Planning a POP Campaign With an AI Agent: Mix, Materials and Budget per Store

A brand marketing manager gets a number, a store count and a date. Say 300,000 €, 180 stores, in store by November. Everything else is undecided: what gets made, in what material, how many of each, and which stores get which pieces. The way that usually gets decided is by describing the launch to an agency or a manufacturer and waiting a week or two to see what comes back.

One thing to clear up first, because search engines still get it wrong. POP here means point of purchase, the physical materials that sell product inside a store: floor stands, glorifiers, shelf strips, stoppers. Not pop-up shops, and not website popups.

I have spent close to four years in point-of-purchase, and campaign planning is the step with the least published help of any part of this job. This post is about what planning actually decides, and how Bellto, an AI agent for POP campaigns that takes a launch from the brief to shop drawings, works through it.

What does planning a POP campaign actually decide?

Four things, in a rough order, and each one narrows the next.

The mix. Which formats the campaign uses. A launch is almost never one display. It is a floor stand doing presence, a glorifier giving one SKU hero treatment, a shelf strip and a stopper holding the position on the shelf where the category is really shopped.

The material tier. Cardboard, PET, acrylic, metal and wood are not interchangeable choices about looks. They are choices about how long the piece has to survive, which is also the biggest single lever on what it costs.

The quantities per store. Not every format goes to every store. The floor stand might go to two thirds of them and the shelf strip to all of them, and deciding that split is the actual planning work.

The spend per store. Which is the constraint the other three have to fit inside.

Most of what is published about POP costs answers a narrower question than this. GMS Industries, for instance, publishes a cost guide with per-unit ranges by display type, from around 5 dollars for a temporary countertop unit up to 1,000 dollars and beyond for permanent pieces. Those are real published figures and they are useful, but they price one display. They do not tell a brand with 180 stores what to make.

Why does the budget per store decide the mix?

Because dividing first changes the question you are answering.

The Bellto landing page carries an illustrative exchange that shows it, a premium cutlery range going into 180 stores in November on a 300,000 € budget. The first thing the agent does with those three numbers is divide them: that is 1,667 € per store. It is the landing page's illustration rather than a real customer, and the arithmetic is the point of it.

That one figure rules things in and out immediately. It says whether a permanent metal and wood floor stand is a candidate for the whole estate or only for part of it. It says whether the campaign can afford a glorifier at every counter. It converts an abstract 300,000 € into the only unit anyone in a store ever experiences, which is what that store gets.

Going the other way round, pricing formats one at a time and totalling them at the end, is how campaigns end up needing a round of cuts after the concepts have already been approved internally. If a campaign like that is on your desk right now, describing it on the Bellto waitlist is where it starts.

What does an agent need to know before it proposes anything?

More than those three numbers, and the useful behavior is asking rather than assuming.

In the same illustration, the agent does not propose a mix straight after the arithmetic. It comes back with two questions: is the corner 1.2 m or 2 m wide, and how many SKUs go in.

Both change the answer completely. Corner width decides whether a floor stand is physically possible in that retailer at all, and a mix built for 2 m that turns up at a 1.2 m corner is a campaign that gets refused at the store. SKU count decides how many facings the pieces have to carry, and facings decide shelf widths, which decide the size of everything and therefore the cost of everything.

That is the difference between an agent and a form. A form asks for whatever fields it has. An agent notices what is missing for the specific decision in front of it, and asks for that before it commits to anything.

Why don't generic AI campaign tools do this?

They were built for other channels, and they say so plainly.

HubSpot's Campaign Assistant takes a campaign description, an audience and a tone, and generates copy: landing pages, marketing emails, and ads for Google, Facebook and LinkedIn. Venngage's AI Campaign Generator walks through goal, campaign type and messaging and produces campaign design assets, layouts and infographics among them, exportable as PNG, PDF or PPTX.

Both do what they describe. Neither is pointed at a physical object. A brief that ends in a floor stand has to resolve into material, thickness, facings, dimensions and quantities, and none of that exists in a tool whose output is copy or a layout. Ask a general marketing AI to plan an in-store campaign and you get a plan written in the vocabulary of media: phases, channels, messages. What the campaign needs is a mix, a split and a piece.

How does a mix get reasoned rather than listed?

By trading coverage against presence, store by store.

The illustration on the landing page resolves into three lines: a floor stand in metal and wood for 120 stores, a counter glorifier in acrylic for all 180, and a shelf strip in PET for all 180, with price ranges per format rather than a quote. The quote itself comes from the manufacturer.

Look at what that split is doing. The expensive, permanent, high-presence piece goes to two thirds of the estate rather than all of it, because stretching it everywhere would force the material tier down in every store. The cheap piece that holds the shelf position goes everywhere, because a campaign that is invisible in the aisle in a third of its stores has a hole in it. That trade is the craft of campaign planning, and it only becomes visible once the budget has been divided per store. If you want the detail on what each format is for, we wrote it up in types of POP displays.

The reason a mix like that can be argued rather than simply presented is that in Bellto every line of it opens as its own design thread afterwards. You are not approving a slide. You are approving the starting point of a piece that then gets designed with you at real scale, with proportions derived from your product dimensions and facings, and the agent argues every decision with you before it moves on. If your launch is further along than the planning stage, the Bellto waitlist takes the campaign as it stands.

What happens to the plan after it is agreed?

This is where campaign planning normally leaks.

A mix agreed in a meeting is a slide. It carries no dimensions, so every manufacturer who receives it reads it differently and prices something slightly different, which is what makes comparing two quotes so unsatisfying. I went through that gap in from AI render to quotable display.

In Bellto the plan does not stop at the slide. Each piece in the mix is designed at real scale in parametric 3D, so a dimension change recomputes the whole piece and re-runs its manufacturing checks in seconds rather than costing an email and days of waiting. Every approved piece then comes out as five things: a 3D model, a part-by-part cutlist, dimensioned drawings, a STEP of the assembly and one cut DXF per manufactured part. That is a package a workshop can quote without redrawing it, and every model goes through engineering review before it is delivered.

On the worked piece published on the landing page, a four-panel acrylic counter glorifier of 200 × 200 × 250 mm, that came to 4 parts and 4 cut DXF files, with 0.0 mm³ of overlap measured across its 6 pairs of parts and a STEP that reimported with no solids lost. At the other end of the scale, a four-shelf floor stand of 1000 × 650 × 1820 mm with a header and casters runs to 241 parts and 250 cutlist rows, 89 manufactured and 152 bought in. Going from 16 to 20 facings recomputed the whole unit from the parametric geometry alone in 1 min 18 s, with the checks green afterwards. The approach behind that is written up in parametric POP display design.

Timing is the other reason to settle the plan early. Manufacturers publish lead times running from roughly 4 weeks for temporary corrugated work to 28 weeks or more for permanent custom pieces, and they start the clock at different points, some only from approved artwork. A mix that leans permanent is a different calendar from a mix of temporary pieces aimed at the same date, so the mix decision is a date decision too.

Where does the concept image fit?

Somewhere specific, and it is worth being precise about it. The concept image is what gets a look agreed internally and with the retailer, which is often the slowest political step in a campaign. That is the job AI POP Displays does, from a plain-language brief with your own product shots as references.

What an image cannot do is carry millimeters. Planning needs numbers that survive contact with a cutting table, which is why the mix, the facings and the dimensions are the things the agent argues with you about.

How do I plan a campaign this way?

Through the Bellto waitlist, and by describing the campaign rather than just leaving an email.

We are building Bellto now, and we are contacting the first brands soon to run the first real campaigns end to end. The list is open to any brand or agency, and POP manufacturers can sign up too, since what reaches them from Bellto already carries dimensions, a cutlist and a defined product. It opens by invitation, in small groups, and the campaign you describe when you sign up sets your place in the order, so a concrete launch with a retailer, a store count and a date moves faster than a general interest. Pricing goes first to the people on the list. There is no card.

If what you have today is a number, a store count and a date, that is already enough to start the conversation. The fuller reference on the agent itself is in what is Bellto.