Key takeaways
- Vending screens sell precision, not reach: the impression and the purchase happen on the same device.
- Closed loop isn't the same as incremental. Use matched holdout devices or on/off periods to measure lift.
- On small networks, sell flights per device-week with a measurement report rather than CPMs.
- Location hosts provide the space and the audience. Give them a revenue share, category controls and approval rights.
- Never let media get in the way of the price, the checkout or operational messages.
Next to a grocery chain’s network of in-store screens, a 10-inch display on a cooler door looks like a rounding error. It might reach a few hundred people a week in a single breakroom. But it has one property the big networks spend heavily trying to approximate: the screen and the shelf are the same device. The impression and the purchase happen within seconds and inches of each other, recorded by the same system. That makes vending screens a small medium with unusually honest measurement, and honest measurement is what you should sell.
Why vending screens are different
- Proximity. The product sits behind the glass the ad plays on. There’s no store to walk across between seeing and buying.
- Repeat audiences. Workplace, hospital and campus devices see the same people most days. That’s bad for reach and great for building habits, testing frequency or seeing whether a trial turns into repeat purchase.
- Closed loop by default. The device knows what played and what sold. No matching between an ad platform and a point-of-sale system is required.
- Fragmentation. Screens are spread across many operators and small sites. Aggregating them into something a brand can buy is the industry’s real challenge.
Large grocers and pharmacy chains have been expanding in-store screen networks, and measurement is a recurring sore point for buyers of that media. Industry groups including the IAB and the Media Rating Council have published retail media measurement guidance that pushes networks toward transparent, well-documented methods. Vending screens can meet that bar more easily than most, if they’re disciplined about it.
Formats that work
| Format | What it is | Best for | What to measure |
|---|---|---|---|
| Attract loop | Short brand spots when no one is shopping | Awareness at high-traffic sites | Promoted SKU sales vs. holdout devices |
| Pre-tap message | A spot or offer on the door display before the shopper opens the door | Nudging choice at the moment of decision | Promoted SKU share of sessions |
| Sponsored discount | A brand-funded price promotion shown on screen | Trial and switching | Incremental units, cost per incremental unit |
| Sampling | A free or deeply discounted first unit for new buyers | New products and new sites | Trial rate, repeat purchase at full price |
| Launch pilot | New SKU plus prime shelf plus screen support for a set period | Testing a product before wider distribution | Rate of sale vs. category, repeat rate |
| Shelf placement | Paid eye-level facings | Ongoing visibility | Rate of sale vs. prior placement |
On a touchscreen spiral machine there’s also the browse moment: a sponsored product can be highlighted while the shopper scrolls. On a cooler, browsing happens with the door open and eyes on the shelf, so the door screen’s job is to shape the decision before the tap.
Measurement: closed loop done right
Closed loop means you can connect exposure to purchase. It doesn’t mean every sale during a campaign was caused by it. People bought that sparkling water last month, too.
Exposure. Log every play on every device: what ran, when and for how long. Add anonymous presence counts if you have them, so you know whether anyone was standing there. You don’t need to identify anyone to measure well.
Outcome. Same-device sales of the promoted SKU, plus the brand’s other SKUs and the category, so switching and cannibalization are visible.
Incrementality. Compare against a control. Two practical designs:
- Matched holdouts. Split devices into test and control groups, matched by site type, size and recent category sales.
- Switchbacks. Alternate campaign-on and campaign-off weeks on the same devices. Useful when the fleet is too small to split.
A worked example, with illustrative numbers. A beverage brand runs a 4-week flight on 40 test coolers, with 40 matched control coolers.
- Before the flight, test devices sold 2.0 units of the promoted drink per device per week; controls sold 2.1.
- During the flight, test devices sold 3.1; controls sold 2.3.
- Change on test devices: 3.1 − 2.0 = 1.1. Background change on controls: 2.3 − 2.1 = 0.2.
- Incremental effect: 1.1 − 0.2 = 0.9 units per device per week.
- Across 40 devices and 4 weeks: 0.9 × 40 × 4 = 144 incremental units.
At $2.75 a bottle, that’s about $396 of incremental revenue. Then check the brand’s other SKUs: if its original flavor fell on the test devices, part of that lift was internal switching, and the brand deserves to know.
That number is small, and that’s the honest pitch. Vending media doesn’t win on scale. It wins on clean answers: which message worked, which site types responded, whether trial turned into repeat. Run flights long enough to see through weekly noise, decide the success metric before launch, and report the control group’s numbers alongside the test group’s.
How long is long enough?
Weekly sales of one SKU on one device are noisy. A device that averages 2 units a week will often show 0 one week and 5 the next. Forty devices per group smooths much of that out, but a 4-week flight can still be swamped by one site’s event or closure. Three practical rules help:
- Remove devices that went offline or changed planogram during the flight from both groups, using the same rule for each.
- Look at results device by device as well as in aggregate, so one outlier site can’t carry the conclusion.
- If the lift is smaller than the week-to-week wobble you saw in the pre-period, report it as inconclusive rather than rounding it up.
Brands remember the networks that told them a test didn’t work.
Pricing models
| Model | How it works | When it fits |
|---|---|---|
| CPM | A price per thousand plays or opportunities to see | Large networks with independent-style reporting |
| Flight fee per device-week | A flat price per screen per week | Small and mid-size networks; simplest to buy and audit |
| Cost per incremental unit | The brand pays on measured lift | Brands with a strong testing culture and a method both sides trust |
| Sampling fee | Product plus a fee per trial delivered | Launches and sampling campaigns |
| Bundle | Screen plus shelf placement plus sponsored discount | Launch pilots |
Below a few thousand screens, CPMs mostly invite awkward comparisons with networks a thousand times larger. A flight priced per device-week, with a holdout-based report at the end, is easier to buy and easier to defend. As an illustrative example, at $6 per device-week, a 4-week flight on 40 devices costs the brand $960, and it can see exactly what it got.
Host revenue share
Three parties have a claim on media revenue: the location host, who provides the space and the audience; the operator, who owns and services the device; and the platform, which sells, serves and measures the campaigns. A fair agreement spells out:
- The host’s share of net media revenue, with what “net” excludes defined in writing
- Category controls: a hospital may block sugary-drink ads, a campus may block energy drinks, a gym may want only hydration and protein
- Approval rights over creative, or at least over new advertisers
- Frequency caps and a share of screen time, so media never crowds out prices and operational messages
- Data terms: what the brand receives (aggregated campaign results, never personal data)
- Reporting cadence and payment terms
Some workplace hosts direct their share into the employer subsidy program, so the ads end up paying for part of employees’ lunch. It’s a good story, and a good reason for the host to keep the program running. See For Location Hosts for how hosts typically set this up.
Creative and privacy rules
- Keep spots short and silent by default; breakrooms are workplaces.
- Show the package exactly as it looks on the shelf, so the shopper can find it in a second.
- Skip fine print that nobody can read from a normal standing distance.
- Avoid rapid flashing; it’s an accessibility problem as well as an annoyance.
- Use no facial analysis and no personal targeting on a shared device.
- Make sure sponsored discounts follow the same price-display rules and guardrails as any other price change, including the ones we describe for expiry-aware pricing.
When it isn’t worth it
If no brand wants your audience, a screen full of house ads is just a screen. If the host forbids advertising, respect that. And if media starts competing with the price display or the door status, you’re costing yourself sales to earn a few dollars of ad revenue. The screen’s first job is helping someone buy lunch; media is a guest on it.
How 8Kava approaches this
Retail Media on Kava OS Autopilot runs campaigns on Kava Cooler door displays and Kava Vend touchscreens, measures outcomes on the same devices that showed the ad, and gives brand partners a read-only campaign view. Shopper Analytics supplies anonymous traffic and dwell counts with no facial recognition, and hosts share in the revenue. As a shopper marketing lead at a beverage brand told us: “Our brand ran a sampling campaign across 60 coolers and could see which screens actually drove purchases.” Brands can start on our For Brands page.