How to brief an AR agency, and what to ask before you sign
Most AR briefs specify an idea and leave out the four things that decide whether it works. This is what to put in, what to ask, and which answers should make you slow down. It includes the questions that are uncomfortable for us to be asked, because a comparison guide that only flatters its author is not worth reading.
What belongs in an AR brief?
Start with the decision you want the experience to unblock, not the experience. "Customers cannot tell whether this sofa fits their room" is a brief. "We want an AR activation" is a budget line looking for a purpose, and it produces work that demos well and converts nothing.
Then state where the customer will be standing when they use it. In an aisle, at a stand, on a sofa at home, holding your packaging. That single fact rules formats in and out faster than anything else, because it determines how many seconds of patience you have and what the lighting and connection will be like.
Name the devices you are willing to lose. Every AR project has a floor, and if you do not set it, the agency will set it at whatever their test phone is. Deciding in advance that a four-year-old mid-range Android must work changes the build; discovering it on launch week changes the invoice.
Finally, say how it will be judged before anyone designs it. A brief without a success metric gets measured on total scans afterwards, which is the metric that flatters everyone and teaches you nothing.
Which questions separate a build partner from a showreel?
A showreel proves someone can make something look good in a controlled shot. It says nothing about whether the experience survives a busy hall on congested wifi. These are the questions that reach the difference.
| Ask this | Why it matters |
|---|---|
| Which device did you test on, and what frame rate did you hold? | A demo shot on a current flagship has not answered the performance question at all. |
| Will this need an app install? | The install prompt is the largest single drop-off in AR. If the answer is yes, ask what participation rate they expect and how they know. |
| What happens when tracking fails or camera access is denied? | There should be a designed fallback. No answer means visitors will hit a dead end in front of an audience. |
| Who prepares the 3D assets, and what do you need from us? | Asset preparation is usually the critical path and the most common cause of a slipped launch. Vagueness here predicts delay. |
| What will this cost to run, not just to build? | Server-side inference and hosting are per-session costs, and campaign traffic is spiky. A build-only quote is an incomplete quote. |
| What will you instrument, and can we see the data? | If the answer is scan counts, the campaign is being described rather than measured. |
| What licences cover the models and libraries you will use? | Some widely used vision models carry copyleft or non-commercial terms that create obligations for you, not the agency. |
| Who owns the source and the assets afterwards? | Agree it before, not during the handover conversation. |
Which answers should worry you?
"It will work on all modern devices." No AR experience works uniformly across devices, and an agency that says otherwise has either not tested widely or is managing you rather than informing you. The useful answer names a floor and what degrades below it.
A quote with no line for asset production, when you have not supplied usable 3D models. Either they have not understood the scope or they intend to raise it later.
Engagement statistics with no source. Industry figures for AR uplift circulate widely, are rarely traceable to a methodology, and are frequently vendor marketing repeated until it sounds like fact. An agency quoting them at you will quote them about you afterwards.
Reluctance to discuss what the experience will not do. Every honest scope has exclusions, and an agency that agrees to everything in a first meeting is deferring the disagreement to the point where it costs you money.
Should you use one agency or two?
Creative and technical AR are different disciplines and are often best bought separately. Creative agencies own the campaign idea, the brand relationship and the media plan. Technical partners build the thing and make it perform on real devices. Plenty of AR work is delivered by a creative agency subcontracting the build, which is a perfectly good arrangement as long as you know it is happening.
Where it goes wrong is when nobody owns the performance question. If the creative agency signs off a storyboard the technical partner was never asked to size, the compromises surface after the concept is approved and the budget is committed.
If you already have a creative or media agency, the useful ask is for the technical partner to review the storyboard before it is locked. It costs a meeting and saves the round of cuts that otherwise happens in build.
What does a realistic timeline look like?
Four to six weeks is a fair planning range for a focused single-product experience where usable 3D assets already exist. Multi-product catalogues and configurators run longer, driven by asset volume rather than by AR complexity.
The variable that moves the date most is what product files you have. Clean CAD shortens everything. Photographs of a product that has never been modelled means commissioning 3D work first, and that is measured in weeks of its own.
Work backwards from the event or launch with at least two weeks of contingency for device testing. That fortnight is the first thing sacrificed under pressure and the reason activations underperform on the day.
Common questions
- What should an AR brief include?
- The customer decision the experience is meant to unblock, where the customer will physically be when they use it, the oldest device that must work, and how success will be measured. An idea without those four things cannot be quoted accurately, and it will be measured on scan counts afterwards for want of anything better.
- How do I compare AR agency quotes fairly?
- Normalise three things first: whether 3D asset production is included, whether the price covers running costs as well as the build, and whether an app install is required. Quotes that differ on any of those are not comparable, and the install question alone can change participation by an order of magnitude.
- Should I hire a creative agency or a technical AR partner?
- Both, usually, and often separately. Creative agencies own the idea and the brand relationship; technical partners make it perform on real devices. The failure mode is a storyboard signed off before anyone technical sized it, so the most valuable step is a technical review before the concept is locked.
- What questions do agencies dislike being asked?
- Which specific device they tested on and what frame rate they held, what the experience costs to run rather than to build, and what licences cover the models and libraries they intend to use. All three are reasonable, all three have concrete answers, and evasiveness on any of them is itself the answer.
