If you already run a data science team and a wall of real-time dashboards, the case for paying an outside consumer insights agency has to be specific.
Dashboards are good at telling you what customers did, how often, and at what value. But they are poor at telling you why, and they quietly carry the assumptions of whoever built them.
This gap is where external research earns its fee, but only when it is pointed at one decision and judged on whether that decision changed.
The same test exposes the dashboards. In a 2022 Gartner survey of 377 marketing analytics users, analytics influenced just over half (53%) of marketing decisions.
This article sets out when to hire, which research model to buy, how to measure the return, and how to make the partnership work with the team you already have.
Write down the single decision you want research to inform, then check it against these five points.
A consumer insights agency researches why customers behave as they do, using qualitative, behavioral, and survey work, and turns the answer into a recommendation tied to a business decision. It's the question your reporting cannot answer. The gap persists even in well-instrumented companies.
Only 37.3% of senior data and AI leaders at Fortune 1000 and leading global companies say they have created a data- and AI-driven organization, according to a 2025 executive benchmark survey. The reasons are worth understanding before you buy anything.
A useful definition comes from a 2024 framework in the Journal of the Academy of Marketing Science, which describes customer insight for innovation as a reasoned understanding of customers' attitudes and behaviors that firms can innovate upon.
Two terms in that definition do the work. Reasoned means the understanding is built deliberately, from evidence. Innovate upon means it is tied to something the business can change.
A metric, by contrast, records an outcome. Repeat purchase fell, basket size rose, share moved to a competitor.
Those are facts about behavior. They carry no information about the attitudes that produced them.
A dashboard can show share moving to a competitor. It cannot say whether price, assortment, delivery speed or a loss of trust drove the switch. That residual question is the one an agency is hired to answer.
Every dashboard encodes the questions someone chose to ask when it was built. The metrics on it, the segments it splits by, and the targets it colors red or green all reflect a theory of the business held at that moment. When a number falls, the people reading it interpret the fall through the same theory.
If the dashboard was built around a belief that price drives switching, the segments will be price tiers and the explanation offered will be about price. The data is accurate, but the frame is inherited.
This is why adding a dashboard rarely resolves an argument. Each side can point to a chart that fits its existing position, because the charts were designed by people who already held those positions.
The usual name for the result is dashboard fatigue: more views of the same behavior, the same debate in the same meeting, and no decision that would not have been made anyway.
Gartner's 2023 survey of 303 sales leaders found the same pattern: 84% said sales analytics had less influence on sales performance than leadership expected.
The audit worth running is simple:
An outside team starts without the inherited frame. It does not know which executive sponsored the current plan or which hypothesis is politically safe.
Inside views of the customer drift, too: in four studies with 480 experienced marketing managers, those who tried to take the consumer's perspective projected their own preferences onto customers and became less likely to use market research results.
An external team can be captured by a brief just as an internal one can.
The practical value is narrower and more reliable: distance makes it legitimate to ask the question nobody inside wants to own.
These questions reframe the problem, and the reframing is what you are paying for.
Hire a market research firm when the question is settled, and you need a number. Hire a consumer insights agency when the question itself is contested.
Classify your pending research need into one of three models before approaching any vendor. Treat the three models below as different purchases.
The legacy service menu covers focus groups, central location tests, concept and pack tests, usage and attitude surveys, and advertising tracking. These methods are mature and defensible, and they are the right buy when the question is already agreed and what you need is a number with a confidence interval.
A CPG team choosing between two pack copy options needs a quantitative concept test. Where this model stops is at the handover. The firm delivers findings; it does not own the decision or the argument about what the findings mean for the plan.
A consumer insights agency, in the sense this article uses, takes responsibility for the reasoning as well as the fieldwork. Its output is a recommendation attached to a specific decision, built from whichever mix of qualitative, behavioral and quantitative work fits the question.
The same CPG team deciding whether to enter an adjacent category needs this model: unmet needs, the motivations behind current substitutes, and a view on whether the brand has permission to play. It costs more per project and takes longer, and its defensibility rests mostly on the quality of the reasoning.
Agile platforms and continuous communities trade depth for cadence. They answer bounded questions in days and suit iterative creative or product development, where many small decisions accumulate.
Their limit is scope. A fast read on which of three headlines lands tells you nothing about whether the campaign should exist.
| Model | Question it answers best | Typical output | Speed and cadence | Who owns the decision |
|---|---|---|---|---|
| Traditional market research | How many, which option, how much | Data tables, tracking reports, test scores | Weeks to months, project by project | The client team, after the report lands |
| Strategic consumer insights | Why, what is unmet, should we | Reasoned recommendation tied to a named decision | Months, bounded engagement | Shared; the agency argues for a choice |
| Agile and continuous research | Which of these, right now | Rapid reads, iterative feedback | Days, ongoing | The client team, decision by decision |
The comparison reflects the GWI distinction between research types linked above and general industry practice; the evidence set does not contain pricing for any model.
The short rule: buy data tables when the question is settled, and you need a number. Buy insight when the question itself is contested.
Build a one-page decision ledger with your agency before kickoff, and schedule its review for the date the decision's outcome becomes observable. That single habit separates research that is defended on evidence from research that is defended on anecdote.
The baseline is not encouraging. In Gartner's 2023 survey of 566 data and analytics leaders, only 44% said their team was effective in providing value to the organization.
Research budgets face the same scrutiny, and a ledger is how they survive it.
Functions that keep a ledger can talk about their contribution in the language the CFO uses.
Agree on five ledger fields before the first interview:
Write the default option down honestly, because it is the counterfactual against which the return is measured, and memory of it will drift once the study lands.
| Decision at stake | Default without research | Choice after research | Expected value difference | Observation date and owner |
|---|---|---|---|---|
| Launch a product line extension | Launch in Q2 with planned media | Shelve; need already met by existing range | Forecast launch and media cost not spent | Planned launch date; head of category |
| Reposition the premium tier | Keep current positioning | Rewrite around a motivation customers named | Projected lift in tier mix over 12 months | 12 months after relaunch; head of brand |
| Reverse a price increase | Hold the increase | Hold, with changed messaging | Avoided margin loss from reversal | Two quarters; pricing lead |
The table is a hypothetical, constructed to show the ledger's structure. The entries are illustrative and do not represent measured results from any client or study.
Most ledgers undercount because they only record upside. Research that stops a launch, kills a campaign, or prevents a price reversal produces a return that never appears in revenue. Record it anyway, conservatively.
These wins matter because launches fail often: a review of empirical studies in the Journal of Product Innovation Management puts new product failure rates at up to 40%.
The method is to take the forecast cost of the action not taken, including media, production and the direct cost of the project, and book it at the date the action would have occurred. Do not add estimated reputational damage or hypothetical lost sales unless the business already models them; the ledger is more credible when it is smaller and unarguable.
In the hypothetical table above, the shelved extension is the clearest entry: the return is a budget line that stayed in the bank.
The moment a research deck is presented is the moment it feels most valuable and proves least. Satisfaction with a readout tells you about the quality of the presentation. The return only exists when the decision has had time to work.
Set the review for the observation date in the ledger, bring the analytics team to confirm what the behavioral data now shows, and record what actually happened against what was expected.
For brand decisions, a brand lift study can supply the observed outcome. Some entries will be wrong. The ledger is still the honest record, and a run of entries that were right is the only argument that survives a budget review.
Procurement cuts what cannot show its value. A ledger with dated, observed entries is the difference between a discretionary study and a standing capability.
Run the scorecard below across your shortlist, and reject any agency that cannot name the downstream decision its findings will change. Most vendor pages list methods and panel sizes. Neither tells you whether the team will improve a decision.
Prioritize commercial judgment. Technical skill in fieldwork and analysis is widely available; the rarer skill is knowing which questions will change a business decision.
The question to ask in the pitch is what the agency would cut from your brief. A team that removes a question because its answer would not change any decision is showing exactly that judgment. A team that adds three methods to the scope is selling capacity.
Send the shortlist your current brief and ask each agency to critique it before proposing anything.
Watch for three behaviors:
Agencies that position themselves on challenging assumptions will claim this posture; the pitch is where you find out whether it is real. The reframing has to be substantive. Renaming your objectives is not the same as telling you the question is wrong.
Method fit is the second half of this test. An agency that proposes the same approach for a pack test, a repositioning, and a category entry is matching the work to its own capacity. Ask it to explain why the proposed method suits this decision and what a different method would have missed.
Research that stays in a slide deck has a return of zero. Before signing, require the agency to specify who in creative, media and product will receive the findings, in what form, and which decision each handover is attached to.
A creative team needs the motivation stated in language it can build on, with the customer's own words preserved; a long appendix of segment tables does not help a brand creative team decide what to say. A media team needs to know which audiences carry the motivation and where they can be reached. A product team needs the unmet need and the evidence for its size.
Ask the agency to show you a past handover, with the client's permission, and look for a decision at the end of it.
| Criterion | What to ask in the pitch | Strong answer | Warning sign |
|---|---|---|---|
| Commercial judgment | What would you cut from our brief? | Removes a question whose answer changes nothing | Adds scope and methods |
| Willingness to challenge | What assumption does this brief protect? | Names it and proposes a test | Praises the brief and quotes it back |
| Method fit | Why this method for this decision? | Explains the trade-off against alternatives | Same method as every case study |
| Handover design | Who receives the findings and for which decision? | Names roles, format and decision for each team | A readout meeting and a PDF |
| Working with internal data | What do you need from our analytics team first? | Asks for trends and current hypotheses | Prefers to start clean |
The scorecard is a framework for the reader to weight.
Weight the rows for your situation. A contested repositioning should overweight willingness to challenge. An innovation pipeline with many small decisions should overweight handover design.
Draft a one-page operating agreement before the engagement starts. It should name the question owner, the data to be shared, the readout format, and the protocol for handling conflicting findings. The agency is a complement to internal analytics, and the agreement is what stops it becoming a duplicate.
The agency's contribution is breadth across categories and distance from internal politics, the argument vendors such as Relish make for external partners. It is not a replacement for the internal team's knowledge of the business, and an agency that behaves as if it were should be corrected early.
Give the agency the trends and the existing hypotheses before it designs anything. The Journal of the Academy of Marketing Science definition cited earlier calls for reasoned understanding, and reasoning starts from what is already known.
If analytics shows a 12-month decline in repeat purchase concentrated in one cohort, the agency should be interviewing lapsed buyers from that cohort. A general study of brand perception would spread the effort too thin.
Sharing the hypotheses matters as much as sharing the data. The agency's job is to test them, and it cannot test what it has not seen.
The readout should be presented jointly. The agency reports what people said and why; analytics confirms whether those stated reasons are consistent with observed behavior and flags where they conflict.
In the hypothetical repeat purchase case, each reason lapsed buyers gave would be mapped back to the dashboard cohorts to see where the stories match the numbers and where they do not. Both outcomes are useful.
Agree on the conflict protocol in advance. When an external finding contradicts an internal view, the worst place for that to surface is a leadership meeting. The protocol should say who sees the finding first, how long analytics has to respond, and how the disagreement is presented: as two readings of one question, with the evidence for each.
Fieldtrip is organized to act on what external research finds. The group brings together specialist teams in strategy, creative, media and measurement, each owning its execution, working to a shared strategy. The point of that structure for consumer insight is the route it gives a finding.
When research explains why customers behave as they do, that explanation has somewhere to go: into the brief the creative team works from, the audiences the media team buys and the experiments the measurement team designs to check whether the change paid.
This describes how the group is set up; the structure does not produce better results on its own. Engagements can begin with one defined problem and sit alongside the internal analytics team and agency partners you already have; scope should grow only when the work has shown its value.
The next useful step is the same, whoever you work with. Pick the decision where your dashboards have stopped changing your mind, write a bounded question against it, open the ledger, and choose a partner who will argue with the brief before executing it.
If your dashboards describe what happened but no longer change what you decide next, Fieldtrip's strategy team runs qualitative and quantitative research and works with internal teams on product, go-to-market and positioning decisions.
It depends on what was studied. Underlying motivations, such as why a category exists in someone's life, tend to shift slowly, while responses to specific offers, prices, and messages move with the market.
Internal analytics can usually track behavior longitudinally; an ad hoc motivational study should be refreshed when dashboards show behavior diverging from the story the study told. The evidence here does not support a fixed shelf life, and any agency that quotes one should be asked how it knows.
Treat the conflict as information. First check whether the two are measuring the same people and the same period, since many apparent contradictions are definitional. If the conflict survives that check, follow the conflict protocol set out above, and have the decision owner record the choice in the ledger so the outcome can be judged later.
Research reduces the risk of a decision; it does not remove it. The ledger is the safeguard, because it records what was expected and when the outcome could be observed, so a wrong call is identified at a known date.
Review why it was wrong: a flawed method, a sample that missed the relevant customers, or a market that changed after fieldwork. Each has a different fix, and only the first is the agency's fault alone.
Start with a fixed-scope project tied to one decision. It forces a sharp brief, produces a clean ledger entry, and lets you judge the partner on observable work. A retainer makes sense once several projects have shown a return and the business has a steady flow of contested questions; before that point, it tends to pay for availability.
