The Maths of Confidence
Virtual Try On Cost vs the Cost of a Return
A business-case analysis published by AI Frame · July 2026
About this analysis: this paper is published by AI Frame, a vendor of virtual try on and size chart software for Shopify. It is a transparent cost analysis, not neutral research. Every external figure is cited; AI Frame's own pricing is the only first-party input; and all worked examples are clearly-labelled hypothetical illustrations. The arithmetic is deliberately simple so that any merchant can re-run it with their own numbers and should.
Executive summary
An apparel return costs a merchant, all-in, somewhere between $20 and $45, a figure documented repeatedly across returns-industry research. A virtual try on costs a merchant between $0.12 and $0.22 to generate. That gap of two orders of magnitude produces an unusual economic property: try on does not need to work often to pay for itself. At these prices, preventing a single return covers the cost of roughly 90 to 375 try ons - meaning the technology breaks even if as little as 0.3–1.1% of try ons prevent a return. Published evidence on fit-technology deployments reports return reductions far above that threshold. This paper builds the cost of a return from its components, states the cost of a try on transparently, derives the break-even arithmetic, stress-tests it in a worked example with pessimistic, central and optimistic scenarios, and states plainly the situations in which the maths does not favour adoption.
1. The cost of a return, built up honestly
The refund is the visible cost of a return. The real cost sits underneath it, distributed across payroll, shipping spend and inventory write-downs, which is why most operators underestimate it - one returns-industry analysis notes that operators who first calculate their true per-return cost typically find it roughly double what they had been reserving [1].
The components of a single apparel return:
| Component | Typical range (USD) |
|---|---|
| Return shipping (reverse logistics) | $8–12 per domestic label [2] |
| Receiving, inspection & grading labour | $10–15 [2] |
| Restocking & repackaging | $2–10 [2] |
| Markdown / resale loss | variable — only ~48% of returned apparel resells at full price [3] |
| Customer-service time | $2–5 [4] |
Independent estimates of the all-in total converge on a consistent band. Radial and associated industry analysis put the true processing cost at $27–41 per item, with direct-to-consumer apparel brands frequently exceeding $40 [2]. Eightx's 2026 apparel-specific cost model lands at a $30 midpoint with an honest band of $20–45, cross-validated against Coresight Research's finding that apparel returns processing runs as high as ~66% of product price, and ShipBob's worked figure of $33 to process the return of a $50 garment [1][5]. Pitney Bowes research places fashion-return processing at $10–30 per item [6], and Statista's compilation at $10–40 [7]. Expressed as a share of order value rather than a fixed figure, Optoro's cross-vertical benchmark is 27% of purchase price, and Pitney Bowes' is 21% of order value [5].
For this analysis we therefore adopt a deliberately conservative range: an all-in cost of $20–45 per apparel return, with $30 as the central case. Merchants with higher item values, international return shipping, or low resale recovery will sit above this band, not below it.
Two aggravating factors deserve mention. First, the volume: online apparel returns run at 20–40% of orders depending on segment [8], so these per-unit costs recur constantly. Second, the majority of these returns are informational failures rather than product failures, McKinsey attributes 70% of fashion returns to fit or style [9], which is precisely the category of return that pre-purchase confidence tools address.
2. The cost of a try on
AI Frame's merchant-facing pricing, stated in full: each plan includes a monthly try-on allowance, and additional try-ons are billed per generation at the plan's rate: $0.16 on Starter ($19.99/month, 100 try ons included), $0.14 on Growth ($59.99/month, 500 included), and $0.12 on Advanced ($199.99/month, 1,250 included). Cached re-views of an existing render cost nothing; only fresh generations are billed. There are no setup or per-store fees beyond the plan.
For the arithmetic that follows, the relevant number is the marginal cost per try-on: $0.12–0.16 depending on plan.
One cost concern should be addressed immediately, because a careful reader will raise it: the merchant pays for try ons generated by shoppers who never buy. This is true, and the honest way to handle it is to fold it into the break-even calculation rather than wave it away, which is exactly what the ratio in the next section does.
The question is never "did every try on convert?" but "across all try ons paid for, did enough returns get prevented to cover the total?" Every figure below is computed on that all-in basis.
3. The break-even arithmetic
The core calculation is one division: how many try-ons equal the cost of one prevented return?
| Return cost → | $20 (low) | $30 (central) | $45 (high) |
|---|---|---|---|
| @ $0.16/try on (Starter) | 125 | 188 | 281 |
| @ $0.14/try on (Growth) | 143 | 214 | 321 |
| @ $0.12/try on (Advanced) | 167 | 250 | 375 |
Read the central column: at a $30 return cost and Growth-plan rates, one prevented return pays for 214 try ons.Inverted, that is the break-even prevention rate: if just 1 in every 214 try ons (≈0.47%) stops one garment from coming back, the try on spend is fully recovered, before counting any conversion benefit at all.
Expressed at the plan level, the arithmetic is even blunter: the Growth plan costs $59.99 per month. At a $30 per-return cost, the entire plan breaks even at two prevented returns per month. Starter breaks even at one prevented return every six weeks.
Is a prevention rate above ~0.5% of try ons plausible? The published evidence sits far above that threshold. Deployed AR try on tools report return-rate reductions of 30–40% where used [10]; a more conservative industry assessment credits size charts, fit tools and AR try-on collectively with 10–20% reductions in apparel returns [4]. Given that 70% of fashion returns are fit- or style-driven [9], exactly the failure mode try on and size charts address, even the most pessimistic published figure implies prevention rates one to two orders of magnitude above break-even. The technology does not need its headline numbers to be true to be profitable; it needs perhaps a twentieth of them.
4. A worked example, stress-tested
The following is a hypothetical illustration with stated assumptions, not a case study or measured merchant result.
Assumed merchant: an online fashion store with 1,000 orders/month, average order value $75, a 25% return rate (250 returns/month - mid-range for online apparel [8]), and an all-in cost per return of $30 (central case, §1). Baseline returns cost: $7,500/month. The merchant runs AI Frame's Growth plan ($59.99/month) and shoppers generate 500 try-ons/month, staying within the included allowance — so the total try-on cost is the plan fee.
Three scenarios for the effect of try-on + size chart on returns:
| Scenario | Assumption | Returns prevented / month | Gross saving | Try on cost | Net position |
|---|---|---|---|---|---|
| Pessimistic | try on influences only a small slice of orders; a 2% reduction in total returns | 5 | $150 | $59.99 | +$90/month |
| Central | 6% reduction in total returns (well below the 10–20% published band [4]) | 15 | $450 | $59.99 | +$390/month |
| Optimistic | 16% reduction - inside the published 10–20% band, far below AR-reported 30–40% [10] | 40 | $1,200 | $59.99 | +$1,140/month |
The deliberately notable feature of this table is the pessimistic row: even at a 2% return reduction — a fraction of any published figure — the tool is net positive. The sensitivity that actually matters is not the prevention rate but the merchant's own inputs: at very low order volume or very low return cost, the fixed plan fee looms larger (see §6). Merchants should re-run this table with their own return rate, return cost and expected try-on volume; the structure of the calculation is the point, not our illustrative numbers.
5. What the return-cost line doesn't capture
The analysis so far counts only prevented returns, which understates the case in four documented directions.
Conversion. Shoppers who use virtual try-on have been reported to convert at multiples of those who don't - platform-reported figures via Forbes cite up to 10x [11]. Self-selection plausibly inflates this, and we deliberately excluded conversion from the break-even maths. But directionally, some portion of try-on spend is buying additional sales, not merely avoided costs.
Bracketing. Ordering multiple sizes with intent to return is estimated to account for ~40% of fashion returns [12], and 63% of consumers admit to the practice [13]. Bracketing is a rational response to size uncertainty; a shopper who has confirmed their size against a chart beside their own try-on has materially less reason to order three. Reduced bracketing shows up as fewer orders-per-purchase and fewer returns - a saving partially distinct from the fit-confidence effect already counted.
Customer-service load and cart abandonment. Fit questions are a staple of pre-purchase support tickets, and return anxiety suppresses checkout - a January 2026 DHL report found 79% of shoppers abandon carts over unsatisfactory return policies [14]. Confidence at the product page relieves both.
Reverse-logistics externalities. A returned parcel's carbon footprint runs roughly 1.5–2x the original delivery [15]. For brands with sustainability positioning, prevented returns are a reportable environmental saving, not only a financial one.
None of these is needed for the case to close - §3 closes it on prevented returns alone - but a merchant modelling full value should note the direction of every one of them is positive.
6. Where the maths does not favour try-on
Credibility requires stating the boundary conditions plainly. The break-even logic weakens or fails in four situations.
Very low return-cost categories. A store whose returns cost $5 to handle (light, cheap, locally-returned items) needs 3–6x more prevented returns to break even than the central case. If your all-in return cost is genuinely low, recompute before buying anything.
Low-return categories. The arithmetic runs on fit- and appearance-driven returns. Merchants selling items with single-digit return rates have little for the tool to prevent; the case then rests entirely on conversion uplift, which is the softer, less-verified half of the evidence.
Very low traffic. A store generating a handful of try ons a month cannot mechanically prevent enough returns to cover even a small plan fee. (This is, transparently, why AI Frame's Free plan exists - 10 try ons/month at no cost lets a low-traffic store measure before it spends. But the honest statement stands: below a certain volume, no paid plan is justified yet.)
Very low average order value. At a $15 AOV, return costs per item are lower and margins thinner; the per-return saving shrinks toward the break-even line. The tool's economics strengthen with item value.
A merchant in one or more of these situations should either start free and measure, or not adopt at all. The worst outcome for this category and for any vendor in it, is deployment where the arithmetic was never going to close.
7. Conclusion
The economic case for virtual try on does not rest on the technology's most impressive claims. It rests on an asymmetry: a return costs $20–45; a try-on costs $0.12–0.22. Between those two numbers lies a break-even prevention rate of well under 1% of try ons, a threshold that every published measurement of fit-technology effectiveness clears by a wide margin. A merchant does not need to believe the 10x conversion figures or the 40% return reductions; they need to believe that answering "will it fit?" and "how will it look?" before purchase prevents one return in every couple of hundred attempts. The prudent path is the same one this paper has taken: adopt at zero cost, measure your own prevention rate against your own return costs, and let your numbers, not ours, make the decision.
### About this paper
Published by AI Frame, which builds virtual try-on and size chart software for Shopify (apps.shopify.com/aiframe-virtual-tryon). AI Frame's pricing is the only first-party data used; all other figures are cited third-party research. Worked examples are hypothetical illustrations.
References
- Eightx, Apparel Returns: The True Cost and the Fixes (2026) — apparel per-return cost stack: $30 modelled midpoint, honest band $20–45; operators typically find true cost ~double their reserve.
- Radial / Synctrack industry analysis — true per-return processing cost $27–41, DTC apparel frequently $40+; component ranges (shipping $8–12, inspection $10–15, restocking $2–10).
- eFulfillment Service / Eightx (2026) — 48% of returned apparel items resell at full price.
- Claimlane, The True Cost of Returns for Ecommerce Brands (2026) — component costs incl. CS; size charts, fit tools and AR try-on reduce apparel returns by 10–20%; true cost 3–4x refund value.
- Eightx, Returns Processing Cost per Item by Vertical (2026), citing Coresight Research/Optoro (apparel processing up to ~66% of product price; $38.0bn returned vs $25.1bn processing, 2023), ShipBob ($33 on a $50 item), Optoro Returns Unwrapped 2024 (27% of purchase price), Pitney Bowes BOXpoll (21% of order value).
- Pitney Bowes logistics research — fashion return processing $10–30 per item.
- Statista compilation — return processing $10–40 per item.
- ShipNetwork / Coresight Research — online apparel return rates 20–40%; Coresight 24.4% twelve-month measure.
- McKinsey & Company — 70% of fashion returns attributed to poor fit or style.
- ShipNetwork, aggregated merchant data (2026) — AR try-on tools associated with 30–40% return-rate reductions.
- Forbes (April 2026) — DressX-reported conversion multiple among try-on users (~10x); platform-reported, not independently audited.
- Narvar consumer survey (2022) — bracketing accounts for ~40% of fashion returns.
- Synctrack industry analysis (2025) — 63% of consumers report ordering multiple sizes intending to return.
- DHL report (January 2026) — 79% of shoppers abandon carts over unsatisfactory return policies.
- Journal of Cleaner Production (2023) — reverse-logistics carbon footprint ~1.5–2x original delivery.
Methodology note: external figures synthesise publicly available returns-industry research and trade reporting, 2022–2026; several originate from commercial parties and have not been independently audited. AI Frame pricing is current as of July 2026. All break-even calculations are reproducible from the stated inputs.