A larger order is useful when it meets the customer’s needs and leaves the store with a worthwhile margin. An AI agent can contribute by explaining why a more suitable product or a relevant additional item fits the request. Pushing the most expensive option into every conversation is a different strategy, and an easy way to lose trust. Start with relevance, then measure whether basket value actually changes.
Define AOV before comparing results
Average order value expresses order revenue divided by the number of orders for a chosen set of transactions. Decide which revenue measure you will use and apply it consistently. Discounts, returns, shipping and taxes can change the interpretation, so do not combine totals from different reports without checking their definitions.
Here is a simplified illustration using merchandise revenue after discounts: 100 orders worth €6,000 produce an AOV of €60. If another comparable group has 100 orders worth €6,400, its AOV is €64. This is arithmetic, not an Obsedia performance result, and it does not tell you whether profit or customer satisfaction improved.
Start with the product that fits the request
A customer who needs a laptop bag with a separate clothing compartment may find a larger bag more suitable than the cheaper compact version. The agent should explain that difference using documented dimensions and compartments. A higher price is justified by the customer’s requirement, not by a hidden instruction to maximize the basket.
Obsedia’s Ecommerce Agent can compare catalog products and show their product cards. Give it useful attributes and test whether its explanation changes when the shopper’s needs change. It should also be willing to recommend the lower-priced item when that is the appropriate match.
Offer an addition only when it serves the purchase
A matching organizer can make sense for someone buying a travel bag. A care product might be relevant if the manufacturer documents that it is suitable for the material. The relationship needs evidence: visual similarity alone does not establish compatibility.
In an illustrative conversation, the customer could ask, “Is there a smaller pouch that fits inside this tote?” The agent can compare the dimensions if the catalog supplies them and present the relevant item. Do not imply that Obsedia automatically bundles products, applies a discount or inserts an extra item without the shopper’s action.
Make the price and choice explicit
Present the product and price clearly. If there are variants, show which one the amount belongs to. A shopper who agrees to a bag has not agreed to an accessory, even when the accessory was mentioned in the same answer.
Obsedia’s supported cart action is explicit: the customer chooses an eligible product variant and adds it. Verify that the resulting cart contains only the items and quantities selected. This protects both the shopping experience and the validity of any basket-value measurement.
Measure margin and completed purchases alongside AOV
AOV can rise while the total number of completed orders falls. It can also rise because a heavily discounted bundle contains more items but earns less contribution. Read the basket-value result alongside completed orders, discounts, returns and product contribution.
Separate recommendation quality from purchase attribution. Inspect whether suggested items fit the request and whether the customer used the product action. To estimate incremental business impact, use comparable visitor groups or another carefully designed test; a purchase after a chat does not by itself prove the chat caused it.
Pilot one product relationship
Choose one main product and one genuinely useful companion item. Document the relevant relationship, such as dimensions or intended use, and test the questions a shopper would ask before buying both. Include a case where the companion is unnecessary or unavailable.
Review the answers before expanding to a whole collection. If the recommendation feels forced, adjust the source or the conversation approach. The goal is to help the customer buy what suits the task; any AOV improvement should follow from that better decision.
Use product knowledge to explain a better fit or a useful addition. Measure completed orders, contribution and returns before treating a higher AOV as success.
