A bundle is a pricing decision wearing a merchandising costume, and the number it is usually judged on hides whether the decision was any good. Average order value goes up almost automatically when you sell more items together. Gross profit per order does not, and that is the number the bundle should be built to move.
The failure mode is specific and common. A store takes its best seller, attaches something that is not selling, and puts a discount on the pair. Average order value rises, the report looks like a success, and the store has just paid margin on its strongest product to move stock it should probably have written down.
We build bundles the other way round. Start from what people already buy together, work out what the set costs you, decide what the discount is buying, and only then set the price. Done that way a bundle raises the order and the profit at the same time, which is the only version worth the operational effort.
Start with the baskets you already have
Your order history has already told you which bundles will work. Export a year of orders, look at every order with more than one item, and count the pairs that appear together most often.
That list is worth more than any brainstorm, because those combinations were chosen by customers without any prompting. They are combinations that make sense in the buyer's life rather than in your category structure, and they are frequently surprising.
Then check the other direction. For your highest-traffic products, what do people who bought that item buy next, and how long afterwards. A pairing that reliably happens six weeks later is a bundle waiting to be offered on day one, or a replenishment message, or both.
For products with too little history, use the support inbox instead. The question what else do I need with this is asked constantly and answering it in a bundle is simply making the answer purchasable.
The reason this matters is that a bundle of things people already combine needs almost no discount to sell. The convenience is the offer. A bundle of unrelated products needs a large discount, which is another way of saying it is not really a bundle.
A bundle of things people already combine needs almost no discount. The convenience is the offer.
Price from the cost of the set, not from the sum of the labels
Once you know the combination, work out the gross margin of the set as a whole. Add up what the items cost you, add the extra fulfilment cost or subtract the saving if they ship in one box, and compare that to the price you intend to charge.
That single calculation prevents the most common error, which is applying one discount across every bundle regardless of what the items cost. A set with a rich margin can carry a real saving. A set containing a thin-margin item cannot, and offering one anyway means you have sold your most fragile product at a loss with extra steps.
A worked example to show the shape. Say product A sells at $50 and costs you $15. Product B sells at $30 and costs you $21. Sold separately, the pair returns $44 of gross profit on $80. Take a tenth off the pair and you charge $72, so gross profit is $36. The average order rose from either item alone, and you gave up $8 to get there. Whether that is worth it depends on whether the bundle created a second purchase or simply repackaged one.
So set the discount against the item that has room, and where nothing has room, sell the bundle on convenience with no discount at all.
Shipping is where the margin quietly goes
The cost that ruins bundle maths is rarely the product cost. It is the fulfilment.
A bundle that ships as one parcel is genuinely cheaper to send than two orders, and that saving can fund the discount without touching product margin. A bundle that ships as two parcels from two locations costs more than the two separate orders would have, and no pricing cleverness recovers it.
So check the physical reality before the offer goes live. Do the items live in the same place. Do they fit in one box. Does the combined weight cross a carrier price break. Does the bundle push the order over your free shipping line, and have you counted that as part of the discount, because it is.
Heavy and bulky combinations are where this bites hardest. Two light items bundled together are usually a gift to your margin. A bundle including anything large should be modelled before it is priced.
The same check applies to returns. A bundle that gets returned as a whole because one part did not suit is an expensive event, so make the return terms for a set explicit before you sell one.
Where the bundle should appear
Placement decides whether a bundle raises the order or replaces it, and this is the part stores get wrong most often.
On the product page, a bundle should be an upgrade to the choice being made, presented alongside the single item with the difference in price and in value shown plainly. Presented well it moves people up. Presented as a separate product with its own page, it competes with the single item and cannibalises it.
At the cart, a bundle offer should be a completion, not a new decision. Something that finishes what is already there, at a price small enough not to reopen the whole purchase. One suggestion, relevant, cheap relative to the cart.
After the order, on the confirmation page and in the first post-purchase email, a bundle is often the easiest sale of the lot, because the buyer is already committed and shipping can be combined if you can get it into the same dispatch.
What rarely works is a bundles collection page. Almost nobody browses for a bundle. They buy a bundle at the moment they are buying the thing inside it.
The numbers that say it worked
Judge a bundle on four numbers together, and the first one is not average order value.
Gross profit per order for the period, compared with before. If the bundle is doing its job this rises. If average order value rose and this fell, you sold a discount.
Items per order, which tells you whether the increase came from behaviour or from price. An average order that rises while items per order is flat means something else changed.
Attach rate, meaning the share of orders containing the bundle out of orders containing the anchor product. This is the number to optimise on the page.
And return rate for orders containing the bundle, watched for a couple of months, because a set with one weak component shows up here before it shows up anywhere else.
One warning about attribution. If the bundle simply replaced two separate items that were often bought together anyway, your gross profit will fall while every activity metric looks healthy. Compare the pairing rate before the bundle existed with the bundle rate after, and be honest about which customers you actually changed.