Case studies
Three deployments. Every number has a "before".
These smart vending case studies come from a Mountain West hospital network, a Midwest vending operator and a Southeast distribution center. The clients are anonymized. Every result is compared with the same site's own prior program, over a stated window.
- revenue at a 3-campus hospital network
- +46%revenue at a 3-campus hospital network
- locations added without a new route
- 140locations added without a new route
- PPE consumption on a 3-shift site
- −24%PPE consumption on a 3-shift site
Start with the site that looks most like yours
Each study covers the starting point, what was deployed, how the rollout ran, the measured results and what we'd tell the next site.
24/7 fresh food for a 3-campus hospital network
14 legacy machines became 22 Kava Coolers and 4 hot-meal lockers. Revenue rose 46% and fresh-food waste fell 31% in six months.
A 640-machine fleet that grew without trucks
520 machines retrofitted with Kava Link in 9 weeks. Service visits fell 29%, and 140 locations were added without a new route.
PPE and fresh food for every shift
3 PPE dispensers, 6 coolers and a micro-market kiosk. PPE use fell 24%, and on-shift PPE stock-outs hit zero after month two.
What moved, and over what window
A number without a baseline is an advertisement
So each study compares a site with itself: the same locations, before and after, over a stated window. Revenue is compared with the prior program's sales. Service visits come from route and ticket records. Waste comes from write-offs logged at restock. Stock-outs come from device telemetry, which counts an empty slot whether or not anyone happened to visit.
We state the window with every figure (first 6 months, first year) because early results and settled results are different animals. New equipment always gets a curiosity bump. A longer window lets it wash out.
- Same sites, before and after. No industry average as the baseline.
- Every figure carries its measurement window.
- Clients are anonymized, and quotes are attributed by role.
Three starting points, three different first moves
| Hospital network | Vending operator | Distribution center | |
|---|---|---|---|
| Scale | 3 campuses, 1,100 beds | 640 machines, 410 locations | 900 workers, 3 shifts |
| The gap | No real food after the cafeteria closed at 7 p.m. | Growth limited by trucks and driver hours | Disappearing PPE and an uneven breakroom across shifts |
| First move | Replace 14 legacy machines with 22 Kava Coolers and 4 Kava Lockers | Retrofit 520 machines with Kava Link in 9 weeks | Add 3 Kava PPE Dispensers, 6 Kava Coolers and a Kava Kiosk |
| What the software did | Automated temperature logs, forecasting by floor, near-date markdowns | Pro, then Autopilot: pre-kitting, routing, then Kava Agent | Badge limits, cost-center allocation, auto-reorder |
| Window | First 6 months | First year | First 6 months |
| Headline result | Revenue +46% | 140 locations added, no new route | PPE consumption −24% |
All three clients are anonymized. Results are measured against each site's own prior program.
What the three have in common
They started with the worst hour
Not the average day. The hospital started with 3 a.m., the operator with the truck that couldn't fit another stop, the distribution center with third shift. The worst hour shows what a site actually needs.
They kept what still worked
The operator kept its machines and added a $349 module to each. The hospital kept its cafeteria and covered the hours it was closed. New hardware went only where the old format couldn't do the job.
They let the data pick step two
Connectivity came first, then forecasting, then automation. The operator moved from Pro to Autopilot once its data was trustworthy. The distribution center set PPE limits from real consumption, not catalog guesses.
Gloves stopped disappearing, and third shift finally has the same food options as first shift.
From deployment to published number
Record the baseline
Capture the prior program's sales, visit counts and known problem hours, so there's a fair "before" to compare with.
Change one site at a time
Phased rollouts give every later site an earlier one to learn from, and they keep surprises small.
Wait out the novelty
Report over 6 or 12 months, not week one. The curiosity bump fades; the real pattern doesn't.
Anonymize and attribute by role
Client names come out and locations shrink to a region. Quotes keep the speaker's role, so you know who is talking without knowing who they are.
Your baseline is disappearing right now.
Every week you run the current program without writing anything down is a week you can't compare against later. Before you change a single machine, pick three numbers: service visits per device, stock-outs you hear about and food you throw away. Six months from now, which of those would you most want to have measured?
Questions about the case studies
Why don't you name the clients?
Because the results belong to them, and so do their vendor relationships. We describe each client by region, size and industry, and attribute quotes by role. That's enough to judge whether a site resembles yours without exposing anyone's operations.
Are these results typical?
They're real, but each one is a specific site fixing a specific problem. Our platform-wide figures are the better guide to averages: −41% out-of-stock events with forecasting, −27% service visits with pre-kitting and dynamic routing, and +18% revenue per device in the first 90 days after a Kava Link retrofit.
Can you estimate results for my site?
Yes. Start with the ROI calculator for a first pass, then ask for a site assessment. We look at traffic, shifts, dwell time and format fit before recommending anything, so the estimate reflects your building, not someone else's.
How long until we'd see results like these?
A smart-cooler program of up to 10 units typically goes live 10 business days after signing. Kava Agent observes for 14 days before proposing autonomous actions. The studies report 6- and 12-month windows because that's when the numbers stop moving week to week.
Where to go next
Bring us your before. We'll help you measure the after.
Tell us about the site, the hours that go unserved and the numbers you already track. We'll suggest a format and a fair way to measure it.