---
title: "AI for beauty and cosmetics stores: how to recommend the step missing from the routine"
date: "2026-10-24"
author: "Rafael Gijon"
canonical: "https://crossup.ai/en/blog/ia-para-tiendas-de-belleza-y-cosmetica"
locale: "en"
---

# AI for beauty and cosmetics stores: how to recommend the step missing from the routine

Almost nobody shops a beauty store for a single product. They shop for a routine, even if it isn't complete yet. Whoever picks a repairing shampoo will need the conditioner from the same line; whoever adds a night serum is one step away from the day one. The sale you lose is the missing step. That is the job of a recommendation AI in cosmetics: read the cart, spot what's missing and offer it when it makes sense.

This guide covers which product relationships sell in beauty, what to recommend at each moment of the purchase, the mistakes that ruin the experience and what to measure, with real data from two stores that use CrossUp: Beauty Web and Tomassa.

## Which relationships sell in a beauty store

Not every recommendation is equal. In cosmetics, the ones that work follow a logic of use the shopper understands in two seconds:

- **The next step of the routine.** Cleanse, treat, moisturize, protect. Shampoo, conditioner, mask. This is the complement par excellence: it doesn't compete with what they chose, it completes it.
- **The same line.** If someone picked a product from a line, the rest of that line is the most natural recommendation. Same actives, same scent, same promise, and the shopper already trusts that brand.
- **The right variant.** Another size, another shade, the version for a different skin or hair type. Here the recommendation doesn't add a product: it helps them choose well.
- **The refill.** When the product runs out, the refill is the obvious purchase. Recommended at the right time, it saves a search.
- **The kit.** Several steps of the routine packed into a single decision. The routine, ready to go.

A good recommendation strategy in beauty mixes these five depending on the moment. "Similar products" on their own tend to show the shopper what they just passed on.

## What two real stores show

The figures in this section cover the **last 90 days to Oct 24, 2026**. They are observational: they show what happened with the recommendations, with no control group, so they don't say how much would have sold without them.

### Beauty Web: the day and night routine

At Beauty Web, **11.5%** of orders included at least one recommended product (n = 23,515 orders). In those orders, the **lift in the ticket of orders with a recommendation (the recommended item over the rest of that order)** was **21%** (n = 2,697 orders with a recommendation and something else in the cart).

Some of its best combinations at add to cart, with the catalog's own names (among the store's best, not the average):

- **Tratamiento Nocturno Regenerador con Melatonina → Tratamiento Diurno concentrado C10 con Vitamina C** (night treatment → day treatment). Shown 67 times: 26.9% added it and 11.9% bought it.
- **Pads Reutilizables para Contorno de Ojos → Tratamiento Diurno C10** (reusable eye pads → the day treatment). Shown 152 times: 15.1% added it, 6.6% bought it.
- **Recarga Crema Contorno de Ojos Energía → Pads Descongestivos para Contorno de Ojos** (eye cream refill → depuffing eye pads). Shown 148 times: 12.2% added it, 5.4% bought it.

The pattern is clear: night calls for day, and an eye treatment calls for its accessory. It's the missing step, not "another similar product". The full story is in the [Beauty Web case](/en/cases/beauty-web).

### Tomassa: same line, next step

At Tomassa, **11.8%** of orders included a recommended product (n = 8,987 orders), and the lift in the ticket of orders with a recommendation (the recommended item over the rest of that order) was **37%** (n = 1,013).

Among its best combinations at add to cart:

- **Fidelite Shampoo Keratina → Fidelite Acondicionador Keratina** (shampoo → conditioner). Shown 60 times: 20.0% added it and 11.7% bought it.
- **L'Oréal Professionnel Acondicionador Absolut Repair → Absolut Repair Shampoo Refill** (conditioner → shampoo refill). Shown 68 times: 25.0% added it, 7.4% bought it.
- **Karseell Shampoo Maca Power → Karseell Máscara Reparadora de Colágeno** (shampoo → mask). Shown 94 times: 14.9% added it, 6.4% bought it.

All three stay within the same line and follow the order of a hair routine: shampoo, conditioner, mask. The second one adds a refill format. Tomassa also runs several moments at once (product page, add to cart, cart, checkout and bundles), and you can see it in detail in [the Tomassa case](/en/cases/tomassa).

## What to recommend at each moment of the purchase

The same relationship can sell or annoy depending on when it shows up. This is how it plays out in a beauty store:

### Product page

The shopper is still deciding. This is the place for **the full line** and **the variants**: the same product in another size, the right shade, the version for their skin type. The routine kit also works here, as an alternative to buying a single step.

### Add to cart

The moment of highest intent: they've chosen. This is where **the next step of the routine** pays off. Every Beauty Web and Tomassa combination above happened here. One recommendation, two or three options at most, added without leaving the page.

### In the cart

While reviewing the order, shoppers are open to completing it. **Refills**, travel sizes and accessories work well: quick decisions that don't compete with the main item.

### At checkout

They're paying: only something **small and obvious** belongs here. A single-dose sachet, an accessory, the makeup remover that goes with everything. Nothing that makes them think again.

### Thank-you page

The purchase is over; the relationship isn't. **Set up the next order**: the refill of what they just bought, or the step they still don't have.

## Bundles: the routine, ready to go

Why choose skincare bundles over individual products? Because a kit that solves a whole routine, at a price that feels like a win, turns three decisions into one. It isn't "three products on sale": it's "your night routine, ready". For the store, the bundle lifts the ticket; for the shopper, it simplifies the choice.

The rule is the same as for any bundle: relevance, incentive and timing. We covered it in [bundles that work](/en/blog/combos-que-funcionan). In beauty, relevance comes from the order of the routine and from the line: if a kit mixes steps that don't belong together, the shopper won't get it and won't buy it.

## Common mistakes in beauty stores

- **Recommending the same step twice.** A second shampoo when there's already one in the cart completes nothing. The recommendation has to move the routine forward.
- **Ignoring shade or skin type.** In makeup and treatments, the wrong variant is worse than no recommendation. Help them choose the right variant instead.
- **Mixing lines without a reason.** The complement from the same line almost always wins, because the shopper already trusts that brand.
- **Showing out-of-stock products.** Shades and sizes sell out fast; recommending something unavailable frustrates.
- **Too many options at once.** Ten suggested products is a catalog, not a recommendation.
- **Measuring only clicks.** A click isn't a sale. What matters is how much of what you recommend ends up in a paid order.

## What to measure, in percentages

To know whether recommendations are working, track these four figures, always with their period and sample size:

1. **% of orders with a recommended product.** How many of your orders include at least one product that came from a recommendation. At Beauty Web it was 11.5% and at Tomassa 11.8%, in the last 90 days to Oct 24, 2026.
2. **Lift in the ticket of orders with a recommendation (the recommended item over the rest of that order).** How much more that order is worth because of what was added. It's not the same as comparing the ticket of orders with and without recommendations, which mixes causes.
3. **The funnel: shown → added → bought.** At Beauty Web, out of 157,100 recommendations shown at add to cart and at checkout, 4.9% were added and 1.0% were bought. At Tomassa, out of 40,635, 7.8% were added and 1.8% were bought.
4. **% of added items that get bought.** Of what shoppers added to the cart, how much ended up in a paid order: 19.8% at Beauty Web and 22.9% at Tomassa. If this number is low, check price, stock or the moment the recommendation appears.

Other stores' figures are a reference, not a target: every catalog is different.

## How to get started

1. **Organize your catalog by routine and by line.** Clear names and categories help tell a step from a variant and a refill.
2. **Turn on "add to cart" first.** It's the moment of highest intent and where the best combinations of both stores showed up.
3. **Add cart, checkout and thank-you page** with different relationships: refills, small accessories and the next purchase.
4. **Build a few bundles, and make them routines.** Three kits that make sense beat ten that confuse.
5. **Review your best combinations every few weeks** and measure in percentages, with the sample size in view.

CrossUp does this with SalesPilot, its AI engine: it decides with your catalog, your stock, your orders, the product being viewed and what's already in the cart, recommends across the 5 purchase moments and attributes every recommended product to the order that bought it. You can try it free for 15 days. See how it works for your industry in [AI for beauty stores](/en/solutions-beauty), or read the [Beauty Web](/en/cases/beauty-web) and [Tomassa](/en/cases/tomassa) cases.

---

[Home](https://crossup.ai/en) · HTML: https://crossup.ai/en/blog/ia-para-tiendas-de-belleza-y-cosmetica