Turning a promotional pallet into a profitable data platform.
A Discovery track for Brambles' Retail Promotions, researching whether the quarter-pallet displays used by Walmart, Costco and major CPG brands like Guinness and P&G could tell brands and retailers something no one could see before: is this promotion actually on the shop floor, or lost in the back of the store? Two rounds of prototyping later, it's now in active customer testing.
What a promotional display actually is
The object at the centre of this whole case study: a branded free-standing display unit (FSDU), stacked with product and built to sit on top of a pallet so it can be delivered, wheeled into place and assembled in minutes. This is what “the pallet” and “the display” mean every time they come up below.
Project snapshot
A Discovery track that paired me closely with a lead UX researcher, moved through two rounds of prototyping, and is now the evidence base behind a product in active customer testing.
The challenge, in short
Brands spend over $500bn a year on trade promotions, yet once a promotional pallet leaves the warehouse, the store is a black box. Somewhere between 40–60% of displays are activated late or never reach the shop floor at all, lost in stockrooms, decanted, or removed early, and nobody finds out until the promotion is already over. Brambles' smart quarter-pallet displays, already inside thousands of stores, were the one asset with a genuine shot at closing that gap without asking a single retailer to install new infrastructure.
The store is the most data-poor environment in commerce, and we already had a pallet inside it.
My role, in short
I was lead designer on the trio driving this Discovery track, working most closely with Olga, our lead UX researcher, whose research underpins most of what follows. I didn't just design from her findings second-hand: I visited stores with her, sat in on user sessions, and personally interviewed stakeholders along the way, including store managers at Tesco, Asda and Sainsbury's and co-packers from Guinness. I designed and built both rounds of prototyping myself, once in Figma, once almost entirely in an AI-accelerated playground, with Olga running the research sessions around them.
Impact, in short
Real-time activation alerts, tested in trials, increased promotional selling time by roughly 80% across a promotion. Early placement-optimisation testing produced a 4.5x increase in display-level sales. The front-vs-back-of-store activation classification capability that came out of this Discovery track is now in active customer testing with brands and retailers, the first rung on a platform whose next horizon, weight/consumption sensing, is already on trial separately: not just whether a display reached the floor, but whether a bottle was actually lifted off it and converted into a sale.
The challenge
Sub-brands like Guinness and P&G pay retailers like Walmart, Costco, Tesco and Sainsbury's significant money for prominent in-store display positions. That's the commercial logic behind trade promotions: a well-placed display drives volume, defends shelf space against ever-growing own-label competition, and is one of the largest line items in many CPG marketing budgets, often bigger than media spend.
The problem is that once a promotional pallet leaves the warehouse, nobody can see what happens to it. Unlike ecommerce, where every click and conversion is tracked, the physical store generates almost no execution data. Brands know what they shipped, and what scanned at the till. Everything in between is a gap, and that gap is where promotions go wrong: pallets sit in stockrooms, get decanted onto shelves to protect availability, launch late, or are removed before the promotional window ends. Across target markets, somewhere between 40–60% of promotional pallets are activated late or not at all, and by the time anyone finds out, the window to fix it has already closed.
Brambles/CHEP already had an asset physically inside thousands of stores that no competitor could easily match: the pallet itself. Because CHEP recovers its pallets, smart sensing technology built into them could be deployed and retrieved at scale, across an enormous store footprint, without asking a single retailer to install new infrastructure or grant system access. The question this Discovery track set out to answer was whether that pallet, instrumented with sensors fusing movement, Wi-Fi and Bluetooth signals to work out where it actually was, could become a genuinely new, profitable data product for Brambles' Supply Chain Illumination platform, and not just a tracking feature nobody would pay for.
The store is the most data-poor environment in commerce, and we already had a pallet inside it.
“What are the additional costs of having an FSDU and not selling those products normally from the shelf? Probably 20K, 30K? Too much. What do I need to do to make it a bit more profitable? What products do I need to add to make it more profitable?”
“The financial impact is in terms of both, any compensation that would be given and also the sales and profit impact, because if something hasn't been implemented well, or not at all, then there's … something that we've missed out as a business.”
My role
I was the lead designer on the trio driving this Discovery track, working alongside a Product Lead and, most closely, Olga, our lead UX researcher.
The research that follows is mainly Olga's work, and I want to be straightforward about that. What I brought to it was staying close enough to be useful rather than reviewing findings after the fact: I personally visited stores with her, including the Asda store-manager walkthrough referenced in Chapter 01, and I sat in on the moderated user sessions across both rounds of prototyping, watching how people actually reacted to what I'd built rather than hearing about it secondhand. Alongside that, I ran interviews of my own with stakeholders along the way, including store managers at Tesco, Asda and Sainsbury's, and co-packers from Guinness, whose day-to-day reality of decanting, congestion and compliance checklists shaped what the product needed to detect and act on.
On the design side, I designed and built both rounds of prototyping myself: the first in Figma, tested with US brands like Green Garden; the second, this year, almost entirely inside an AI-accelerated playground, using the same VS Code and Claude-driven workflow covered on the AI Playground page, with far less reliance on static Figma files.
Understanding what actually happens on the shop floor came before any design work started.
Into the stores
Before designing anything, we needed to see, physically, what happens to a promotional display once it leaves the warehouse. Olga led that research; I visited stores and interviewed stakeholders alongside her.
Olga's research programme combined retail safaris across Tesco, Sainsbury's and Asda stores in the Birmingham area, a dedicated field visit and store-manager interview at an Asda supermarket in Coventry, and direct conversations with the people who actually execute promotions day to day: store managers, CPM merchandisers, and, on my side, co-packers from Guinness. I joined the store visits and the store-manager walkthrough in person, and ran my own interviews with retailer and supplier stakeholders alongside Olga's programme, rather than working from a synthesis deck after the fact.
What came back reframed the problem. Execution maturity varies a lot by retailer: Tesco's aisles were visibly more structured, run under strict congestion rules like a "five-tile rule" to keep walkways clear; Sainsbury's had more promotions live on the shop floor, helped by a barcode-scanning system that gives HQ visibility of activation, but that same volume came with lower compliance, mixed products and inconsistent execution; Asda's approach was availability-first, weaker on promotional hygiene, with staff focused on getting stock onto shelves rather than a specific display looking right. The underlying constraints were the same everywhere, though: limited backroom space, aisle congestion, theft risk, and store teams working from fragmented systems with no single view of everything running in their store at once.
Store managers think like traders, not merchandisers
Store managers are measured on store profit above almost everything else. Their instinct is to move stock onto the shop floor as fast as possible to protect sales and availability, whether or not it lands in the right promotional position. Getting something onto the floor beats getting the right thing onto the floor in the right place.
Decanting is a rational response, not a failure
When space is tight, congestion rules are strict, or theft risk is high, staff move promotional stock out of its intended display and onto ordinary shelves. It looks like non-compliance from a brand's dashboard. On the shop floor, it's the sensible trade-off available to the person actually managing the space.
Compliance is tracked on paper, and rarely leaves the store
Store managers walk the floor after set-up and record compliance on a physical checklist, kept mainly in case a senior visit asks for it. It almost never reaches HQ, and it never reaches the brand that paid for the placement.
Activation and compliance are two different things
A display can be "activated", present on the shop floor, while still being non-compliant: mixed products, the wrong location, degraded packaging. Any solution had to be able to tell those two states apart, not just report a single activation flag.
These weren't abstract personas. They were the specific people, store managers walking us through their own stockrooms, merchandisers explaining why a display that looked "activated" on paper was mixed and unpriced on the shelf, that the two rounds of prototyping that follow were designed and tested against.
Before designing anything, the trio mapped the whole ecosystem: every stage, every actor, and the physical store itself.
Mapping the ecosystem
Before any design work, we mapped the full retail promotions journey, every entity involved, and the physical store itself, so the product being designed matched reality rather than a plan.
We did this through service blueprints, mapping the activation and compliance journey stage by stage, capturing both the user-facing actions (a merchandiser scanning a display, a store manager walking the floor) and the backstage systems and processes behind them. Blueprinting the two together, rather than just the front-of-stage steps, is what surfaced friction points and opportunities a simple flow diagram would have missed.
Seven stages, four entities, one moment we chose to focus on first.
Manufacturer (M), Retailer (R), Co-Packer (CP) and Field Team (FT) touch different stages of the same journey. Activation & Compliance, stage five, is where this Discovery track and both rounds of prototyping focused.
Campaigns are conceived, volumes estimated, terms negotiated between brand and retailer.
The creative and technical development of the display itself.
Displays are physically built and loaded ahead of shipment to retail DCs.
Displays move from assembly through to individual stores.
Displays are set up, activated on the shop floor, and checked against plan. Our focus.
Display performance is tracked and restocking decisions are made.
Campaign results are analysed against the objectives set in stage one.
The location: the retail store
Stage five happens inside a physical space most of the team, and every brand paying for a display, never actually walks through. Mapping it mattered as much as mapping the process.
Splitting the store this way is what made the store-manager research in Chapter 01 make sense as a system rather than a list of anecdotes. Decanting happens at the boundary between these two halves. Compliance checklists get filled in on one side and rarely cross to the other. A pallet's whole value proposition, "is this display on the sales floor", only means something once you can point to where the sales floor actually starts.
The actors
Five roles are directly responsible for what happens to a display once it reaches a store. Every one of them shaped how the prototype needed to behave.
Store Manager
Accountable for overall store performance, with profit as the primary goal. Activation and compliance compete with a much longer list of KPIs.
Section Manager
Oversees day-to-day activity within one product category, making sure goods, including promotional stock, are on the shop floor and available.
Store Employee
Carries out the daily operational tasks that get a display from the stockroom onto the shelf, and back off it again once a promotion ends.
Merchandiser
Visits stores to verify a campaign has been activated to standard, documenting compliance with photos, on a scheduled itinerary they don't fully control.
Store Director
Oversees a group of stores. The senior visit store managers describe checking compliance for, more than any system does.
“You're really getting into the realms of adding value because you're not calling on a store where the display is already there and just taking a photograph. You're making sure that every store you go into, you're making a difference.”
The same underlying execution data means something different to each of them. A Store Manager needs to know, right now, whether a specific display in their store is live. A Regional Manager needs to compare stores to prioritise where to send limited field-team time. A Category Manager needs the same data rolled up across every campaign, every region, to decide what to plan next time.
| Role | What they need from the same data |
|---|---|
| Store Manager | Real-time, store-level: is this display active right now, so an issue can be resolved before it costs a sale. |
| Regional Manager | Weekly, region-level: which stores are falling behind, so field-team visits go where they'll actually change the outcome. |
| Category Manager | Monthly and quarterly, estate-wide: what drove performance across every campaign, to plan the next one better. |
Understanding the store reframed what the product actually needed to be: not a report, an intervention.
From tracking to a data product
The commercial case only became real once we reframed the pallet from "an asset we recover" to "a sensor already deployed at scale inside thousands of stores".
Brambles/CHEP's quarter-pallet promotional displays already carry a small stack of sensors: Bluetooth and Wi-Fi for indoor positioning, an accelerometer and magnetometer to detect when a pallet has actually moved and been reoriented rather than just nudged, a temperature sensor to confirm it's sitting in a customer-facing space rather than still in cold storage or transit, and cellular connectivity for coarse location when nothing more precise is available. GPS hardware is on board too, currently disabled to preserve battery life until it's needed. Fused into a single classification model, that's enough to answer the two most commercially critical questions in the whole promotional lifecycle: did this pallet arrive at the correct store, and did it get activated to the front of the store rather than left in the back room? That's the "Now" capability, real-time front-versus-back-of-store activation classification, and it's the one now in active customer testing. This Discovery track, and both rounds of prototyping, focused deliberately on that single moment: activation, not the full promotional lifecycle.
It's also just the entry point. The same pallet is a platform: every additional sensor opens a new category of action. Beyond activation detection sits a separate, later-stage trial already running on-device in stores: weight/consumption sensing, precise enough to know whether a bottle of Coca-Cola or a case of Guinness was actually lifted off the display and converted into a sale. That trial sits outside the scope of this Discovery track, but it's the next horizon on the same platform, alongside quality monitoring in transit, catching temperature, shock or humidity damage before a display is ever activated, aisle-level location precision, and a compliance score that rolls all of it into a single number brands, retailers and merchandisers can act on together.
Not "did the pallet arrive." Did it get to the front of the store in time to sell.
“[Retail promotion] campaigns are a pretty big investment for us. In the past, when these pallets got picked up from our warehouse, they went into a black hole. … We need end-to-end visibility, and if a pallet hasn't arrived, I want to see where it's stuck.”
Zero retailer friction, by design
Any solution requiring a retailer to install hardware or grant system access faces a slow, political approval process most retailers simply won't grant. Everything here works because the sensing lives on the pallet, not the store, which is exactly why Brambles can do this at a cost and scale nobody else can.
Real-time alerting, not a post-mortem
The value isn't a report confirming a display failed after the promotion ended. It's alerting a merchandiser or store team while there's still time to move a pallet from the stockroom to the floor, and trial data shows that intervention can meaningfully recover lost selling time.
increase in promotional selling time across an entire promotion, when real-time activation alerting was trialled against unexecuted displays.
increase in display-level sales from early testing of placement optimisation, using front-vs-back-of-store activation data rather than guesswork.
of promotional pallets activated late or not at all across target markets, the baseline problem the whole product exists to close.
Two rounds of prototyping turned that opportunity into something real enough to test with actual users.
Round 1: proving the concept with Green Garden
Last summer, before any of this was validated with UK stakeholders, the first prototype tested the core activation concept with US brands.
Before a single Figma screen existed, we storyboarded the concept specifically for Green Garden: pre-campaign set-up, the end-of-day delivery notification, store activation with photo evidence, and compliance monitoring, walked panel by panel through the same Operations Manager and Merchandiser roles the rest of this journey is built around. That storyboard is what got tested first, before we invested in building the higher-fidelity prototype.
The first round of prototyping, designed and built by me in Figma, tested concept activation and compliance visibility with US brands including Green Garden. Olga interviewed users through the sessions; I sat in on them, watching in real time how an Operations Manager reasoned through the key moments of the promotional journey, planning, delivery, activation and post-campaign analysis, and where a design decision landed or didn't.
That grounding mattered going into Round 2. It confirmed the activation moment specifically, not the full promotional lifecycle, was the sharpest wedge into the problem, and it surfaced the first version of the interaction patterns, status states, and alerting logic that Round 2 would rebuild almost entirely inside code rather than static screens.
Why Green Garden was the right test case
Green Garden ships its promotions on standard full-size pallets fitted with an early location-tracking device, and at the time, the brand cared about exactly one thing: had a display been delivered, activated, and completed on time. That's a precise match for what this Discovery track and both rounds of prototyping actually tested, activation and timeliness, not consumption. It's why Round 1 was built and tested against Green Garden's real promotional programme rather than a hypothetical one.
In parallel, and run separately from this Discovery, Brambles' physical product team was already piloting the next generation of hardware in the field: a quarter-pallet display fitted with a weight sensor, trialled in Spanish retail. That trial is aimed at the consumption/weight-sensing horizon on the roadmap, not the front-vs-back-of-store activation capability this Discovery track validated, but it's the same underlying platform, and the photos below are from that same period of hardware development.
A quarter-pallet display on the shop floor
A Solis-branded promotional display trialled in a Spanish retailer, physically the same category of asset this Discovery track was designing around, photographed for the product team's own field research.
Quarter pallet with a weight sensor
An early V1 weight-sensor build, location on this version could only be captured manually. This is the next-horizon hardware referenced throughout this case study, developed and trialled on a separate track from the activation work.
Standard pallet with a location device
The hardware behind Green Garden's own displays: a full-size pallet fitted with a location device, used only for delivery, activation and completion timing, exactly the scope this Discovery track focused on.
Green Garden Products
The US CPG brand whose real promotional programme Round 1 was designed and tested against, alongside the fictional data used in every session and every screenshot in this case study.
Wireframing before the visual prototype
Before any of the Figma screens in Round 1 were made to look real, I worked through the service-design thinking in lower-fidelity wireframes: what a dashboard listing hundreds of displays needed to show, how a merchandiser would confirm a display was compliant on their phone, and the logic behind every notification the system could send. That thinking, not the visual polish, is what Round 2 later rebuilt in code.
The merchandiser's compliance check
A field-side flow: confirm the store and zone are correct, capture a photo of the activated display, and get confirmation the campaign is live, the mobile counterpart to the dashboard above.
The logic behind every alert
Before any screen existed, I mapped the actual decisions a notification system needs: what counts as late, who gets told, and when, the content model the alerting logic in Round 2 was built on.
Round 2 took what Round 1 validated and rebuilt it almost entirely inside an AI-accelerated playground.
Round 2: built almost entirely in code
This year's round returned to the same question, activation, but tested it as a working, data-driven prototype built in VS Code with Claude, not a set of static screens.
Where Round 1 leaned on Figma, Round 2 leaned on the same AI-accelerated, playground-first workflow covered on the AI Playground page: enterprise-grade Svelte and TypeScript components, built and iterated directly in code, so what users tested in sessions was a functioning prototype rather than a click-through mock. That mattered specifically for a data product like this one: activation status, store maps, and display-level detail only feel real when the underlying data actually moves.
Olga ran the research sessions again; I sat in on them, and built and adjusted the prototype in response to what we were seeing in real time between sessions, something that would have taken far longer working purely in Figma. As with every prototype in the Playground, the data behind it, retailers, products, the promotional campaign itself, was entirely fictional, invented specifically so the concept could be tested honestly without exposing a single piece of real commercial or customer data.
Building the prototype in code rather than Figma meant every change Olga and I wanted to make between sessions, a different alert threshold, a new status state, a different way of grouping stores, could be made and re-tested within the same day, not queued up for the next design pass.
Two rounds of prototyping, grounded in real store visits and stakeholder interviews, is what got this to customer testing.
From Discovery to customer testing
The evidence from both rounds of prototyping, and from the store research behind them, is what took front-vs-back-of-store activation classification from a Discovery hypothesis into a capability now being tested with real brands and retailers.
Real-time front-versus-back-of-store activation classification is the "Now" horizon on Brambles' Retail Promotions roadmap, and it's in active customer testing today. A separate, later-stage trial is already running on-device in stores to test the next horizon, weight/consumption sensing, and beyond that sits precise in-store location and a unified compliance score, all sequenced on the same activation foundation this Discovery track and both rounds of prototyping established.
None of that would have been credible without staying close to the store floor throughout: the trader mindset that shapes how a store manager actually behaves, the decanting that looks like non-compliance but is really a rational trade-off, the compliance checklist that never leaves the store. That's the case for why I stayed personally involved past the design file, into the store visits and the stakeholder interviews themselves, not just the screens that came out the other end.
What started as an assumption about a promotional pallet is now evidence behind a real product, in testing with real customers.
Where this stands today
Evidence from the store floor, two rounds of prototyping, and a capability now in front of real customers.
What this demonstrates
The store isn't a UI problem. It's a data-visibility problem that happens to need a UI once you can finally see it. Getting that right meant staying close enough to the shop floor, the stockroom, the store manager's checklist, to design something that matched what actually happens there, not what a plan says should happen.
Not "did the pallet arrive." Did it get to the front of the store in time to sell.
This was buddying up closely with a lead UX researcher rather than working from her findings at a distance, visiting stores and interviewing stakeholders personally, designing and building two rounds of prototyping myself, once in Figma, once almost entirely in an AI-accelerated playground, and staying in the room for the sessions that tested both. The result is a capability now in active customer testing, on a roadmap toward turning a recovered pallet into one of Brambles' most promising new data products.
See the same discipline applied at a different scale.