Put an add-on on the item page when it belongs closely to the item a guest is choosing, such as extra sauce for a particular entrée. Consider the cart when the offer makes sense across the order, such as a shared side. Then compare the placements with equivalent eligible traffic before making the choice permanent.
The best location is not necessarily the one that gets the most clicks. A prompt should be relevant, easy to dismiss, and unlikely to interfere with finishing the order. It also needs to produce orders the kitchen can fulfill correctly.
Map what the guest knows at each point
On an item page, the guest is deciding what goes into one dish. They can see its name, price, and available choices together. An optional topping or sauce is easy to understand there because the connection is immediate. The same placement can feel cluttered if the guest must work through several prompts before adding a simple item.
In the cart, the guest can see the order as a whole. That gives a shared side or extra drink more context, but it is also the point where many guests want to check the total and pay. A late prompt that hides the checkout button or resets the cart can create friction at a costly moment.
Start with one specific offer and write down the decision inputs:
- Relevance: Which items or combinations make the add-on useful? Your menu lead or kitchen team should define these pairings.
- Price and cost: What will the guest pay, and what ingredients and packaging does the add-on use? The owner or manager supplies these numbers.
- Preparation: How should the add-on appear on the kitchen ticket, and can the team fulfill it during a rush? Ask the people preparing and packing orders.
- Placement controls: Can the ordering system show the same offer at either point, limit who sees it, and report the results? Ask the software provider to demonstrate the setup.
- Guest experience: Can someone decline the offer and continue without losing choices? Check this on a phone as well as a larger screen.
Your restaurant website and ordering flow should make the menu easy to use before you add promotional prompts. If you are also reviewing the ordering system itself, the online ordering buying guide covers a broader set of purchasing questions.
Compare relevance with interruption
Imagine a restaurant deciding where to offer an extra dipping sauce with a chicken entrée. On the item page, the guest sees the offer while choosing the entrée. That placement keeps the sauce beside the relevant food and lets the guest decide before moving on. It is a sensible starting point if the sauce is useful mainly with that entrée.
A cart offer can remind a guest who skipped the option earlier. But the cart must identify orders containing the eligible entrée; otherwise, it may offer the sauce to someone buying only a salad. The restaurant should also decide what happens when the cart contains two eligible entrées: can the guest choose two sauces, and is the quantity clear?
Now consider a family-size side intended to accompany several dishes. The cart may give the guest a better view of whether the order needs another side. An item-page prompt attached to only one entrée could miss eligible orders or suggest that the side belongs to one person rather than the table.
Keep the offer itself consistent when comparing placements. The name, price, image, quantity rules, and description should stay the same. If one version changes both the location and the deal, you will not know which change influenced the result.
Also ask how the offer behaves when an item is removed, a quantity changes, or the guest returns to the menu. A sauce should not quietly remain in the cart after its associated entrée disappears unless that is an intentional, clear choice. The manager should agree on the desired behavior before the provider configures it.
Test one placement against the other
First run a simulation in a safe practice environment. Walk through adding and declining the offer, changing quantities, removing the eligible item, and checking the kitchen-facing order. Use test data and confirm that the simulation cannot take a real card payment, send a guest message, or dispatch a courier. Fix confusing behavior before exposing the offer to guests.
For a live service pilot, ask whether the provider can assign eligible visitors consistently to one of two versions: item-page placement or cart placement. Each visitor should remain in the same version during the visit. Run both versions at the same time, using the same menu, price, service area, hours, and device mix as far as practical. Do not put one version on weekday lunches and the other on weekend dinners and treat the results as equivalent.
Define eligibility before counting. For the sauce example, an eligible visit might be one in which a guest views the qualifying entrée. A cart offer can only appear if that entrée reaches the cart, while an item-page offer can appear earlier. Record both eligible visits and actual prompt views. That distinction shows whether the location changes exposure as well as whether guests accept the offer.
If the system cannot split traffic or report these measures, ask what it can export and how an equivalent comparison could be arranged. Alternating placements by time period may be useful for exploration, but changes in demand, promotions, and staffing make the result less certain. Do not call it a controlled comparison.
Agree in advance on the pilot length, who checks the numbers, and what would stop it. Stop or correct the setup if orders are being misrouted or guests cannot complete checkout. A busy shift should not become an experiment the kitchen has to untangle.
Measure completed contribution and errors
Track the add-on through to a completed order, not just a click on the prompt. Review eligible visits, prompt views, add-ons accepted, completed orders, add-on refunds, and kitchen or packing errors. Compare checkout completion as well: an add-on gain is less useful if the placement appears to discourage guests from finishing their orders.
Hypothetical example: A restaurant assigns 1,000 equivalent eligible visits to each placement during the same pilot. Each group has 400 completed orders. The sauce costs the guest $5; ingredients and packaging cost the restaurant $2 per sauce. The table shows completed add-ons before later refunds. It assumes equal offer terms and no other changes.
| Measure | Item page | Cart |
|---|---|---|
| Eligible visits / completed orders | 1,000 / 400 | 1,000 / 400 |
| Completed add-ons | 60 | 50 |
| Add-on sales / ingredient and packaging cost | $300 / $120 | $250 / $100 |
| Later add-on refunds | 2 at $5 = $10 | 1 at $5 = $5 |
| Illustrative add-on contribution | $170 | $145 |
The calculation is sales minus ingredient and packaging cost minus refunds: $300 − $120 − $10 = $170 for the item page, and $250 − $100 − $5 = $145 for the cart. The example assumes refunded sauces cannot be recovered. Ingredient and packaging costs are included as variable costs; fixed costs, payment processing, and other variable costs are excluded. Refunds are included as an exception cost, but replacement work and other error costs are excluded. These are not all-in operating totals.
Under these assumptions, the item page is the better candidate for a longer trial: it produced $25 more in measured add-on contribution while both groups had the same number of completed orders. That is not proof it will always win. The groups are small, and the table does not show whether the placements changed sales of other items or created different amounts of staff work. Check prompt views and actual error records before deciding.
For a fuller view, compare the contribution of the entire completed order, including any change in other items and the incremental costs your records can support. The related buying guide explains how contribution differs from sales when comparing ordering channels. For placement, keep the channel and offer terms consistent so the question remains focused.
Use this checklist on your current setup: choose one eligible offer, confirm the desired cart and ticket behavior, and identify the reports you need. Then ask Fleksa about restaurant online ordering, the placement controls available for your setup, and verified US availability.
Frequently asked questions
Is the item page always better for add-ons?
No. It is often a good starting point for an option tied closely to one item. A cart offer may fit better when the guest needs to see the whole order to judge whether the extra is useful.
What if guests see the item-page offer but never reach the cart?
Count eligible visits and actual prompt views separately. The item-page placement may reach more people because it appears earlier; measure completed orders and contribution too, rather than comparing acceptance rates alone.
How long should a placement test run?
There is no universal duration. Include a representative mix of service periods, set a review date before starting, and avoid drawing a firm conclusion from a handful of orders. If the result is close or traffic is limited, continue observing or treat the choice as provisional.
