A fashion ecommerce director sees shoppers use virtual try-on, explore more looks and place orders. The next budget meeting asks a harder question: how many of those orders would have happened anyway? A convincing demonstration and an active experience do not answer that question.
Virtual try-on may help increase sales when visual uncertainty is blocking a purchase, but a brand should test the additional sales it creates before committing to a rollout. For buyers evaluating iKawn Mirror, the useful outcome is a credible comparison with the current shopping experience, followed by a cost and margin decision. This guide proposes that evaluation; it does not report an established Mirror conversion uplift.
Why the sales claim needs a closer look
Deloitte’s survey of 200 retail and consumer-products executives found that 75% considered AI a top strategic priority, while only 16.5% could quantify a return. This is broad consumer-sector research, not a fashion virtual try-on success rate. It nevertheless highlights a relevant budget question: what evidence will justify the next investment? Source: Deloitte, June 18, 2026.
Public retailer examples need the same discipline. In April 2023, Zalando announced a virtual fitting-room pilot covering 22 jeans items and reported that around half of customers tried more than one size on an avatar. The announcement describes engagement; it does not give a controlled sales-lift estimate for that pilot. Its separate size-advice results should not be attributed to virtual try-on. This is a historical example of a different experience, not a Mirror benchmark. Source: Zalando, April 24, 2023.
Ask a vendor what an uplift figure compares, who entered each group, which products were included and whether returns had settled. A large number without those boundaries is a weak basis for your forecast.
Write a sales hypothesis tied to one buying decision
Start with an observed hesitation: shoppers struggle to picture an occasionwear look, or store associates cannot easily show alternatives beyond the garments on the rack. Test whether an optional preview helps those shoppers move toward a purchase. If unavailable sizes or delivery charges are the main obstacle, address them before expecting try-on to solve the problem.
A useful pilot brief might say: “For eligible shoppers browsing our selected jackets, does offering Mirror increase completed orders per shopper compared with the current product journey, without unacceptable costs or returns?” Agree the pilot product range, channel, eligible audience, purchase window and owner before launch. Keep stock, pricing and promotions comparable between the experiences.
Compare the offer of try-on, not only people who use it
Shoppers who choose to use a preview may already be more interested in the product. Comparing their conversion rate with non-users mixes that existing interest with any effect of the experience. Use participation to diagnose adoption; do not make it the sole basis of the sales claim.
For an ecommerce pilot, ask your analytics team to randomly assign eligible shoppers to the current journey or the journey with the optional Mirror offer. Keep each shopper’s assignment consistent across repeat visits within the agreed identity boundary. Count everyone assigned, including people who decline, abandon or encounter a failed preview. This estimates the commercial effect of offering the experience as it actually operates.
The underlying distinction between attribution and incrementality is established in marketing measurement: Google describes randomized controlled experiments as a way to assess lift. Applying that principle to a Mirror pilot is our proposed evaluation approach, not a Google endorsement or a claim that Google Ads measures Mirror usage. Methodology context: Google, October 12, 2020.
In a store, staff assistance and shared displays can blur the comparison. If everyone can see and use the same Mirror, a shopper-level split may be impractical. Discuss assigning comparable stores or time blocks with an analyst, accounting for traffic patterns, promotions and the number of independent groups. One launch weekend versus the previous weekend is directional learning, not strong evidence of causation.
Put commercial outcomes above activity counts
Choose one primary sales outcome before results arrive, then use the rest of the scorecard to explain it. For a first fashion pilot, the following structure keeps the discussion focused:
- Primary outcome: completed orders per eligible assigned shopper, within a fixed purchase window. State how repeat orders and cancellations are handled.
- Economic check: contribution per assigned shopper after product costs, discounts, fulfilment and mature returns, using a cost definition agreed with finance.
- Journey diagnostics: offer seen, preview started, usable preview completed, product handoff and purchase. Confirm which events can actually be recorded in the scoped deployment.
- Operating limits: preview failures, waiting time, staff assistance, complaints and interruptions to the normal shopping journey.
Do not credit every order after a preview to Mirror. Where a store-to-online purchase cannot be linked reliably, show the measurement gap. A QR scan is evidence of a handoff, not evidence of a completed sale.
Read a positive result without overstating it
Hypothetical arithmetic, not an iKawn result: suppose 10,000 shoppers are assigned to each experience. The current journey produces 300 orders and the Mirror-offer journey produces 330. The observed conversion rates are 3.0% and 3.3%: a 0.3 percentage-point difference, or 10% relative lift. The additional observed orders are 30, not all 330.
That arithmetic does not establish a reliable effect. An analyst should report uncertainty and check whether the experiment ran as designed. Agree the smallest improvement worth funding and the required sample before launch; do not stop because a daily dashboard briefly looks positive. Run long enough to cover the planned trading cycle and purchase window, then allow return outcomes to mature.
Finance should compare any credible additional contribution with software, usage, integration, hardware and staff costs relevant to the deployment. Separate one-time setup costs from recurring costs and avoid counting either twice. The Mirror ROI calculator can support scenario planning; measured pilot outcomes should replace planning assumptions before a rollout decision.
What if the brand lacks traffic or measurement capacity?
A smaller retailer may not have enough volume to distinguish a commercially useful lift from ordinary variation within a short pilot. In that case, use a bounded feasibility stage to test product suitability, shopper completion, staff workload and the purchase handoff. Those findings can justify a better-designed next test; they do not justify publishing a conversion promise.
If data quality is weak, repair it before extending the sales experiment. If engagement improves but contribution does not, investigate the handoff and cost structure. If the estimate remains too uncertain, record “inconclusive” and decide whether more learning is worth the expense. Agree a stop, revise or expand decision with the budget owner in advance.
Evaluate iKawn Mirror against your sales question
iKawn Mirror supports live visual exploration of selected looks. Start with a demonstration of your buying journey: choosing an approved product, viewing it and continuing to an associate, physical fitting or product page. Confirm product suitability, supported devices, integration scope and the evidence your team can collect. Do not assume automatic experiment assignment, order attribution or analytics exports are included.
Bring your target category, current conversion baseline, typical order economics and intended channel to the demo. The aim is to establish whether Mirror addresses a real visual-discovery problem and whether a scoped pilot can produce evidence useful enough for your next investment decision.