An experiential quote is five businesses in one number. Buyers evaluate agencies on the first, pay for the second, and discover during the campaign that the third and fourth were what mattered.
Worth taking them apart, including the parts where an agency is genuinely irreplaceable and the parts where the margin is buying coordination you might already have.
The five things being bought
Concept and creative
The idea, the reason anybody stops walking, and how it connects to whatever the brand is trying to say. This is the hardest part to buy elsewhere and it is what the good agencies are actually selling.
Production
Building the thing. Fabrication, transport, install, power, teardown, and the hundred details that decide whether a stand looks expensive or improvised. Specialist work with real capital behind it.
Permits and logistics
Site agreements, council permissions, insurance, health and safety documentation, licences where samples involve food or alcohol. Unglamorous, occasionally the reason a campaign does not happen, and genuinely hard to do yourself in a city you do not operate in.
Staffing
The people who are actually there. Brand ambassadors, demonstrators, hosts, sampling teams. This is the largest variable cost on most multi-city campaigns and the part most frequently brokered to a third party rather than provided directly.
The fifth is measurement, which usually appears as a line item and functions as an afterthought. More on that below, because it is where the money leaks.
Agencies are hired for the idea and judged on the execution, and the execution is mostly the people they did not employ.
Where the money goes
| Component | Typical share of budget | Notes |
|---|---|---|
| Concept and creative | 10 – 20% | Fixed early, hard to compare between agencies |
| Production and build | 30 – 50% | Dominates single-flagship campaigns |
| Permits, logistics, insurance | 10 – 15% | Scales with the number of sites, not the size |
| Staffing | 20 – 40% | Dominates multi-city, multi-day campaigns |
| Measurement | 0 – 5% | The line most often cut to fund the others |
The shape flips depending on the campaign. One spectacular installation in a single city is a production project with some staffing attached. Forty sampling shifts across eleven cities is a staffing project with some production attached, and it is frequently priced as though it were the first kind.
Staffing rates commonly run $20 to $45 per hour per person before agency margin, higher for demonstrations needing product knowledge or a licence. Margin on top varies widely and is rarely itemised, which is a reasonable thing to ask about directly.
The staffing layer, and why quality varies
Most agencies do not employ the field staff. They work through staffing partners, who work through local agencies, who work through their own rosters. That chain is normal and it explains most of what goes wrong.
Each link takes margin, so the rate reaching the person on the floor is lower than the rate you paid, sometimes considerably. Below a certain point you are recruiting from a pool that has already priced the job as not worth preparing for.
Each link also loses brief fidelity. What you wrote, what the agency summarised, what the staffing partner briefed and what the person was told at the start of a shift are four documents, and only the last one affects anybody.
This is not agency incompetence. It is what happens when supervision is centralised and delivery is not, and it is exactly the failure that structured evidence exists to catch. How each layer of targeting narrows the pool and moves the rate covers why coverage in secondary cities costs what it does.
Writing a brief that survives the chain
Since brief fidelity degrades at every handoff, the practical response is to write the field-facing version yourself rather than trusting a summary of a summary.
It fits on one page and it is not the creative deck. The creative deck explains why; the field brief tells somebody standing in a shopping centre at nine on a Saturday what to do.
Four things belong on it. The three actions that must happen at every interaction, in order, phrased as instructions rather than objectives. The words that may and may not be said, particularly any claim about the product that carries a regulatory weight. The evidence to capture and when — before doors open, during, at teardown. And a named person to contact when something is wrong, with a real number, because the alternative is somebody improvising and nobody hearing about it.
What does not belong is brand strategy, audience personas, or the campaign's internal objectives. Every paragraph of context competes for attention with the instruction underneath it, and the instruction loses.
The same principle governs any distributed work: the specification that reaches the person doing the task is the only one that affects the outcome, and everything upstream of it is a hope about what they were told.
Measurement, honestly
The numbers reported after an activation are usually activity: footfall past the site, samples distributed, conversations had, content captured, hashtag mentions.
Those are worth collecting and none of them is an outcome. Footfall past a stand in a busy station measures the station. Samples distributed measures how quickly people accept free things.
The outcome questions are harder and answerable: did the people who interacted differ afterwards from comparable people who did not, and by how much. That needs a control group and a survey, which is a separate design and a separate budget line. What a brand lift study can establish and the sample size that claim needs sets out the arithmetic, and the short version is that a few hundred intercepts cannot support a small effect.
Capture at the moment
A short structured interaction record taken during the shift — what was demonstrated, what was asked, whether a code was issued. Recorded at the time, not reconstructed at the end of the day.
Follow up later
A survey to people who opted in, fielded within days rather than weeks, with the same questions asked of a comparable group who were not there.
Count what is countable
Codes redeemed, links followed, sign-ups from a site-specific route. Weak signals individually and useful because they are not self-reported.
Keep the activity numbers
Footfall and samples still matter for planning the next one. Just do not present them as evidence that anything changed.
Verification: the question nobody asks until it goes wrong
For a single-city activation the marketing lead was there. For eleven cities over three weekends, nobody was, and the reporting is a summary written by the party being paid.
That is not a trust problem so much as a design gap. The fix is specifying what evidence comes back from each site before anybody is booked: timestamped photographs of the setup before doors open, a completed checklist against the brief, counts logged during the shift, a photograph of the teardown. Defined in advance, returned per site, and checkable by somebody who was not there.
What counts as evidence and how to specify it before the work starts covers the formats. The same discipline underlies in-person retail and location checks, which is the mature version of this problem — an industry that worked out decades ago that a report without an artifact attached is a report about somebody's memory.
An activation you did not see, reported by the people you paid to run it, is a story rather than a measurement.
Where agencies are worth every penny
Being fair about this, because the answer is frequently yes.
A flagship build. Anything involving fabrication, structural safety, power, or a site that will be photographed — that is production work with real risk attached and it belongs with people who do it constantly.
Permits and licensing in cities you do not know. The time cost of learning one city's requirements exceeds the margin, and the failure mode is a campaign that cannot legally happen.
Anything on a compressed timeline. Agencies hold relationships with sites, fabricators and staffing partners that take years to build, and a campaign that has to be live in five weeks is buying those relationships.
And the concept itself, when it is good. A genuinely strong idea is rare and worth paying for, and no amount of operational efficiency substitutes for people walking past without stopping.
Where the model strains
Two situations, both about spread rather than quality.
Many cities, many days. Agency margin scales with staffing volume while the coordination work does not scale at the same rate, so the fortieth shift costs what the first did and needs considerably less thought. This is where buyers start asking what the margin is buying.
Secondary markets. Rosters are deep in the cities agencies operate in and thin elsewhere, which produces either a longer timeline, a substitution of the cities they can cover for the cities you asked for, or a chain of subcontractors long enough that nobody can tell you who will actually be there.
What to ask before signing
Six questions, and the answers describe an agency more accurately than a reel of previous work.
Who employs the field staff — you, a partner, or a partner's partner? Ask how many links there are.
What rate reaches the person on the floor, and what is the margin on staffing specifically?
What evidence comes back from each site, in what format, and is it defined before booking or after?
Who is physically present in each city, and what is their span of control? One supervisor across four cities on the same day is a number worth hearing out loud.
What happens when a site underperforms mid-campaign — who decides, how fast, and out of whose budget?
And how will we know whether it worked, in terms that are not footfall?
Where we fit, and where we do not
Plainly: we do not build activations, hold site relationships, or handle permits. Those are agency competences and there is no distributed version of a fabrication shop.
What distributes is the part where somebody has to be in a place, do a defined thing, and prove it. Coverage across cities where no roster exists, shift capacity that scales without a longer subcontracting chain, and structured evidence returned from every site rather than the ones somebody visited. For a single flagship that is not a useful offer. For eleven cities across three weekends it is the difference between knowing what happened and receiving a summary.
For the people doing that work, and what it actually pays, what brand ambassador work involves covers the supply side honestly. How an ambassador programme is designed so it produces something is the longer-term version of the same relationship, and how a brief becomes reserved capacity and verified evidence describes the mechanism end to end.