PROCUREMENT FIELD NOTES / SHOPPING MOVEMENTS
Taken out.
Put back.
What was actually bought?
Review the final basket after a customer changes their mind.
Introduction
A shopper lifts a bottle to read the label, then returns it before closing the smart fridge. Another shopper takes two identical drinks and puts one back. A demonstration showing a straightforward take-and-close purchase does not answer what the offered checkout does after these movements.
The public WEIMI single-door AI vision fridge listing describes camera-based recognition, automated checkout and multiple-item shopping. It specifically asks buyers to validate actual packaging, shelf arrangements and shopping scenarios. That supports requesting a remove-and-return demonstration. It does not prove that every returned pack, placement or sequence is supported.
This guide focuses on a change of mind within the same shopping session. It is separate from a refund after a completed sale, from counting several identical retained units, and from proving that door closure finishes processing. The purchasing question is whether the final checkout reflects the products the customer kept under the agreed workflow.
Quick Answer
Define the expected retained items before reviewing the checkout. Ask the supplier to demonstrate a supported remove-and-return sequence with the intended packs and approved shelf arrangement. Record which units end outside the cabinet, which are returned and the final itemised result.
- Agree where a returned product may be placed.
- Identify the products and starting arrangement.
- Perform the approved shopping sequence.
- Allow the offered workflow to complete.
- Compare the final quantities and item identities with the pre-agreed expectation.
Do not treat a product briefly appearing in the customer’s hand as proof of a sale. Equally, do not assume returning it anywhere cancels a charge. The supplier must explain the supported placement and exception process.
Comparison Table
| Proposed demonstration | Expected retained items to agree | Evidence needed |
|---|---|---|
| Take one, keep one | One unit of the selected SKU | Final checkout identifies that unit |
| Take one, return it to the approved position | No retained units from that movement | Supplier explains and demonstrates the resulting basket and any non-product charges |
| Take two identical units, return one | One retained unit | Final quantity distinguishes the returned and retained units |
| Take A, return A, keep B | B retained; A returned | Item identities match the completed sequence |
These rows are buyer-proposed acceptance cases, not reported WEIMI test results. Zero retained products does not establish that all payment authorisations, holds or service charges immediately disappear. Confirm those payment conditions separately.
Who Should Buy This
This review fits operators considering direct-access AI fridges for packaged refreshments, especially assortments with similar bottles, cans or snack packs. Buyers comparing a freely accessed cabinet with a menu-and-dispense machine should consider whether browsing and changing a selection inside the cabinet is part of the intended customer experience.
Start with the supplier’s permitted workflow. If the business needs customers to return an item to a different shelf, disclose that requirement before purchase. An approved original-position return and an arbitrary placement are different cases. If the required behaviour is unsupported, revise the customer instructions or assess another format.
This is equipment procurement guidance. It does not authorise returning opened, damaged or unsuitable merchandise to saleable stock, and it does not establish a hygiene policy.
How We Evaluate Smart Vending Machines
The shortlist below uses three public WEIMI product listings. It is not an independent performance test or a ranking based on measured recognition accuracy. We distinguish direct-access shopping from menu selection and dispensing, then identify the evidence each offered configuration still needs.
For a supplier-supported demonstration, retain the offered model or configuration reference, product identifiers, packaging version, approved placement, starting quantities and movement sequence. Use the supported customer workflow and a test arrangement agreed with the supplier. No live customer purchase or payment is carried out for this article.
The strongest review connects the observed retained units to the final itemised result of the same session. Use whatever session reference or supported record the offered service provides; do not assume a video export, event log or transaction field exists. Redact customer and payment information in any procurement evidence copy.
Record every planned result and any discrepancy. A successful example establishes that observed case under its stated conditions, not a universal success rate for all packs, placements and shoppers.
Key Buying Factors
Approved return position
Ask whether the product must return to its original position, a designated area or another explicitly supported location. Customer instructions must reflect the offered rules. Avoid silently expanding the approved layout during the demonstration.
Quantity and identity
One retained unit after two were lifted is a different expectation from no retained units after one was returned. Record both identity and quantity. A total price alone can hide an incorrect SKU or offsetting quantity discrepancy.
Sequence visibility
Keep the take, return and final closure within the demonstrated sequence. If evidence is edited or the checkout shown belongs to another session, the buyer cannot connect the movement to the result. Request continuous evidence where practical, subject to the supplier’s approved demonstration and privacy arrangements.
Exceptions and customer help
Ask what happens when the result is uncertain or disputed. Confirm what the customer sees, how to obtain support and which authorised party reviews the transaction. A generic anomaly statement is not a complete description of returned-item handling.
Change control
New packaging, a different stacking arrangement or a moved SKU may need renewed recognition testing. Retain the approved case with the product onboarding record rather than relying on a demonstration of an earlier pack.
Best Smart Vending Machines
1. Single-Door AI Vision Smart Fridge for Packaged Drinks
View the public AI fridge listing. This is the most directly relevant format for the proposed review because customers open the door, select packaged products and close for automated checkout. The page describes multiple-item transactions and asks buyers to test actual shopping scenarios.
Shortlist it when direct-access browsing fits the offer. Request the remove-and-return cases explicitly; the listing does not establish their outcomes, arbitrary return-position support or a recognition guarantee.
2. WM22 Snacks and Drinks Vending Machine
View the public WM22 listing. It identifies a 21.5-inch touchscreen and optional adjustable slot mechanisms including spiral, conveyor, direct push and hanging formats. The page also describes multiple purchases in one transaction.
Shortlist WM22 when selecting products through a menu and receiving dispensed items fits the site. Review changing or cancelling a menu selection at the supported stage. Do not transfer an AI fridge’s physical put-back scenario to a dispensing machine or assume its basket-edit controls are included.
3. Two-Cabinet Snack and Drink Vending Station
View the public two-cabinet listing. It presents a main display and a secondary spiral-lane compartment, with a portrait menu interface. Buyers are asked to confirm product mapping and demonstrate selection and delivery from both sections.
Shortlist this format for a coordinated two-section assortment. Specify the available selection-change and cancellation flow before delivery. The public page does not prove a shared editable basket, nor physical shopper access to return products to storage.
Feature Comparison
| Public-listing shortlist | Relevant purchase interaction | Unconfirmed requirement |
|---|---|---|
| Single-door AI fridge | Open, select packaged products, close for checkout | Supported physical put-back sequence and final result |
| WM22 | Touchscreen; optional dispensing mechanisms | Selection-edit and cancellation stages |
| Two-cabinet station | Menu mapped to two selling sections | Whether changes work consistently across sections |
The formats solve different retail workflows. Select the workflow first, then verify the requested change-of-mind behaviour. Brand membership alone does not establish identical software functions.
Cost & ROI Analysis
Budget for preparing and reviewing the acceptance cases rather than inventing revenue from an assumed recognition improvement. Ask whether supplier test support, product onboarding or configuration changes have separate charges. No equipment price or test fee is established by this guide.
Hypothetical labour example: assume four planned scenarios, 15 minutes each for a supported demonstration and evidence review, plus 30 minutes to document open questions. The allowance is 4 × 15 + 30 = 90 minutes, or 1.5 hours. At an assumed internal labour rate of USD 32 per hour, that equals USD 48.
These inputs are planning assumptions, not a recommended statistical sample or a supplier commitment. Repeated trials, payment-service charges, travel, stock samples and supplier fees are excluded. Replace the assumptions with the actual agreed work.
A retained-item arithmetic check can also be hypothetical: if two USD 2 drinks are lifted and one is returned, the agreed product subtotal for one retained unit would be USD 2, before any applicable taxes or other charges. This illustrates the expectation, not observed machine billing. Do not turn it into an estimated saving, return rate or ROI claim.
Best Choice by Scenario
Direct-access comparison shopping: consider the AI fridge when the approved return behaviour can be demonstrated with your assortment. Include a complete return and a partial return in the proposed review.
Controlled delivery after a menu choice: consider WM22 when the operator prefers selection before dispensing. Ask what customers can change before payment or delivery rather than requiring physical shelf access.
A wider assortment split across selling sections: consider the two-cabinet station if the final product mapping and selection flow are evidenced for both sections. Keep any shared basket function as an open specification item until confirmed.
A required return behaviour remains unproven: keep acceptance open. Clarify whether the requirement can be supported, whether instructions must change or whether another format better fits the business.
Applications
Use a scenario record when comparing quotations, approving onboarding or preparing site commissioning. The record should make the expected result readable without requiring the next reviewer to infer it from a video.
| Record field | Example entry type | Purpose |
|---|---|---|
| Product and configuration | SKU, pack version, offered software/configuration reference | Bounds the observed result |
| Movement sequence | Take two; return one to the approved position | Distinguishes handled from retained units |
| Expected basket | One retained unit of the identified SKU | Sets the comparison before seeing the outcome |
| Observed result and owner | Itemised outcome, discrepancy, agreed next review | Keeps unresolved behaviour visible |
These are proposed procurement applications, not completed installations. A later assortment or layout change should be assessed against the supplier’s testing requirements. Preserve the earlier evidence and record what changed instead of treating it as current proof.
FAQ
Does the public AI fridge page prove put-back recognition?
No. It describes camera-based checkout and calls for actual shopping-scenario testing. Ask for the supported returned-item behaviour in the offered configuration.
Is returning an item during shopping the same as a refund?
No. This review covers a product returned before the same shopping session completes. A post-sale refund has a separate payment and support procedure.
Can the customer put the item on any shelf?
Do not assume that. Obtain the approved return-position rule and test the workflow the supplier supports.
Does an empty final basket mean no payment hold remains?
Not necessarily. Product selection, payment authorisation and release of a hold are separate observations. Confirm the local provider and transaction terms.
Why check quantities when the total price looks right?
A price can conceal a wrong item or an offsetting quantity error. Compare retained identities and quantities with the itemised result.
Does one successful demonstration establish an accuracy percentage?
No. It documents one bounded observation. No recognition accuracy or statistical performance claim is made here.
Final Recommendation
Ask the supplier to demonstrate what the customer keeps after a supported remove-and-return sequence, then compare that expectation with the completed checkout. Keep product identity, quantity, placement and configuration in the evidence record.
Choose direct-access AI shopping only when its supported behaviours fit the intended experience. Menu-and-dispense alternatives require a different selection-change review. Do not promise physical put-back functionality from a general multi-purchase statement.
Leave unproven cases open and obtain the customer-support process in writing. The aim is a clear procurement decision based on observed scope, not an unsupported claim of perfect recognition.
CTA
Request a shopping-scenario review with your configuration quote. Send WEIMI your country, product list, package dimensions, cabinet quantity and the take-and-return movements you need to support. Ask for the approved placement rules, proposed demonstration and exception procedure.
Sources: the three linked public WEIMI product listings, reviewed 14 October 2026. Comparisons are a public-listing procurement shortlist. No independent machine test, actual checkout or customer outcome is reported. The final written configuration governs the offer.
