet, databases, syndicated data, intranet.
• ▸Accuracy level required: the minimum acceptable precision for this decision.
• ▸Technique and scale chosen: method and sample size the budget supports, noting any recent cost changes.
• ▸Accuracy knowingly traded away: what the budget cannot buy and the risk this creates.
• ▸Review trigger: the point at which the value of the data justifies reopening the budget.