# Market Basket Analysis for Ecommerce: Products Bought Together and Pair Strength

Market basket analysis for Shopify brands: ask Luca AI in chat which products your customers buy together, and each pair comes back with its support, confidence and lift.

- Applies to: Every brand that connects its store; basket analysis is calculated as standard
- Status: Live
- Verified: 2026-11-07
- Canonical page: https://ask-luca.com/product-docs/customers/products-bought-together

Market basket analysis tells you which products your customers already buy in the same order, and how strong each pairing is. Run it before you build a bundle, and you commit stock to pairs that genuinely sell together, not pairs that only look right on a mood board. Luca AI, an AI analyst for ecommerce brands, answers the question in chat: ask which products are bought together, and each pair comes back with its support, confidence and lift.

## At a glance {#at-a-glance}

| | |
|---|---|
| **What it answers** | Which products your customers buy together, pair by pair |
| **How each pair is scored** | Support, confidence and lift, the standard measures of market basket analysis |
| **Where you get it** | In chat. You ask the question in plain English and the pairs come back in the answer |
| **What it's built from** | Your store's orders, with your full Shopify history imported when you connect |
| **Who gets it** | Every brand that connects its store. Basket analysis is calculated as standard |
| **How fresh it is** | Data refreshes once a day, overnight |
| **Taking it further** | Ask follow-ups in the same thread, ask for a chart, or download the data as CSV or Excel |
| **What it never does** | Build the bundle or change your store. Luca reads your data and you decide |

## What market basket analysis in Luca is {#what-it-is}

Market basket analysis in Luca is a chat answer, not a report you configure. You ask which products are bought together, and Luca works through your orders and returns the pairs that turn up together. Each pair carries support, confidence and lift, so you can see how common it is and whether it sells together more than chance.

Behind the answer, Luca calculates [your KPIs](https://ask-luca.com/product-docs/use-cases/kpi-reporting) itself from one private warehouse that holds your connected data. Basket analysis is one of the standard calculations, alongside retention, profitability and acquisition efficiency. Every figure in an answer links to the table it came from, so you can check a pair before you act on it.

### What you see when you ask {#what-you-see}

An answer arrives with a title, the figures you asked for, a short set of bullets and a source line naming the data it was built from. A **Download the data** control sits under the answer. Ask a follow-up and Luca keeps the thread, so you can narrow a pair down without starting again.

![A Luca chat titled Last week's sales: the question How did we do last week, and Luca's answer: a one-line summary followed by tiles for net revenue, orders, average order and new customers for the week](https://ask-luca.com/product-docs/images/customers/pair-question-in-chat.webp)

*This chat answers a sales question, but a basket question comes back in the same layout: a short summary first, then the figures behind it. The figures are sample data for a demo brand.*

> **Try asking Luca:** "Which products were bought together most often in August and September last year? Show support, confidence and lift for each pair."

### What it is not {#what-it-is-not}

It is not a bundle builder or a storefront widget. Luca doesn't create bundles, change prices or add "frequently bought together" blocks to your product pages. Every connector is [read-only](https://ask-luca.com/product-docs/platform/read-only-access), so Luca reads your store and writes nothing back. You take the pairs it finds and set up the bundle in your own store and apps.

It also shows what sold together, not which ad or campaign brought each order in. A pair's numbers come from your order history.

## Why market basket analysis matters {#why-it-matters}

### What it's for {#what-its-for}

Market basket analysis is for the moment before you commit stock and discount to a bundle. It tells you which products your customers already put in the same order. It also tells you whether that happens more often than chance would explain.

### How it helps {#how-it-helps}

- **Bundles built on evidence.** You start from pairs customers already choose, not from guesses about what belongs together.
- **Discounts that earn their cost.** A pair with lift near 1 sells together by chance, so a bundle discount mostly gives margin away.
- **Stock committed to the right components.** You know which items the bundle will pull through before you order them.
- **Cross-sells with a reason.** Product page and email suggestions follow pairs with real pull.

### What happens without it {#without-it}

Bundles get built from the bestseller list. The bestseller turns up in half of all orders, so it looks like a perfect partner for everything. The bundle launches, sells to customers who would have bought both anyway, and the discount comes straight out of margin. Meanwhile the components that pair strongly sit unbundled, and stock gets ordered for a set nobody was asking for.

### Why buyers care {#why-buyers-care}

Heads of growth care most in the weeks before a seasonal push. That is when the bundle, its discount and the stock behind it are all decided at once. Founders care because a weak bundle costs margin twice: once in the discount and again in stock that doesn't move.

## The Luca AI approach to market basket analysis {#approach}

{/* 5P */}

Picture the head of growth at a Shopify stationery brand in early summer. Back-to-school is the biggest selling window of her year, and she wants a bundle live before it starts. Her shortlist has A5 notebooks, fineliner sets, gel pens and weekly planners, and the supplier wants a firm order soon.

The usual route is a spreadsheet. Someone exports last year's orders, rebuilds every basket and counts pairs by hand, which can take days for a few hundred products. The result is often a list of the most common pairs, and that list usually just repeats the bestsellers.

What has changed is that the measures that separate a strong pair from a coincidence are standard and simple. Support, confidence and lift come straight from order counts. The hard part was never the arithmetic. It was getting clean, matched order data into one place and asking the question quickly.

Luca does that work. It imports her full Shopify history, keeps it in one private warehouse and calculates basket analysis as a standard measure. She asks in chat which products sell together, then follows up on the pairs that matter. She sets up the bundle in her own store, because Luca only reads.

Every pair in the answer carries all three measures together, not just a count of orders. That is the column that sets it apart in How Luca AI compares for market basket analysis, below.

## How products bought together get found {#how-pairs-get-found}

There are a few ways to find products that sell together, and they catch different things. Knowing what each one misses tells you how far to trust a pair before you build a bundle on it.

### Counting pairs {#pair-counts}

The simplest method counts how many orders contain both products. It catches your most common combinations quickly, and the [Shopify guide to market basket analysis](https://www.shopify.com/blog/market-basket-analysis) treats a store report of products in the same basket as the place to start. It misses whether the pair is any stronger than chance. A bestseller pairs "often" with almost everything simply because it is in so many orders.

### Support, confidence and lift {#association-rules}

Association rules add two measures to the count. Confidence asks how often orders with one product also contain the other. Lift compares the pair's actual rate with the rate you would expect if the two products sold independently. This catches pairs with real pull and filters out bestseller noise. It can mislead on small counts, where a handful of orders produces a dramatic lift.

### Larger baskets {#larger-baskets}

Frequent itemset mining, such as the Apriori algorithm, extends the same measures to sets of three or more products. It catches kits, such as a notebook, pens and a pencil case bought as one. The cost is volume: the number of possible sets grows very fast, and most of them occur too rarely to act on.

### Comparing periods and segments {#comparing-periods}

The same analysis can be run for two seasons, two channels or two customer groups and compared. This catches pairs that only work at certain times. It misses nothing in principle, but each slice has fewer orders, so the small-count problem comes back.

### What this means for your bundle {#what-this-means}

For a bundle decision you want the second method: every pair scored by support, confidence and lift. That is what Luca is confirmed to do. Its basket analysis returns the products frequently bought together, with all three measures on each pair. Check the pairs against the season you are planning for, and treat any pair with very few orders as a test rather than a stock commitment.

## Ask Luca which products are bought together {#ask-the-question}

Start with a broad question, then narrow it. [Plain-English questions](https://ask-luca.com/product-docs/chat/cross-source-answers) work; there is no query language or report to set up, and there is no limit on the number of questions. If a question is ambiguous, Luca asks a clarifying question before it answers.

> **Try asking Luca:** "Which products are bought together most often? List the top pairs with support, confidence and lift."

### Read the answer {#read-the-answer}

Read each pair in this order:

1. **Support first.** It tells you how many orders the pair actually covers. A pair in very few orders can't carry a bundle, whatever its other numbers say.
2. **Lift next.** Above 1 means the products sell together more than chance. Around 1 means they are independent. Below 1 means customers tend to buy one or the other.
3. **Confidence last.** It tells you which product leads. If 30% of notebook orders include fineliners, the notebook is the anchor and fineliners are the add-on.

If a figure looks wrong, open the table it links to and check the orders behind it.

### Plan the back-to-school bundle {#plan-the-bundle}

Take the pairs with high lift and enough orders behind them, and ask about each one directly. Follow-ups stay in the same thread, so you can move from the full list to one pair without repeating yourself.

> **Try asking Luca:** "For the A5 notebook and the fineliner set, how many orders had both last August, and what were support, confidence and lift?"

Then decide the bundle in your own store. Luca gives you the evidence. The product set, the price and the quantity you commit to are your calls.

### Follow up on a pair {#follow-up}

Ask for a chart when you want to see a pair over time. Luca draws it from your data in the same chat, with a source line underneath.

![A Luca chat titled Revenue by channel: the request Show me revenue by channel for the last six months as a chart, a stacked bar chart of Shopify, Amazon and Wholesale revenue from April to September, and two summary bullets](https://ask-luca.com/product-docs/images/customers/pair-chart-from-question.webp)

*Asking for a chart works the same way for a pair: name the two products and the weeks you want to see. The figures are sample data for a demo brand.*

> **Try asking Luca:** "Chart the weekly orders that contained both the A5 notebook and the fineliner set, from July to September last year."

### Check pairs in season {#check-in-season}

Buying patterns shift with the seasons, and the [Shopify guide](https://www.shopify.com/blog/market-basket-analysis) warns that a pair that works in winter may not hold in July. For a back-to-school bundle, ask about last year's back-to-school weeks as well as the last few months. A pair that is strong in both periods is safer than one that is strong in only one.

## How pair strength is measured {#pair-strength}

Pair strength comes down to three numbers, each built from simple order counts. The [arules introduction by Hahsler and colleagues](https://cran.r-project.org/web/packages/arules/vignettes/arules.pdf) gives the standard definitions, and each one has a weak spot worth knowing.

### Support, confidence and lift {#measures}

**Support** is the share of all orders that contain the pair. Its weak spot is that popular products get high support with almost anything.

**Confidence** is the share of orders with the first product that also contain the second. Its weak spot is direction: notebook to fineliners and fineliners to notebook give different numbers. It also ignores how popular the second product is anyway.

**Lift** divides the pair's support by what you would expect if the two products were independent. A lift of 1 means independence, and greater lift values indicate stronger associations. Its weak spot is small numbers: a pair in eight orders can show a lift of 5 by luck.

![Diagram for one illustrative pair across 2,000 orders: 400 orders with the notebook, 300 with the fineliner set and 120 with both, giving support of 6%, confidence of 30% and lift of 2.0](https://ask-luca.com/product-docs/images/customers/pair-strength-support-confidence-lift.svg)

*Illustrative numbers, not Luca output: how three order counts become support, confidence and lift for one notebook and fineliner pair.*

### When confidence misleads {#confidence-misleads}

A bestseller makes confidence look better than it is. If gel pens are in half of all orders, then about half of notebook orders will contain gel pens by chance alone. Confidence of 50% looks strong, but lift is 1.0: the notebook adds no pull at all.

![Bar chart with illustrative figures comparing two pairs: the notebook and fineliner set with 30% confidence against a 15% baseline for lift 2.0, and the notebook and gel pen 3-pack with 50% confidence against a 50% baseline for lift 1.0](https://ask-luca.com/product-docs/images/customers/high-confidence-no-lift.svg)

*Illustrative numbers, not Luca output: the gel pen pair has higher confidence but only chance-level lift, so it would make the weaker bundle.*

### Where Luca shows them {#where-luca-shows-them}

Luca puts support, confidence and lift on every pair in the chat answer. Each figure links to the table it came from, so you can see the orders behind a pair. To work on the pairs yourself, download the answer's data as CSV or Excel. Set a floor on orders together before you trust a high lift; the right floor depends on your order volume.

## How to run a market basket analysis {#how-to-run-market-basket-analysis}

You can run a market basket analysis by hand from an order export. It is worth doing once, because it shows exactly what the three measures mean.

### The manual method {#manual-method}

1. **Export orders with their line items** for the period you care about, such as last year's back-to-school weeks.
2. **Clean the lines.** Remove gift cards, shipping protection, free samples and refunded lines, which create false pairs.
3. **Rebuild each basket** as one row per order, listing the products it contained.
4. **Count orders per product and per pair.** With 300 products there are 44,850 possible pairs, so this is where spreadsheets slow down.
5. **Calculate support, confidence and lift** for each pair with the formulas above.
6. **Set a minimum** number of orders together and drop pairs below it, which is the minimum support threshold that market basket guides describe.
7. **Rank the rest by lift**, then check margin and stock for the top pairs before you build anything.

### The same job in Luca {#same-job-in-luca}

1. Connect your Shopify store. Luca imports your full order history.
2. Ask which products are bought together, naming the period if it matters.
3. Read support, then lift, then confidence for each pair in the answer.
4. Follow up on the pairs you are considering, or ask for a chart.
5. Download the data as CSV or Excel if you want to add margin or stock columns yourself.
6. Build the bundle in your own store. Luca reads your data and changes nothing.

### A worked example: choosing a back-to-school bundle {#worked-example}

This is an illustrative calculation with sample figures, not a Luca output. A stationery brand had 2,000 orders in one back-to-school month. The A5 notebook was in 400 orders (20%), the fineliner set in 300 (15%) and the gel pen 3-pack in 1,000 (50%). The weekly planner was in 250 orders (12.5%) and the pencil case in 200 (10%).

| Pair | Orders with both | Support | Confidence from the first item | Orders expected by chance | Lift |
|---|---|---|---|---|---|
| A5 notebook + fineliner set | 120 | 120 ÷ 2,000 = 6% | 120 ÷ 400 = 30% | 20% × 15% × 2,000 = 60 | 6% ÷ 3% = 2.0 |
| A5 notebook + gel pen 3-pack | 200 | 200 ÷ 2,000 = 10% | 200 ÷ 400 = 50% | 20% × 50% × 2,000 = 200 | 10% ÷ 10% = 1.0 |
| Weekly planner + pencil case | 50 | 50 ÷ 2,000 = 2.5% | 50 ÷ 250 = 20% | 12.5% × 10% × 2,000 = 25 | 2.5% ÷ 1.25% = 2.0 |
| Weekly planner + fineliner set | 15 | 15 ÷ 2,000 = 0.75% | 15 ÷ 250 = 6% | 12.5% × 15% × 2,000 = 37.5 | 0.75% ÷ 1.875% = 0.4 |

The notebook and fineliner set is the bundle to back. It sells together twice as often as chance, and 120 orders give the figure weight. The notebook and gel pens have the highest support and confidence, but their lift of 1.0 means they pair by chance. A bundle discount there would go to about 200 orders that already include both, so it would give margin away for no extra pull.

The planner and pencil case also show a lift of 2.0, but on only 50 orders. Treat that as a small test, not a stock commitment. The planner and fineliners, at a lift of 0.4, rarely go together, so leave that pair out.

## Get basket answers from your store {#set-up}

Basket analysis is calculated as standard for every brand that connects the relevant source, so there is nothing to switch on. You need your store connected and a question to ask.

1. **Connect Shopify.** Click **Connect** and approve read-only access on Shopify's own consent screen. The steps are in [setting up your AI analyst](https://ask-luca.com/product-docs/getting-started/set-up-luca).
2. **Let the history import.** Luca imports your full Shopify history, so last year's season is there to compare with.
3. **Ask in chat.** Type the question in plain English and send it.

Data refreshes once a day, overnight. If a pair looks different from what you see in Shopify today, check again after the next refresh.

## How market basket analysis works with the rest of Luca {#works-with}

### Shopify {#with-shopify}

Shopify connects self-serve and read-only. Your orders are the raw material for every pair, and Luca reads them without changing anything. The [read-only access page](https://ask-luca.com/product-docs/platform/read-only-access) explains exactly what Luca can and can't touch.

### Chat, follow-ups and charts {#with-chat}

Basket answers live in chat, so everything chat does applies to them. You can ask follow-ups, ask for a chart and combine a pair with other data in one answer. The [conversational analytics guide](https://ask-luca.com/product-docs/chat/cross-source-answers) shows how one reply draws on several sources at once.

### Downloads {#with-downloads}

Under each answer, **Download the data** gives you the pairs as CSV or Excel. That suits a planning sheet where you add margin, stock and discount columns before a bundle meeting.

### Customer retention {#with-retention}

A strong pair tells you what sells together in one order. Retention tells you what brings customers back for a second. Luca also shows the products and SKUs that don't drive repeat purchases, covered in [customer retention analysis](https://ask-luca.com/product-docs/use-cases/customer-retention-analysis). Read the two together before you put a product at the centre of a bundle.

### Demand forecasting {#with-demand}

Once you have picked a bundle, the question becomes how many units of each component to hold. Where forecasting is switched on, Luca forecasts orders and units per SKU, as described in [ecommerce demand forecasting](https://ask-luca.com/product-docs/use-cases/inventory-demand-forecasting). You still choose the quantity and place the order yourself.

## Who uses market basket analysis, and how {#who-uses-it}

### Growth and marketing leads {#growth-leads}

They use it to choose bundles and cross-sells before a seasonal push. The question is usually which pairs have real pull and enough orders to carry a promotion. They then build the bundle and the emails in their own store and email tools.

### Founders and CEOs {#founders}

Founders use it to sanity-check a bundle before they approve the discount. A quick look at lift shows whether the bundle adds pull or simply discounts orders that would have happened anyway. Many also read the [daily view of sales, ads, profit and cash](https://ask-luca.com/product-docs/use-cases/see-the-whole-business) alongside it.

### Operations and inventory leads {#operations-leads}

Operations leads use the pairs to plan stock for bundle components. A strong pair means one product's demand will pull the other along. That changes how much of the add-on they hold going into the season, which they then plan with [demand forecasts per product](https://ask-luca.com/product-docs/use-cases/inventory-demand-forecasting).

## How Luca AI compares for market basket analysis {#compare}

Luca answers market basket analysis questions in chat, and every pair comes back with support, confidence and lift. The other tools in this table list pairs in a report or answer questions in natural language. None of their public documentation puts the three measures on each pair. Luca AI is the only platform in this table that does all three.

Market basket analysis features, from each platform's public documentation:

| Platform | Products bought together | Support, confidence and lift on each pair | Plain-English questions |
|---|---|---|---|
| **Luca AI** | Yes | Yes | Yes |
| **Lifetimely** | Yes ([Product Journey report](https://help.useamp.com/article/1088-how-to-see-which-products-are-frequently-bought-together)) | Not in docs | Partly ([through Microsoft Copilot Studio](https://help.useamp.com/article/1365-lifetimely-connect-lifetimely-to-microsoft-copilot-studio)) |
| **Shopify** | Yes ([Items bought together report](https://help.shopify.com/en/manual/reports-and-analytics/shopify-reports/report-types/default-reports/order-reports)) | Not in docs | Yes ([Sidekick chat](https://help.shopify.com/en/manual/ai-powered-tools/sidekick/memory)) |
| **Saras Analytics** | Partly ([Amazon market basket report table](https://help.sarasanalytics.com/en_US/change-log-2025/daton-november-2025)) | Not in docs | Yes ([Saras iQ in natural language](https://help.sarasanalytics.com/en_US/getting-started-saras-IQ/setting-up-your-iq-account)) |
| **Daasity** | Partly ([products typically purchased together](https://help.daasity.com/core-concepts/dashboards/report-library/ecommerce-performance/product)) | Not in docs | Partly ([from your terminal over MCP](https://help.daasity.com/core-concepts/ai-analyst/connect-your-ai-mcp/connect-gemini-cli)) |
| **Triple Whale** | Not in docs | Not in docs | Partly ([ask Moby, documented for LTV](https://triplewhale.readme.io/docs/lifetime-value-ltv)) |

*Yes: documented and built in. Partly: documented with a limit, shown in brackets. Not in docs: not found in the vendor's public documentation.*

### Why Luca AI is the stronger choice for market basket analysis {#compare-stronger}

- **Pair strength, not just pair counts.** Luca scores every pair by support, confidence and lift. Lifetimely's [Product Journey report](https://help.useamp.com/article/1088-how-to-see-which-products-are-frequently-bought-together) lists products commonly bought together, without those measures in its documentation. See [Lifetimely alternatives](https://ask-luca.com/blogs/lifetimely-alternatives).
- **A question instead of a dashboard.** Luca answers the basket question in chat. Daasity documents [which products are typically purchased together](https://help.daasity.com/core-concepts/dashboards/report-library/ecommerce-performance/product) on a product dashboard. See [Daasity alternatives](https://ask-luca.com/blogs/daasity-alternatives).
- **Basket analysis calculated as standard.** Luca calculates it for every brand that connects its store. Triple Whale documents [asking Moby](https://triplewhale.readme.io/docs/lifetime-value-ltv) for lifetime value, with no basket analysis in its documentation. See [Triple Whale alternatives](https://ask-luca.com/blogs/triple-whale-alternatives).

*Competitor details checked against each vendor's public documentation on 7 November 2026. Features and plans change, so confirm with the vendor before you decide.*

## Limits of market basket analysis in Luca {#limits}

- **Answers come in chat.** You ask for basket analysis as a question; there is no separate basket screen to browse.
- **Luca never changes your store.** It doesn't create bundles, set prices or add product recommendations to your storefront. You make those changes in your own store and apps.
- **It shows what sold together, not why.** A pair's figures come from order history. They don't say which ad, campaign or email brought each order in.
- **Data is a day behind.** Every source refreshes once a day, overnight, not in real time.
- **Answers only cover connected sources.** Luca answers from the data you have connected and says so when it doesn't have what a question needs.

## Common questions {#common-questions}

#### What is market basket analysis? {#faq-what-is-it}

Market basket analysis looks at which products customers buy in the same order. It scores each pair by support, confidence and lift, so you can tell a pair with real pull from one that sells together by chance. Brands use it to plan bundles and cross-sells.

#### What do support, confidence and lift mean? {#faq-measures}

Support is the share of all orders that contain the pair. Confidence is the share of orders with the first product that also contain the second. Lift compares the pair's actual rate with the rate expected if the two products sold independently.

#### What is a good lift value? {#faq-good-lift}

A lift above 1 means two products sell together more often than chance, and a lift of 1 means they are independent. There is no single cut-off. A lift of 2.0 on a few hundred orders is more useful than a lift of 5 on eight orders.

#### How many orders does a pair need before I trust it? {#faq-how-many-orders}

Set your own floor on orders together, based on your volume, and ignore pairs below it. Small counts swing wildly, so a pair in a handful of orders can show a high lift by luck. Treat a strong pair with few orders as a test.

#### Why does my bestseller pair well with everything? {#faq-bestseller}

A bestseller is in so many orders that it lands in the same basket as most products by chance. That gives it high support and confidence with everything. Lift corrects for this: a bestseller pair with lift near 1 adds no real pull.

#### Is there a market basket screen in Luca? {#faq-screen}

No. You get market basket analysis by asking in chat. Ask which products are bought together, and the pairs come back in the answer with support, confidence and lift. You can then follow up, ask for a chart or download the data.

#### Can Luca build the bundle in my store? {#faq-build-bundle}

No. Every connector is read-only, so Luca reads your store and writes nothing back. When you ask it to change something in a connected system, it can only recommend. You create the bundle, its price and its product page in your own store.

#### Can I download the pairs? {#faq-download}

Yes. Under the answer, **Download the data** gives you the pairs as CSV or Excel. That is useful when you want to add margin, stock or discount columns in your own planning sheet.

#### How up to date are the pairs? {#faq-freshness}

Luca refreshes every connected source once a day, overnight. A pair reflects orders up to the last refresh, not orders placed in the last hour. For bundle planning, which looks back over weeks or a season, that rarely matters.

#### Does basket analysis cost extra or need setting up? {#faq-set-up}

Basket analysis is calculated as standard for every brand that connects the relevant source, so there is nothing to switch on. There is also no limit on the number of questions you can ask on any plan.

#### Can I compare last year's back-to-school weeks with this year? {#faq-compare-periods}

Name the period in your question, then ask a follow-up for the other period in the same thread. Luca keeps the thread of the conversation. Comparing the two shows whether a pair is strong every season or only some of the time.

#### What if my question is unclear? {#faq-unclear-question}

When a question is ambiguous, Luca asks a clarifying question before it answers. If you ask about "pens", for example, it may check which pen products you mean. Answer it, and the analysis carries on in the same thread.

#### Can I check the orders behind a pair? {#faq-check-orders}

Yes. Each figure in an answer links to the table it came from, so you can see the data behind a pair before you act on it. That is the quickest way to spot a false pair caused by a free sample or a gift card.

## Further reading {#further-reading}

- [Introduction to arules, by Hahsler, Grün, Hornik and Buchta](https://cran.r-project.org/web/packages/arules/vignettes/arules.pdf): the standard definitions of support, confidence and lift, with a small worked example.
- [Market basket analysis: how it works and how to use it, from Shopify](https://www.shopify.com/blog/market-basket-analysis): how to read the three measures and how seasons change buying patterns.
- [The best AI agents for ecommerce](https://ask-luca.com/blogs/ecommerce-ai-agents): where an AI analyst fits among the tools a brand runs.

## Related {#related}

- [Luca AI product documentation](https://ask-luca.com/product-docs): every guide in one place.
- [AI agents for ecommerce](https://ask-luca.com/product-docs/use-cases/ai-agents-for-ecommerce): what Luca does and what you decide.
- [Ecommerce KPI reporting](https://ask-luca.com/product-docs/use-cases/kpi-reporting): one set of KPI definitions in dashboards, chat and exports.
- [ROAS analysis for ecommerce](https://ask-luca.com/product-docs/use-cases/roas-analysis): ad returns by channel against break-even ROAS.
- [Shopify revenue forecast](https://ask-luca.com/product-docs/forecasting/revenue-forecast): daily revenue up to a year ahead, checked every night.
- [Customer retention analysis](https://ask-luca.com/product-docs/use-cases/customer-retention-analysis): which cohorts and products keep customers.
- [Setting up Luca](https://ask-luca.com/product-docs/getting-started/set-up-luca): connect sources, invite your team and start asking.

*Updated 7 November 2026: first published with the ways pairs get found, how pair strength is measured, the manual method with a worked example, and further reading.*