Build audience segments around a decision your organization needs to make, then use evidence about people’s situations, needs, behavior, and desired outcomes to define groups that would benefit from different treatment. Demographics are not inherently irrelevant: use them when they affect access, eligibility, language, geography, or a measured decision. The mistake is treating a shared demographic label as proof of a shared motivation.
Start with the decision, not a demographic column
Before gathering data, write down what a segment will change: a content pathway, product feature, message, service, or channel. If no decision would change, the proposed segmentation may be descriptive rather than useful. Salesforce recommends setting a project goal and assessing whether the work is relevant, timely, feasible, and measurable; it also advises comparing the results with the wider audience. Salesforce’s market segmentation guide offers a useful framing.
For a personal-finance publisher or service, a decision might be whether to create different guidance for someone comparing ways to pay down debt and someone trying to understand an unexpected bank fee. Those situations could call for different information even if the people involved share an age range or income band. Start with the action you need to improve, not with whichever demographic field is easiest to export.
How do you figure out your audience’s problems?
Write a testable hypothesis
Describe a situation and a need you expect to find before inventing a persona. For example: “People who have just received a collection notice may need a clear explanation of what to check and what steps are available.” That is a hypothesis to investigate, not a claim about everyone who receives such a notice. Google News Initiative recommends developing an audience hypothesis and testing it through user research. Its audience-focused journalism playbook frames the work around learning about people’s needs.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
#1 Best Overall
Ask about real, recent behavior
Ask people to recall a specific occasion rather than inviting broad speculation about what “people like them” do. Useful questions include:
- What were you trying to do?
- What prompted you to do it then?
- What did you try?
- What got in the way?
- What would a good outcome have looked like?
A job story can capture the answer in a compact form: “When [situation], I want to [need or action], so that [outcome].” The situation and outcome help distinguish the underlying motivation from the task itself. NSW Government describes a similar approach: “We ask questions about why the user came to the site today, what problem they are trying to solve, and what is the job to be done.” The OneCX Program’s user research page explains its focus.
Rank #2
- Ideal for Gifting
- Ideal for a bookworm
- Compact for travelling
Combine what people say with what they do
Interviews and open-ended survey answers can reveal reasons, constraints, and desired outcomes. Site or app analytics, content choices, customer feedback, reviews, and social interactions can show what people did in a particular context. Salesforce lists these as possible segmentation inputs, while Google Ad Manager documents first-party audience segments based on site or app use and the content people see. Google Ad Manager’s first-party audience segment documentation describes that product capability.
Neither source of evidence should be asked to do the other’s job. A page view or sequence of clicks records an action; on its own, it does not establish a durable motive. A person may read about refinancing because they are considering a change, helping someone else, or simply trying to understand a term. Use qualitative research to explore why, behavioral evidence to observe actions, and compare the two without assuming they mean the same thing.
Recommended Free Tools
Google News Initiative suggests starting with 10 to 20 interviews to test assumptions when recruiting relevant potential users. This is a practical starting suggestion from its playbook, not a guarantee of statistical representativeness or a universal rule for when research is complete. The playbook also gives a 2 to 6 percent active-user goal as an example for a news business, subject to market size, marketing budget, competition, and organic growth; that figure is not a general conversion benchmark or a segmentation target.
Group people only when the difference changes what you do
Compare the situations, needs, barriers, and outcomes you have observed. Ask whether two possible groups would need different information, product support, messages, or channels. If the answer is no, keep them together rather than multiplying segments without a practical reason.
Rank #4
Assess candidate groups against the decision they are meant to support. For campaign activation, a useful group also needs to be reachable through an appropriate channel, while the cost of misclassification and the sensitivity of the data need to be considered. These checks are not a substitute for evidence: they help establish whether an observed difference is clear and actionable enough to use.
- Decision relevance: Would a distinct response help people achieve the outcome they want?
- Evidence quality: Is the group supported by more than an assumption or one isolated signal?
- Assignment clarity: Can the membership rule be applied consistently?
- Reach and channel fit: Can the intended group be reached in a suitable way?
- Misclassification, privacy, and cost: What errors or data risks might result, and is serving the group worth the effort?
Define membership so another person can apply the rule
For each segment, record the qualifying evidence, relevant time window, exclusions, data source, confidence, and owner. A label such as “people interested in saving” is difficult to review: it does not say what someone did, when they did it, or what evidence establishes membership. A rule tied to a specific observed event or research finding is easier to inspect and revise.
Free tools Windows power users keep installed
One-click scans. No signup required.
Best Value
- It can be a gift option
- Comes with secure packaging
- Helpful in various ways
Google Ad Manager gives “Restaurant Enthusiasts” as an example of a segment defined by two page views or app activities within three days. That is a platform example of an operational rule, not a universal threshold and not proof of a person’s motivation. The same distinction matters when building a finance-related segment: an observable action can qualify someone for a test or pathway, but it should not be presented as certainty about why they acted.
Technology availability also varies. The cited Google Ad Manager Audience Solutions feature is limited to Ad Manager 360, so it should not be treated as a capability available to every publisher. The documentation recommends moving from third-party cookies to PPID for greater coverage and durability; that product recommendation does not remove the need to govern data appropriately or establish that a particular segment is valid.
Use demographic and contextual attributes when they matter
Demographic or contextual information can be relevant when it demonstrably changes access or a decision—for example, eligibility rules, language, geography, or a measured outcome. Treat those attributes as evidence about the specific constraint, not as a shortcut for inferring a person’s needs or preferences. The IAB Tech Lab’s Audience Taxonomy distinguishes demographic, interest-based, and purchase-intent labels to improve clarity and comparability across providers. Its version 1.1 followed a 30-day public comment period in October 2020; that history describes a standards release, not proof that one way of segmenting audiences performs better. The IAB Tech Lab Audience Taxonomy page provides the standard’s context.
Validate, challenge, and revise the segments
Check whether each group behaves or achieves outcomes differently in a way relevant to the decision. Compare results with a sensible baseline, and actively look for cases that contradict the rule. If people assigned to two segments respond similarly, or the evidence does not support the distinction, reconsider whether they need separate treatment.
Keep names focused on a need or situation rather than an identity. Personas can help a team remember a set of findings, but a composite character is not evidence that every member shares the same story. Google News Initiative notes that personas may evolve after interviews and can become less useful than jobs-to-be-done. NSW Government says its first research callout produced over 70 personas and that its audience does not fit neatly into groupings; the program’s website saw more than 14 million users over the first six months of the COVID-19 pandemic. Those are facts about that program and period, not universal segment-size guidance. Revisit definitions when evidence or the decision changes.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
