Is Virtual Try-On Worth It? Separate Attributed Sales from Profit

The decision depends on incremental contribution

Virtual try-on is worth its cost when the additional contribution it creates exceeds the cost of providing it. A dashboard showing orders after a try-on helps you describe shopper activity. It cannot, by itself, tell you how many of those orders would have happened anyway.

Keep four questions separate: do shoppers use it, do they buy afterwards, does offering it change purchases, and does the change leave more money after returns and costs? The examples below are synthetic calculations, not merchant results or forecasts.

1. Measure usage before discussing lift

Track eligible product views, entry-point visibility, try-on starts, completed results, cart additions and purchases. Keep failed attempts visible. A feature that is hidden, slow or frequently fails needs a product investigation before an effectiveness claim.

Define whether you count distinct shoppers, sessions, orders or units. Use the same definition in each comparison, and document what consent choices and missing identifiers prevent you from observing. Do not count one shopper's ten generated images as ten independent customers.

2. Treat attribution as a record of activity

aisthetix reports orders and revenue associated with tried-on products under its attribution rules. That is useful for understanding the journey from product view to try-on, cart and purchase. Record the product match, attribution window, order deduplication and treatment of refunds when reconciling any vendor's dashboard with your store.

Suppose 30 orders worth $2,400 are attributed to try-on. That does not mean the app generated $2,400 of additional revenue, and it does not mean it generated $2,400 of profit. Some shoppers would have bought without it; some purchases may later be refunded; product and fulfilment costs still apply.

3. Compare the offer of try-on, not only its users

People who choose to use try-on may already be more interested in buying. Comparing their purchase rate with non-users describes an association. Matching the product reduces one source of difference, but it does not remove that self-selection.

Where practical, assign eligible shoppers at random to a group offered the experience or a control group, and keep that assignment stable. Compare all assigned shoppers, including those who never use the feature. This is an experiment you must arrange; a standard attributed-sales dashboard is not an automatic randomized test.

Before starting, define the primary outcome, eligibility, sample plan, observation window and stopping rule. Keep pricing, promotions, product availability and tracking consistent across groups. A short-term conversion result still needs follow-up for returns. If random assignment is unavailable, a before/after comparison can guide investigation, but label it observational and record changes in traffic, season and merchandising.

Worked example: a 10% relative change is not 10 percentage points

Imagine equally sized groups with the same purchase definition and a completed observation window:

• Control: 5,000 assigned shoppers and 200 purchasing shoppers, or 4.0%.
• Try-on offered: 5,000 assigned shoppers and 220 purchasing shoppers, or 4.4%.
• Observed difference: 0.4 percentage points.
• Relative difference: (4.4% − 4.0%) ÷ 4.0% = 10%.

The point estimate corresponds to 20 additional purchasing shoppers among the 5,000 offered try-on. Random variation can produce a difference like this: the counts alone do not establish a reliable improvement. Assess uncertainty under the experiment's planned analysis rather than stopping when a percentage looks attractive. Also count actual orders and returned units; purchasing shoppers and orders are not interchangeable.

4. Compare contribution with the full incremental cost

Use the same period and currency for benefits and costs. Contribution should reflect revenue retained after discounts and refunds, less the variable costs relevant to those sales: goods, payment processing, fulfilment, return handling and any loss on returned inventory. Add subscription, usage charges and incremental implementation or support costs to the tool side of the comparison. Avoid subtracting a refund or cost twice.

For a simple break-even example, assume an extra order contributes $15 after the applicable return and variable-cost allowance, and the feature costs $75 for the period. Those values are assumptions, not current aisthetix prices or a promised merchant margin.

Break-even extra orders = $75 ÷ $15 = 5.

Five genuinely additional orders cover that assumed cost; six exceed it. Five orders merely attributed to try-on do not establish break-even. If the contribution assumption falls to $5 per extra order, the requirement becomes 15. If it is zero or negative, increasing that type of order does not cover the tool's cost.

For a fuller experiment, compare contribution per assigned shopper across the groups, multiply the difference by the number exposed to the feature, then subtract its incremental cost. This also captures outcomes that a purchase-rate headline can miss, such as more returned units or a different basket mix.

Include returns before making the decision

Wait for the relevant return window and processing delay before treating a cohort as settled. Separate physical returned units from refunds and other adjustments. Shopify's sales-report definitions explain why sales reversals and physical returns are different measures.

If sizing is the suspected issue, use the fit-return formula and blank worksheet. A visual try-on previews appearance; it is not proof that a garment fits. Size guidance depends on the configured chart and available product information. Evaluate those functions against the problem each is meant to address.

Make the next decision explicit

• Little use or frequent failures: investigate visibility, speed and output quality before making a revenue claim.
• Attributed purchases but no credible comparison: you have evidence of usage and association. Improve the evaluation before claiming incremental value.
• An uncertain experiment: report the uncertainty; decide whether a longer, properly planned test is worthwhile.
• Credible improvement that exceeds full cost after returns: consider continuing, and check that performance persists as products and traffic change.
• No worthwhile benefit after a fair evaluation: change the intervention or stop paying for it.

If the feature matches the problem you want to test, the Shopify installation guide explains the setup. For problems a preview cannot solve, start with body and garment measurements or returns management versus prevention.