Target audience research helps you decide which customers to spend your budget on, what to offer them and how to explain the price. It is useful when launching a product, entering a new market or reviewing campaigns that generate enquiries without sales.
If customers walk away after receiving a quote, resist the urge to cut the price. Interviews may reveal that they are comparing services with different scopes and cannot see what yours includes. Test whether a clearer explanation of the work and the expected result helps them understand the price.
How to research your target audience
- Choose a decision. Define whether you need to find buyers for a new product, understand lost deals or refine an offer.
- Gather evidence. Compare enquiries, sales, site behaviour, search terms and reviews. Retain sources and customers’ exact words.
- Fill gaps. Use interviews to understand decisions, surveys to estimate how widespread a known pattern is and observation to see what people do.
- Identify segments. Compare buying situations, goals, constraints and criteria. Look for differences that call for different decisions.
- Profile each segment. Record confirmed requirements, barriers, evidence and open questions.
- Test a change. Choose what to change in the product, terms, advertising or site and decide how you will measure it.
Which decisions does research inform?
A new service calls for evidence about how people currently solve the problem and what an offer must include. If you are investigating lost deals, talk to people who chose another supplier or postponed the purchase. The decision determines who to talk to and which records to review.
For a local clinic, access to a branch may matter. For an equipment supplier, procurement requirements and the people involved in approval may matter more. Include demographics where they help explain a decision or find the audience in a relevant channel. A fictional biography cannot tell you which programme or delivery terms to offer.
What has changed in audience research?
The idea of looking at the job a customer needs done predates generative AI. Harvard Business School’s 2011 account of Clayton Christensen’s milkshake research illustrates the approach. What has changed is the speed at which teams can process evidence. Dovetail’s documentation describes AI tools for working with research material. Choose participants, check transcripts and validate interpretations as part of your process.
Synthetic respondents are model-generated answers designed to simulate research participants. NIQ discusses their use for early exploration and the need to validate against real consumer data. Asking a chatbot to answer on behalf of your customers gives you hypotheses to investigate.
Start with a question you can act on
An English-language school might ask why people book a trial and fail to attend, who needs interview preparation, why learners stop taking lessons or how to explain the cost of individual tuition. Each question requires different evidence. Decide what you will act on first and list what you still need to learn about timing, alternatives, requirements and concerns.
Record your working hypothesis separately. You might expect people to miss a trial because they do not know what will happen in it. If the evidence points to inconvenient times or slow replies, keep that finding even when it contradicts the team’s expectations.
Where to find evidence
Use a consistent record for each observation: ID, date, product, source, buying situation, exact words and outcome. Store your interpretation separately. A lost enquiry because a school offered a general course could indicate a mismatch with the buyer’s goal. The record should show the customer’s words and identify the inference.
CRM records and lost deals
Review the product, deal stage, payment, decision time and reason for loss. A salesperson’s note that something was “too expensive” could mean a budget shortfall, unclear scope or a comparison with a different service. Include purchases, lost deals, cancellations and repeat enquiries. Keep records from different products, periods and markets distinct when the conditions explain a difference.
Website analytics
Compare sources, devices, pages and configured events. A visit ending on the pricing page does not establish why someone left. In Russia, Yandex Metrica’s long-term interests report (Russian) estimates interests from online behaviour, and its coverage can differ from other reports. For UAE projects using GA4, begin with observed acquisition and site events and follow up on the reasons separately. Website visitors represent only the people you have already reached.
Search terms
Terms such as “English for a job interview”, “English for work” and “online conversation practice” can help frame research questions. For a Russian-market study, use Yandex Wordstat (Russian), recording region and device filters. For the UAE, use Google Keyword Planner with the relevant locations and languages. Search estimates measure activity on the selected platform. They do not give you the number of potential buyers.
Reviews and public discussion
Save the review and its URL. The same complaint about scheduling can mean different things for a shift worker and a parent. For Russian-market research, monitoring tools such as Brand Analytics can help locate public discussions. Reviewers may differ from other customers, and repeated posts may distort apparent frequency.
Match the method to the question
- Interviews: reconstruct the circumstances and sequence of a decision. Compare recollections with records where possible.
- Surveys: estimate how common a known pattern is in a defined group. Include meaningful answer options and document who was invited.
- Observation and usability tests: see whether people can choose a course, find rescheduling terms or book a lesson. Our guide to product UX analysis explains how to investigate obstacles on a site.
- 5W and Jobs to Be Done: organise your evidence around who, what, why, when and where, or around the situation, desired outcome and alternatives. A completed framework still needs evidence.
Combine methods around one decision. Review enquiries, interview people about unclear refusals and then test the proposed change. The level of evidence needed depends on the cost of being wrong.
How to prepare and run an interview
Recruit actual or likely users with experience relevant to the question. For lost deals, find people who considered the offer and declined it. For repeat purchases, speak to people who bought again. Screen for what they were choosing, when and what happened.
Follow the sequence of a recent decision. Ask what they did before searching, where they found alternatives, how they compared them, what they ruled out, who helped decide or pay and what happened after the purchase. The GOV.UK interview guide covers open questions and follow-up.
If someone says they wanted a good teacher, ask how they judged that before paying. To investigate scheduling, ask what happened when plans changed. Record events and exact words separately from your explanation.
Segmentation and audience profiles
Group people by the differences that affect the offer or the way you reach them. Preparing for one interview and practising for recurring meetings are different jobs, even if both groups need flexible lessons.
Interview preparation: a specific event, a rehearsal and feedback. Workplace conversation: repeated practice for meetings. Shared requirement: scheduling may matter to either group. Check its importance separately.
A working profile can record the trigger, desired outcome, alternatives, criteria, constraints, who decides, evidence and unknowns. Add demographics only when the research supports their relevance. In Serenity’s Evrostroy project in Russia, the team used buyers’ motivations and barriers to plan content sequences for different properties. Reported outcomes reflect the full campaign.
Which segment should you test first?
Compare your confidence in the need, your ability to deliver and your access to the audience. If you have deal data, add acquisition and servicing costs. Interview frequency cannot establish profitability. Check whether the school has suitable teachers before testing a separate interview preparation programme.
Worked example: research for an English-language school
These four fictional records illustrate the method. They are not research findings.
- I01, bought: “I have an interview in a month. I wanted to practise answering questions about my work experience.”
- I02, bought: “I understand colleagues in meetings, but it takes me a long time to put my replies together. I need regular speaking practice.”
- I03, declined: “I wanted interview preparation and was offered a general course lasting several months.”
- I04, stopped lessons: “I work shifts. When my rota changes, I cannot reschedule in time under the school’s rules.”
I01 and I03 share an interview-related goal. Their different outcomes suggest investigating the offer and how it was explained. Only I01 specifies a one-month deadline. I02 describes recurring meetings. I04 raises a scheduling constraint and gives no reason for studying English.
Evidence: I01 and I03 mention interviews. Finding: there is a shared job with different purchase outcomes. Next question: what kind of preparation did each person expect?
A profile for interview preparation might record a specific outcome, the two source IDs, the one known deadline and questions about interview format, current level, teacher choice, workload and budget. Test whether the school can deliver a rehearsal with feedback. Four records cannot tell you how big or profitable the segment is.
Using AI to analyse audience evidence
Use AI for an initial pass over transcripts and enquiries. Remove unnecessary personal or commercial details, check whether you can share the material with the tool and keep internal source IDs. Check negations, sums, dates and names in transcripts before uploading them.
Use only records I01–I04. Extract the trigger, goal, deadline, choice criteria, barrier and outcome for each record. Cite its ID and exact supporting words for each field. Mark missing information as unknown. Suggest groups by buying situation and list differences within each group. Keep proposed marketing changes separate from the evidence.
Check each claim against its source and look for counterexamples or omitted evidence. The four records do not support a claim that everyone is an employed professional seeking higher income. I04 cannot be assigned to the interview group. Pilot the coding on a small batch before using it across the dataset. Count themes by a defined unit, such as participant, and report the limits of the sample.
Use the findings in marketing
- Offer and website: test a page explaining who the interview programme is for, how the rehearsal works, what feedback learners get and what affects preparation time. Avoid promising a job outcome.
- Advertising: test a message about the specific need and send it to the matching offer. Evaluate enquiry quality and enrolment, alongside clicks. Work out separately how to reach the segment through paid social.
- Terms: review rescheduling records and check whether a different rule would work for learners and teachers. One observation is enough to ask a question. Larger changes need more evidence.
When findings affect several offers, use them in your marketing strategy. Keep observations, interpretations and actions separate so that the team can trace decisions back to evidence. The GOV.UK guide to analysing a research session describes a way to do this. To turn recurring customer questions into publications, use our monthly content plan guide and template.
Questions about a first study
How many interviews should I run?
The GOV.UK planning guide suggests four to eight participants for many qualitative research rounds. Use that as a starting point for a focused question. Add relevant participants as new situations emerge. A few interviews can uncover reasons and hypotheses. They cannot estimate a segment’s share of the market.
What if we have no customers yet?
Talk to people who already face the problem and use a competitor, solve it themselves or put it off. Ask about their actual experience. Test willingness to pay with a concrete offer and clear terms.
When should we revisit the research?
Return to it when the product, terms, market or mix of enquiries changes. Identify which previous findings may no longer hold and collect the evidence needed to update them.
If you need help turning audience and competitor evidence into a decision, contact Serenity to discuss your project.
