For DTC beauty brands and multi-brand retailers
MatchLayer builds your catalog into a product graph that knows which items genuinely belong together across brands and categories — so the module works on day one, without waiting for purchase history. No rules to build, no merchandising to maintain: we build it, launch it and operate it.
+13.2% AOV — blended lift on the PDP module at Minimalist. Live in 7 working days.
Share your store URL. No feed or integration required for the sample.
Rice extract and niacinamide, no alcohol denat — protection that won't sting the skin you're calming.
The toner is humectants with no lipid. Ceramides are what hold that water in overnight.
Beauty teams we've worked with
PDP module outcomes
AOV and attach rate are the primary commercial outcomes for the recommendation module. Product graph outcomes are shown separately below.
Blended lift in average order value
Order-level attach rate, with a 12.3% lift in revenue per eligible session measured against a holdout group
Soft Services was measured control-versus-treatment against a holdout group over 90 days, on a store averaging 45,000 monthly sessions — not pre-post. Full test methodology, including the eligible-session definition and holdout split, is shared during evaluation.
The module
One consistent interaction pattern, adapted to each category’s decision logic and styled to your storefront.
Two high-confidence picks by design — not another endless recommendation carousel.
Invisible and non-pilling over a treatment cream — daily UV is what undoes the brightening you're paying for.
The same peony complex, worked in overnight while the cream's barrier support holds.
A warm rose-brown two steps deeper — same undertone, so the nude keeps its edge.
This satin grips dry patches. A mild AHA clears them before morning.
Every pick carries a shopper-facing reason. MatchLayer writes and maintains it — your merchandisers don't.
How every recommendation is chosen
The module does not wait for purchase history before it can work. Product intelligence creates a safe eligible set immediately; store-level signals sharpen the commercial order over time.
Formula, ingredients, routine step, category, finish, shade and use-case logic determine which products genuinely belong with the anchor.
Where sufficient signals exist, product-level buying patterns help rank the strongest compatible options. Behaviour never overrides compatibility.
Brand adjacency, duplicate products and conflicting actives are resolved when relationships are built. Inventory and commercially blocked SKUs are checked at serve time against webhook-synced stock, so a sold-out product never renders.
Every recommendation receives a concise explanation of why it belongs, so the module helps the shopper decide rather than showing another thumbnail grid.
Control and measurement
MatchLayer manages catalog enrichment, relationship logic, exclusions, new-product ingestion and ongoing QA. You also get a dashboard covering the whole catalog—search any SKU, read its ranked pairs and the reasoning behind them, flag or remove anything. Nothing obliges you to open it: the module runs whether you review one product or none.
Relationships are generated across the entire catalog before launch, so there is nothing to wait for and nothing to build. Review is available, never required.
We monitor the module against eligible PDP traffic, surface what changed and review commercial performance with your team.
The product graph
One real SKU, as it sits in the product graph. Automated resolution and consistency checks structure the catalog; flagged or low-confidence fields are routed for expert review before they can serve.
A soft pale peachy-beige nude with subtle warm undertones, reminiscent of bare lips with a whisper of peach.
Suitability is described by likely visual effect, so the module can explain when a liner or prep step improves the result.
The same product graph can later support search, filtering, PLPs and lifecycle journeys—without rebuilding the product logic.
Product graph outcome
This implementation used the MatchLayer product graph to improve product discovery and segmented lifecycle communication. It is shown separately because it was not a PDP-module result.
Where this came from
MatchLayer began inside Kult, a multi-brand beauty retailer carrying 200+ brands, where many SKUs arrived as little more than a brand name, product name and shade.
The product graph was built to structure that catalog, map compatible next steps and support clearer product decisions across a database of 60,000 beauty SKUs. The taxonomy was developed with dermatologists and cosmetic chemists; 11,000 colour-cosmetics SKUs ran live in the module.
These figures are internal reporting from that launch period, not a controlled A/B test. The client results at the top of this page are.
Connect Shopify through the MatchLayer custom app or send the catalog as it already exists. We build ingestion around your structure, enrich and map the products, then enable one app-embed block on your PDP template, with QA and a rollback plan.
Connect Shopify or send a CSV or product feed. No replatforming or catalog migration.
We map your existing catalog structure onto the beauty taxonomy and resolve variants.
Attributes are generated and cross-checked; flagged or low-confidence fields are reviewed.
Compatible products are mapped, with inventory, conflicts and brand-adjacency rules applied.
The dashboard opens with the full catalog already mapped, so you can spot-check any SKU before launch. Approve the placement, reasoning style and any standing restrictions. Reviewing individual pairings is optional.
The MatchLayer custom app is installed and one app-embed block is enabled on your PDP template after responsive QA, approval and rollback checks. The block can be toggled off from theme settings without editing theme files.
After launch, MatchLayer handles the operating work. Your team joins periodic performance reviews and steps in only when a commercial exception requires approval.
What beauty operators say
Our assortment is complex, so recommending the right complementary products is hard. MatchLayer goes beyond generic “You may also like” — it identifies compatible products, explains why they work together, and helps customers buy with more confidence.
Mohit Yadav
CEO, Minimalist
We wanted intelligent recommendations to increase conversion. MatchLayer gave us a scalable way to surface compatible products directly on the PDP — strengthening the customer journey without adding manual merchandising work for our team.
Stephanie DiSturco
Chief Revenue Officer, Soft Services
The enrichment work created a more structured and intelligent catalog — improving search, filters, merchandising, recommendations, and the quality of information on every product page.
Shouvik Roy
Chief Technology & Product Officer, Mosaic Wellness
Why MatchLayer
MatchLayer combines beauty-native compatibility with store-level commercial signals, without handing your merchandising team another rule system to maintain.
Data and storefront safety
Each client receives an isolated product graph. Product attributes, relationships and store signals are not used to improve a competitor’s recommendations.
Attributes, relationships and shopper-facing reasoning are generated and reviewed ahead of time, then stored. No language model runs on the PDP. At request time the module does one thing: filter that stored set against inventory and your commercial rules. Stock reaches the graph by webhook, so that filter is a local lookup: no external API call happens in the render path.
Conflicting actives, redundant steps, duplicate variants, out-of-stock products and commercially blocked SKUs can be removed before they reach the shopper.
The behavioural layer can work from SKU-level and aggregated store patterns. Individual shopper quizzes or customer profiles are not required for the module to operate.
Pricing
MatchLayer includes the PDP recommendation module, shopper-facing reasoning, ongoing catalog ingestion, managed QA and performance reporting. Every plan is fully managed—no rule-building or ongoing recommendation merchandising required.
We are capping onboarding through September 30, 2026 so each catalog gets the same enrichment review. Until then, catalog enrichment — normally a one-time $900 up to 500 active SKUs, or $2,500 for 501–2,000 — is included. Attribute extraction, taxonomy mapping, variant resolution and expert review of flagged fields.
For complex DTC catalogs
90-day initial rollout — the window needed to measure lift against control. Month to month after, cancel with 30 days' notice.
For larger DTC and multi-brand retailers
90-day initial rollout — the window needed to measure lift against control. Month to month after, cancel with 30 days' notice.
For enterprise catalog complexity
Term, scope and SLA agreed during evaluation. No per-click, per-order or revenue-share fees at any tier.
Flat pricing replaces two ongoing costs: a revenue-share recommendation engine, which scales with your GMV rather than with the work involved, and the merchandiser time it takes to build and maintain pairing rules by hand. Catalogs above the listed limits are scoped separately.
Questions
It is designed for DTC beauty brands and multi-brand retailers where shoppers must choose between multiple products, categories, shades, ingredients, routines or use cases. Very small catalogs with obvious product relationships may not need this depth.
For the free five-PDP sample, a store URL is enough. For implementation, connect Shopify through the MatchLayer custom app or provide a CSV or product feed, then approve one app-embed block on your PDP template.
No. MatchLayer is a fully managed recommendation layer. We operate catalog ingestion, enrichment, relationship mapping, exclusions, new-product coverage, QA and ongoing tuning. Your team approves the initial setup, retains override authority and reviews performance with us.
No. The module works from catalog intelligence on day one. Store-level signals, where available, only refine the order — they never make an incompatible product eligible.
We define eligible PDP sessions and can configure control-versus-treatment measurement. The primary commercial metrics are recommendation attach rate, AOV and revenue per eligible session, supported by coverage and empty-slot reporting.
Yes, two ways. In the dashboard you can search any SKU, read its ranked pairs and the reasoning behind each one, and flag or remove a pairing directly. Or share standing restrictions with us and we implement and maintain them. Reviewing is optional — most teams spot-check a category before launch and then leave it alone — but the whole catalog is visible whenever you want to look.
Out-of-stock or commercially blocked products can be suppressed before serving. New products enter through the ongoing catalog feed, receive compatibility coverage without waiting for order history, and are included after the configured QA or approval step.
Your catalog is built into a product graph hosted on our side, holding the ranked recommendations and their reasoning for every anchor product. None of that is generated while a shopper waits — no language model runs on the PDP. At request time the module reads the ranked set for that product, drops anything out of stock or commercially blocked, and returns the top picks that remain. Inventory changes are pushed to the graph by webhook, so that check is a local lookup — not a synchronous call to your store while the page renders. Nothing in the render path depends on an external API being up or within its rate limit. Because the stored ranking runs deeper than the two slots shown, a sold-out product is usually replaced rather than leaving an empty slot.
The module can show one recommendation or render no slot at all. The stored ranking runs deeper than the slots shown, so a stock-out normally promotes the next compatible product rather than emptying the module. When nothing compatible remains, we would rather decline than pad the slot with a duplicate, an incompatible product or the store bestseller.
No. The module works for anonymous first-time traffic. It uses the product being viewed, catalog intelligence and store-level patterns rather than requiring an individual profile.
Free catalog-module sample
Share your store URL. We’ll select five anchor products and show the compatible cross-brand or cross-category recommendations, the shopper-facing reason for each, and one relationship we would deliberately exclude.
We'll review your catalog and send your 5-PDP sample within about three working days. If we need anything else to build it, we'll email you first.