AI Retail Kiosk ROI — Worked Math for a Single Liquor Store

Most ROI claims for retail technology are made by the vendor selling it. So when I write about Remi's ROI I want to do something different: show the actual levers, show the math, and let you plug in your own numbers. The honest answer is that the ROI on an AI retail kiosk depends on three or four variables specific to your store, and if you don't know those numbers, the kiosk is also a useful tool for finally measuring them.

This post walks through the math for a single independent liquor store. The same framework applies to convenience, wine, specialty food, and grocery — the variable values change, the structure doesn't.

The simple frame

Every ROI calculation in retail tech reduces to:

Monthly value created  -  Monthly cost  =  Monthly net benefit

For an AI concierge kiosk, "monthly value created" comes from four levers:

1. Average-ticket lift on customers who interact with the kiosk.
2. Conversion lift — customers who would have left empty-handed but now don't.
3. Staff time freed — clerks doing inventory, customer service, or restocking instead of answering "do you have any" questions.
4. Inventory intelligence — knowing what customers asked for that you didn't have, so your next purchase order is tighter.

I'll show each lever with a worked example, then the cost side, then the net.

The example store

A single-location independent liquor store. Numbers chosen to be representative of stores I've talked to, not any specific store:

• Daily customer count: 200 transactions
• Average ticket: $28
• Monthly revenue: ~$170,000
• Gross margin: 25%
• Owner-operator plus 1.5 FTE clerks
• Hours open: 70 per week

These are placeholder numbers. Plug your own in.

Lever 1: Average-ticket lift

The mechanism: a customer interacts with the kiosk, gets a recommendation that includes a complementary item or a slightly better-fit primary product, and adds the item to the basket they would have already bought.

The honest range I've seen people cite for AI-driven recommendation systems in retail varies widely depending on category, time period, and how the lift is measured. I'm not going to anchor a number — I'll show the math at a few values so you can apply your own assumption.

Assume a meaningful share of daily customers interact with the kiosk. Of those, some take a recommendation that adds to their ticket. The math:

Daily customers:                                  200
Engagement rate (customers who use kiosk):         ?  (call it E, as a fraction)
Recommendation acceptance rate:                    ?  (call it A, as a fraction)
Average added value when accepted:                 $X

Daily added revenue = 200 × E × A × X
Monthly added revenue (30 days) = 200 × E × A × X × 30
Monthly added margin = monthly added revenue × 0.25

A worked instance, with conservative assumptions: if 25% of customers engage, 30% of those accept a recommendation, and the average added value is $4 (one mixer, one snack, one upgrade), then:

Daily added revenue = 200 × 0.25 × 0.30 × $4 = $60
Monthly added revenue = $60 × 30 = $1,800
Monthly added margin = $1,800 × 0.25 = $450

Re-run with your own numbers. If your engagement rate is higher (a destination store), the number scales. If your average added value is higher (wine recommendations, specialty pairings), the number scales harder.

Lever 2: Conversion lift

The mechanism: a customer walks in unsure, can't find help, walks out. With a kiosk, they ask Remi, get an answer, and complete a purchase.

This is harder to measure because you don't have a baseline for "customers who walked out." But you can estimate it. Spend a Saturday at the door counting people who entered vs people who left with a bag. The difference, conservatively halved, is your "could have been converted" pool.

Worked instance: if 5% of daily walk-ins currently leave empty-handed and the kiosk converts a quarter of those:

Daily walk-outs converted = 200 × 0.05 × 0.25 = 2.5
Monthly converted = 2.5 × 30 = 75 transactions
Monthly added revenue = 75 × $28 = $2,100
Monthly added margin = $2,100 × 0.25 = $525

Even at lower conversion rates (1% walk-out, 10% recovered), the number is non-trivial: about $42 of added margin per month. The lever is most powerful in stores with high foot traffic and a deep catalog where customers regularly need help.

Lever 3: Staff time freed

The mechanism: clerks who would have spent time answering "do you have" and "what's like" questions during shift can now do inventory, restocking, customer service for the harder questions, or just have a faster line at the register.

You can value this in two ways: as labor saved (you can run the same store with one fewer clerk-hour per shift) or as labor reallocated (the existing clerks get more done).

Worked instance, valuing as reallocation: if each clerk-shift currently includes 30 minutes of fielded discovery questions that the kiosk now handles, and you have 3 shifts per day at $20/hour fully loaded:

Daily time freed = 30 min × 3 shifts = 1.5 hours
Daily value = 1.5 × $20 = $30
Monthly value = $30 × 30 = $900

Whether to call this "savings" depends on your model. If the freed time becomes inventory work that would have required overtime, it's real cash savings. If it just becomes "clerks are less stressed," that's still valuable but doesn't show up in the P&L.

Lever 4: Inventory intelligence

The mechanism: every conversation Remi has logs the products customers asked for, including the ones you didn't have. Over a month, you have a list of "things 47 customers asked about that I don't carry." That list is purchasing gold.

Worked instance: if the inventory data lets you stock 2 new SKUs per quarter that turn out to be hits at $200/month in incremental sales each:

Monthly added revenue from better SKU mix: $400
Monthly added margin: $100

This is the slowest-to-show-up lever and the hardest to attribute cleanly, but in stores I've talked to it's often the lever owners value most after a few months because it changes how they buy.

The cost side

Subscription cost for Remi is published on the pricing page. For ROI purposes, the cost has two parts:

Subscription: flat monthly per kiosk. No per-conversation fees. No usage-based surprises. Plug your tier from the pricing page into the math.

Hardware: one-time, amortized. From our hardware walkthrough, a single-kiosk hardware bill is typically a few hundred to under a thousand dollars including tablet, mount, UPS, and cabling. Amortize over a 3-year tablet life and the monthly cost is small — call it $25-$30.

Setup time: about an hour of your time, plus any data prep on your catalog. Treat as zero ongoing.

So the all-in monthly cost is the subscription tier plus roughly $25-$30 of amortized hardware.

Putting it together

Using the worked numbers above (which are conservative for an active liquor store):

Lever 1 (avg ticket lift):          +$450/mo margin
Lever 2 (conversion lift):          +$525/mo margin
Lever 3 (staff time freed):         +$900/mo value
Lever 4 (inventory intelligence):   +$100/mo margin
Total monthly value:                ~$1,975/mo

Subtract your subscription cost and amortized hardware. The remainder is your monthly net benefit. For most pricing tiers in single-store deployments, the net is positive in month one if even half the levers materialize at the levels above.

The phrase to internalize: payback period in months ≈ (subscription + setup) ÷ monthly net value created. At reasonable assumptions, payback is under 90 days for a typical single-store deployment. At aggressive assumptions, it's faster. At pessimistic assumptions where only one lever fires, it's slower but still within a year.

What I won't claim

I won't tell you Remi increases average ticket by a specific percentage. I've seen vendors cite 15%, 20%, 30% lifts. The honest answer is that lift depends on engagement rate, acceptance rate, the category, the time of year, and how the store currently operates. Some stores will see double-digit ticket lift. Some will see low single digits and gain everything from levers 2 and 3 instead. The math above is structured so you can plug in conservative numbers and still see the case.

I also won't tell you the staff time freed is "savings" if you don't actually reduce labor. Treat it as savings only if you're cutting an hour or eliminating overtime. Otherwise it's reallocation, which is real but not P&L cash.

The cheapest way to test the ROI

The cheapest way to validate this for your store is to run a 30-day pilot, measure your average ticket, walk-out rate, and staff productivity before and after, and compare. We structure pilots to make this easy — a single kiosk, the basic configuration, dashboard analytics turned on. Book a demo or read about how the solutions for liquor stores page describes what we measure during pilots.

If the math doesn't work for your store, we'd rather you know in 30 days than in 18 months.

Frequently asked

What's a realistic payback period for a single store?

At conservative assumptions on the levers above, payback is typically under 90 days. At pessimistic assumptions where only one or two levers fire, it can extend to 6–12 months. The structure of the math is the same — only the inputs change.

How do I measure average-ticket lift cleanly?

Compare the average ticket of customers who interacted with the kiosk to those who didn't, over the same time period and ideally similar dayparts. Most POS systems can be tagged at the register or, if not, the kiosk can hand off a code that ties the conversation to the eventual transaction.

Does the ROI hold for smaller stores?

The math is the same; the absolute dollars are smaller. A store doing $40k/month rather than $170k will see proportionally smaller monthly value. The subscription is also smaller-store-friendly — see the pricing page.

What if my customers don't engage with the kiosk?

Engagement rate is the single most important variable. If you have a destination store with curious customers (wine, specialty), engagement is high. If you have a quick-grab convenience format, engagement is lower and the kiosk is less valuable. Honest pre-pilot scoping should tell us which you have before you sign anything.

Where do the numbers actually come from in a deployed kiosk?

The dashboard shows daily and monthly engagement, recommendation acceptance, common asked-for-but-not-stocked items, and a per-conversation log. You see the same data we do. There's no black box.