.svg)
.svg)
.svg){kind=icon}
.svg)
.png)
Our Apps Have Driven
5X AOV | 2X Conversions | $30M+ Additional Revenue
5X AOV | 2X Conversions | $30M+ Additional Revenue
.svg)
5.0 (900+ Reviews)
Shopify Apps
.svg)
Most Shopify brands bundle on instinct, not data. The Bundle Recommendation Report analyzes your actual order history to tell you what to bundle, which format to use, and how to price it.
The engine scores every product pairing across four signals: co-frequency, support, confidence, and lift. It then classifies each pair into a relationship type (variant pairs, routine steps, anchor-accessory, theme, collection, or size upsell) and recommends the right bundle format for each.
What you receive: prioritized bundle moves ranked by signal strength, revenue projections from your actual basket data, customer cohort breakdown, full bundle specs, and a placement map for every recommendation.
Two ways to start: free diagnostic (no order data needed) or full report with order history access.
Works with Easy Bundle Builder or any other platform. The strategy is standalone.
What should your store bundle? Specifically: which of your products have a real reason to go together, what kind of bundle each grouping should become, and what offer should sit on top.
That’s the question the Bundle Recommendations Report answers. We point a proprietary AI bundle recommendation engine at your store and hand you a complete bundle strategy, built for your catalog and the way your customers actually shop. You don’t need to have installed Easy Bundle Builder, and you don’t need to know which bundle format fits which products. The engine works that out. Here’s how it works.
Bundle Recommendations Service is an AI-Powered report that answers three questions for your store:
1. What to bundle. Which products in your catalog have a real reason to go together? Not just pairs, but the full grouping that holds up as a bundle.
2. What kind of bundle to build? Quantity break, frequently-bought-together prompt, fixed product bundle, mix-and-match builder, subscription bundle, or gift kit. The right format depends on the products and the customer, and the engine chooses it for each grouping.
3. What offer to attach. A discount range and structure that fits the bundle type and the data behind it, including where no discount is the right answer.
Most bundle advice stops at the first question, and even then, it’s a guess. The report answers all three, grounded in your store’s own data & our experience handling thousands of stores over the years.
Most Shopify brands run bundles. Almost none run them at full potential.
Every merchant we talk to wants a higher average order value, and they know bundles are one of the most reliable ways to get there. The trouble is what happens after they decide to try. These are the four things we hear most often:
"We launched a bundle six months ago. It got traffic, almost zero add-to-carts. Now it just sits there."
"I threw three of our bestsellers together at ten percent off. I thought that was a bundle. Customers completely ignored it."
"I have no idea which products my customers naturally buy together. I’m just guessing at this point."
"We installed three different bundle apps. They conflict with each other and our theme is broken on mobile."
Every one of these is the same underlying problem: bundling done on instinct instead of evidence. The store owner, the marketing manager, the e-commerce manager, the CRO lead, the agency running a store on someone else’s behalf, everyone in that chain knows bundles work. Almost nobody has a confident, data-backed answer to "which bundle, with which products, at what offer?"
That’s what this report exists to solve.
There’s a reason this comes from the Easy Bundle Builder team and not from somewhere else.
We’ve spent years working hands-on with Shopify merchants of every shape and size, across supplements, apparel, outdoor gear, baby products, food and beverage, beauty, and home goods. Every category, every catalog size, every kind of buyer. That experience & data is what’s built into the engine. We’ve seen which bundle types convert and which fall flat, how a routine-driven catalog behaves differently from a one-big-purchase catalog, and when a discount drives real behaviour versus when it just cheapens a premium brand.
Most "AI bundle" tools have one of two things behind them. Either a generic large language model guessing at what makes a good bundle, or a basic frequently-bought-together query dressed up in modern language. Neither carries any real merchandising judgment.
What we built is different. It’s proprietary, built in-house from the ground up. We’ve iterated enough. The first generation was built to fill the gaps every off-the-shelf approach leaves behind. The current generation adds a layer of AI-powered classification that lets the engine reason about why two products belong together, not just whether they happened to sell together.
And the part that matters most: the recommendations are based on deep analytics of your own store data. The engine reasons over your store’s own order history, your products, your customers, your patterns. We bring the accumulated judgment of working with many merchants; your store brings the specifics. The recommendations come from the combination.
This is the engine’s logic, in plain English. Every customer-friendly name has the internal technical name in brackets right after it, so anyone going deeper into our documentation can map back cleanly.
Every store has its own pattern. The engine reads your catalog size, your shopping behavior, your industry, your typical order shape, and adapts the recommendations to match. A snack brand doesn’t bundle the same way a skincare brand bundles. A skincare brand doesn’t bundle the same way a furniture store bundles. The engine figures out what kind of store yours is first, and lets that shape everything downstream. (Internally: store-shape classification + industry classification with a confidence score.)
Before the engine recommends anything, it does the kind of analysis a merchant could never do by hand. It takes every possible pairing of products in your catalog and scores each one across several mathematical dimensions:
• How often the two products end up in the same order (co-frequency)
• What share of multi-item orders contain both (support)
• How predictive each product is of the other, calculated in both directions because the answer is different each way (confidence A to B and confidence B to A)
• How the actual co-occurrence compares to what pure statistical chance would produce (lift)
All four signals are combined into a single composite score that weighs them against each other. The composite is built so that no single dimension can dominate the ranking. A pairing with high frequency but weak lift won’t outrank a pairing with both. A pairing with high lift but very few observations won’t outrank a steady pattern with many.
Sitting on top of the scoring is a multi-threshold gate. Each pairing has to clear minimums on co-frequency, support, and lift simultaneously before it advances. Pairings that don’t clear the gate stop here. The ones that do move forward. (Internally: co-frequency, support, confidence in both directions, lift, composite score, strength gate with multi-dimensional thresholds, hybrid rescue.)
Once a pairing has cleared the math, the next question is what’s actually going on between those two products. Because the answer changes what kind of bundle they should sit in.
Two products ending up in the same order can mean very different things. Years of working with how customers actually shop surfaced a clear pattern: those reasons cluster into six distinct, repeatable types. The engine then sorts each pairing into one of them:
• Variant pairs (internally: peer_variant). Same product, different scent or color or flavor. "Cedar Bar Soap" and "Pine Bar Soap." The bundle here is "try the set."
• Routine steps (internally: routine_step). Products used in sequence. Cleanser, toner, moisturizer. The bundle here is the routine.
• Anchor and accessory (internally: anchor_accessory). A main product and its add-on. DSLR camera and lens cap. The bundle here is "everything you need to get started."
• Theme pairs (internally: theme_line). Different product types unified by a theme. Dog leash and dog bowl. The bundle here is the kit.
• Collection pairs (internally: collection_member). Products the merchant has already grouped together. Summer Collection Tote and Summer Collection Hat. The bundle here matches what the merchant already intended.
• Size upsells (internally: upsell_larger_format). Same product, bigger pack or larger size. The bundle here is "go bigger," and only in that direction.
Different relationships call for different bundle formats. Routine-step products want to be packaged as a fixed bundle (cleanser + toner + moisturizer at a single price). Variant pairs want to be quantity breaks (buy two scents, save fifteen percent). Anchor-and-accessory pairs want to be frequently bought together, prompting on the anchor’s product page. Theme pairs and collection pairs want to be mix-and-match builders ("pick three of these four"). Reusable products want subscription wiring. Gift-driven products want pre-curated gift kits at clear price tiers.
The engine doesn’t dump every pair as a frequently-bought-together suggestion and call it a day. It chooses the format that fits across the full range of bundle types.
This is the step most engines skip. A statistical co-occurrence is not a bundle. Two products bought in the same order could be a real bundle (phone and case) or a coincidence (sunscreen and a dog leash, because the store sells outdoor gear).
Every candidate bundle has to pass a sense-check: would a knowledgeable merchant confidently put this on a product page? Groupings that wouldn’t make sense as bundles get dropped here. (Internally: coherence test / LLM quality gate.)
If a camera and a lens cap are bought together, the camera page should recommend the lens cap. The lens cap page should not recommend the camera. The engine figures out which product is the trigger and which is the companion, and only places the recommendation where it makes sense. (Internally: direction assignment.)
For bundles that work both ways, like variant pairs, where shoppers looking at either scent might want the other, the engine intentionally emits the recommendation on both product pages. (Internally: bidirectional emission.)
If a store sells "Vanilla Candle Small" and "Vanilla Candle Large," those shouldn’t show up as two separate recommendations on the same product page. The engine collapses them into one and keeps the slot for a genuine complement. (Internally: variant deduplication.)
Some products co-occur with almost everything in the store. Cheap add-ons, popular best-sellers, free gifts. Lazy engines recommend the same five generic items on every product page. Ours demotes them, so each product page gets recommendations that feel specific to that product, not a copy of the store’s overall best-seller list. (Internally: portfolio diversity.)
A five-dollar sticker recommended alongside a five-hundred-dollar camera looks like a mistake. The engine flags pairs that are too far apart in price unless the data behind them is overwhelming. (Internally: price compatibility filter.)
Every recommendation comes out with a confidence score. The engine doesn’t just produce recommendations; it knows which ones are strong, which ones are borderline, and which ones it wouldn’t stake its reputation on. Borderline recommendations are demoted or held back. (Internally: confidence scoring / self-grading.)
When a product has no purchase history, the engine doesn’t go silent. It falls back through tiers (same collection first, then similar product titles, then category and price range) so every product page has at least three recommendations from day one. The confidence label tells you what’s data-backed and what’s a thoughtful starting point. (Internally: cold-start handling / fallback layer.)
Get Your Bundle Recommendation Report Now
The output isn’t a dashboard you have to interpret. It’s your Bundle Revenue Strategy: a written plan, built for your store, that a founder or a marketing lead can read in one sitting and hand straight to whoever implements it.
A typical report includes:
• A decision brief. The handful of moves to ship first, ranked by signal strength and ease of execution. If you read nothing else in the report, read this.
• A diagnosis. The specific revenue leaks in your store today. The anchor product with no cross-sell. The bundle that excludes its own strongest partner. The subscription rail that isn’t wired to your bundles. The accessory line has a zero percent attach rate.
• Your customer cohorts. The two or three distinct ways people shop your store, because a bundle that works for one cohort actively annoys another. Each recommendation is tagged to the customer it serves.
• The bundles themselves, across every type. Frequently-bought-together prompts, build-your-own fixed pre-curated kits, mix-and-match builders, quantity breaks, subscription bundles, and gift kits. Each one is grounded in your data, tagged to a customer type, a placement, a directionality flag, and the evidence behind it.
• A placement map. Where each bundle should live: on which product page or on a dedicated landing page.
• Revenue framing. Once you’ve shared order data, bottom-up projections are grounded in your actual basket patterns. No invented numbers, and no projections at all until the data supports them.
Every bundle in the report can be built with Easy Bundle Builder, but the strategy stands on its own. It’s a plan, whether you implement it with us or not.
It starts with a free diagnostic. We audit your live storefront and show you the bundle opportunities visible from the outside. No order data required, no commitment. If the direction looks right, we go deeper: with your order history, the engine produces the full Bundle Recommendations Report, grounded in your actual basket patterns. From there, you implement it with Easy Bundle Builder, with our help, or on your own.
Most brands run bundles. Almost none run them at full potential. This report how you close that gap.
Get your Bundle Recommendation Report Now
.avif)
