Introducing the AI Kiosk ROI Calculator — Methodology and How to Use It

The single most common question we get from independent retailers evaluating an AI kiosk is some version of "what is this actually going to do for my numbers." It's the right question. The wrong way to answer it is with a hand-wavy ROI promise. The right way is to give the operator a calculator with their own inputs, methodology that's transparent enough to argue with, and outputs that tie to operating metrics they already track.

We're shipping that calculator at /customers/roi-calculator/. This post is the methodology behind it, written for the operator who wants to know what the math is doing before trusting the output.

What the calculator does

In one sentence: it estimates the monthly economic impact of deploying Remi in your store, expressed in dollars, in payback period, and in staff-hour equivalents. You put in numbers about your store; it gives you an estimated range, not a single number. Ranges, not point estimates, because retail is variable and anyone giving you a single confident number is selling you something.

The calculator is intentionally conservative. We'd rather under-promise and have a customer pleasantly surprised than the opposite.

What inputs it needs

Six inputs. None of them require accounting forensics — every operator we've talked to can pull these from their POS in fifteen minutes.

1. Average daily transaction count. How many sales per day, averaged across a typical week. Your POS has this report.
2. Average ticket size. Total sales divided by transaction count. Same report.
3. Vertical. Liquor, convenience, beauty, supplements, hardware, specialty grocery, other. The vertical matters because ticket-lift sensitivity differs across categories — liquor and wine respond more to recommendation than convenience does, mostly because the products carry more explanation.
4. Store hours per week. Used to compute the labor-hour equivalent.
5. Average staff hourly cost (loaded, including taxes and benefits). If you don't know, $22/hr is a defensible national average for retail; the calculator defaults to that and lets you override.
6. Number of kiosks you're considering. Most independent stores deploy one. Multi-counter stores sometimes do two.

The calculator runs locally in the browser. We don't store the inputs, and we don't email-gate the result. You'll get a "want to talk to us" prompt at the end, but it's optional and ignorable.

How the math works

There are three output numbers, each with its own methodology.

Payback period

Payback is the simplest:

payback_months = (subscription_cost + amortized_hardware_cost) / monthly_incremental_gross_profit

The subscription cost comes from our pricing — we use the tier most operators choose. The amortized hardware is $400 spread over 36 months ($11/month). The hard part is monthly_incremental_gross_profit, which is where the assumptions live.

We compute incremental gross profit as:

monthly_incremental_gross_profit = 
    daily_transactions × ticket_lift_pct × avg_ticket × gross_margin_pct × days_per_month

We do not assume Remi increases transaction count — only ticket size. The reason: ticket-lift evidence from our pilot stores is more consistent than transaction-count evidence, which depends on customer acquisition that we don't directly drive. Conservative bias.

ticket_lift_pct is set by vertical, with a low/mid/high range:

• Liquor & wine: 3% / 6% / 9%
• Convenience: 1% / 2.5% / 4%
• Beauty & supplements: 4% / 7% / 11%
• Hardware: 2% / 4% / 6%
• Specialty grocery: 2% / 4% / 6%
• Other: 1.5% / 3% / 5%

These ranges come from our pilot data and are calibrated against published industry benchmarks for in-store recommendation systems. They are ranges, not promises. The calculator shows all three.

gross_margin_pct defaults by vertical too — 25% for liquor, 30% for convenience, 45% for beauty/supplements, 35% for hardware, 30% for specialty grocery — and you can override.

Average ticket lift sensitivity

The calculator includes a sensitivity slider. Move the ticket-lift assumption and watch payback change in real time. This matters because the operator's reasonable disagreement with us is usually about ticket lift — they think their store is more or less responsive than our base rate. The slider lets them argue with the assumption directly.

Practical interpretation:

• At low end of vertical range: payback typically falls in the 6–12 month window for stores with 80+ daily transactions and a $25+ ticket.
• At mid range: payback falls in the 3–6 month window for the same store profile.
• At high range: payback falls below 3 months, but we caution against banking on the high end of the range; it's there for completeness, not as a planning number.

If the calculator shows a payback period over 18 months at the mid range, the kiosk probably isn't right for your store. We'd rather you walk away than buy it under bad expectations.

Staff-hours equivalent

This is the output operators ask for that other AI vendors don't usually compute. It answers: how many staff-hours per week is Remi handling that would otherwise be handled by a human, or not handled at all?

We compute it as:

weekly_staff_hours_equivalent = 
    daily_transactions × interaction_rate × avg_interaction_minutes / 60 × 7

Where:

• interaction_rate is the share of transactions that involve a customer-staff conversation about a recommendation. We use 25% as the default — about one in four shoppers in indie retail asks a staff question pre-purchase.
• avg_interaction_minutes is the average length of that conversation, defaulting to 2 minutes.

A 100-transaction-per-day store with these defaults gets about 6 staff-hours per week of conversation handled by Remi. At $22/hr loaded, that's roughly $570/month in staff-time equivalent — separate from the ticket-lift calculation.

We don't add the staff-hours-equivalent dollar value to the ROI calculation by default. It's shown as a separate output, because most stores don't actually reduce staff hours when they deploy a kiosk; they redirect staff to higher-value work like inventory, regulars, and complex conversations. Counting both lifts and labor savings would be double-counting in most stores.

If your store is genuinely understaffed at peak — a Friday-night liquor store with a one-person register — the staff-hours equivalent is real economic value the calculator just doesn't put a dollar on. That's a conservative choice on our part.

How to interpret the output

A few rules of thumb when reading the calculator's output:

Trust the mid-range. Plan around it. Use the low range to test whether the deployment is still defensible if the kiosk underperforms. Treat the high range as upside, not as a planning number.

Look at payback in months, not in dollars. Dollar values feel concrete but are sensitive to assumptions you may want to revise. Payback period is a more stable comparison number across scenarios.

Compare to your alternative. The right comparison isn't Remi vs. nothing — it's Remi vs. the next best thing you could spend that money on. Hiring another part-time clerk? More inventory? A better POS? The calculator gives you Remi's numbers; the comparison is yours to make.

Re-run after 90 days. If you deploy, re-run the calculator after a quarter using your actual ticket lift and your actual interaction rate. Most stores find their real numbers are inside the range we predicted; some find they're above. A few find they're below, and we want to know about those because that's how the model improves.

What the calculator can't do

Honest limits:

• It doesn't model seasonality. A liquor store's December is not its February. The calculator gives you an annualized monthly average; your peak months will be better and your trough months worse.
• It doesn't model acquisition. If you have a marketing problem (people aren't walking in), Remi doesn't fix it. The calculator assumes you have the foot traffic you have today.
• It doesn't model competitive response. If a competitor across the street puts in their own kiosk a year later, that's not in the model. Most indie retail markets don't see this kind of competitive response yet.
• It doesn't model your team's adoption. A kiosk that the staff resents will underperform. Adoption is a soft variable; the calculator assumes a normal-engagement deployment.

Where to find it

The calculator lives at /customers/roi-calculator/. It's free, takes about three minutes, and doesn't require an account. If you want to walk through your numbers with us instead of doing it solo, book a demo and we'll run it together.

If you're in liquor or convenience specifically, we have vertical-specific guides at /solutions/liquor/ that drill into the assumptions for that category.

Frequently asked

Where do the ticket-lift ranges come from? A combination of our pilot store data (early but real) and published benchmarks for in-store recommendation systems. We're transparent about the small sample on our pilots — that's why we publish ranges rather than point estimates.

Is the calculator biased toward making Remi look good? We've tried hard to bias the other way. We don't add staff-hour savings to the ROI by default, we use the bottom of vertical gross-margin ranges, and we cap ticket-lift at conservative numbers. If you think the math is generous somewhere, tell us and we'll show our work.

Can I export the result? Yes — the calculator generates a PDF you can share with a partner, a CFO, or a banker.

What if my numbers are way outside the assumptions? Run the calculator with overrides. Every default can be changed. If your store profile is unusual (very high ticket, very low transaction count, multi-vertical mix) the calculator may not give you a useful answer and you should talk to us directly.

Will you publish the full methodology somewhere I can audit? Yes — once the calculator is live, the methodology page links from inside it. This blog post is the prose version; the calculator page itself includes the math in a collapsible "show your work" section.