Advanced AI Enrichments: what they are and what they let you analyze

Standard monitoring tells you that your brand was covered. Advanced AI Enrichments tell you what that coverage was actually about: which product, which collection, which Voice, which event, and whether it was corporate news or product editorial.

Same coverage, a completely different conversation. Every brand mention is read by an AI model and tagged against one shared taxonomy, applied identically across Online, Print and Social, in every language. The result is structured, standardized data you can filter, report on and benchmark against competitors, instead of a keyword list you have to maintain.

🎯 What you can answer with Advanced Enrichments

🧩 Three layers of tagging

Your coverage carries three distinct layers of tagging. They behave differently and are set up differently, so it helps to know which is which.

1. Standard enrichments (always on, included)

Foundational classification applied to every placement:

• Brand
• Industry (barebones)

2. Advanced enrichments (the unified taxonomy)

Brand-level AI enrichments, standardized and cross-channel. This is what this article covers:

• Product Category & Sub-category
• Collection & Product Name
• Corporate (Yes/No) and Corporate Categories
• Gender and Universe
• Events and Mentioned Voices

3. Custom enrichments: Smart Tags

The layer specific to your organization, for anything the universal taxonomy doesn't cover: your own campaign calendar, your spokespeople roster, your key messages, your brand themes.

⚙️ How it works

1. Collection. Launchmetrics collects raw content across Print, Online and Social.
2. Extraction. An AI model reads each document and extracts the relevant information in a semi-structured form.
3. Matching. The extracted information is matched against the shared taxonomy, built and maintained by the Launchmetrics Fashion, Lifestyle & Beauty experts.
4. Human cross-check. Launchmetrics teams cross-check a sample of results per brand to hold the quality bar and detect where the models need tuning.

The output is richer, standardized data that enables deeper insight and genuine benchmarking.

🔎 One mention, many data points

Here is what a single article mentioning a handbag can carry:

A single mention can carry one or many taxonomy entities. All entities attach directly to the brand, and not every relationship between entities is mandatory. A mention can be corporate only, corporate and industry, or industry with Voices and events, and so on.

📈 Four questions, answered end to end

Each of these takes a handful of filters. None of them is answerable with standard monitoring alone.

Which products actually drove the quarter?

• How you filter: Category = Bags, then Product Name, your brand plus your closest competitor, over the quarter.
• What you learn: how concentrated your coverage is, whether one hero product carried the period while the rest of the line stayed invisible, and whether your competitor's coverage is spread across more products than yours.
• What you do with it: reallocate seeding and sample priorities toward the products that convert coverage, and re-brief press on the ones that don't.

Did that event pay off, and who carried it?

• How you filter: Event = Watches & Wonders or Paris Fashion Week, plus Mentioned Voices, Corporate = No, broken down by market.
• What you learn: the MIV the event generated for your brand specifically, which Voices amplified it, and which markets picked it up. Run the same filters on the previous edition for a like-for-like comparison.
• What you do with it: build the next edition's Voice and market plan on what actually worked.

How much of our coverage is institutional rather than product?

• How you filter: Corporate = Yes and No over time, then Corporate Categories to see the type: Financial news, Social responsibility, Governance news, Brand partnership.
• What you learn: whether a results cycle or a governance story is drowning out product storytelling, and whether your social responsibility messaging is being picked up rather than just published.
• What you do with it: rebalance the PR calendar, and report corporate and brand performance separately instead of as one aggregate number.

Are we visible on the segments we're investing in?

• How you filter: Gender = Men with Category, by market and period. Same read with Universe for Sportswear, Haute Couture or Wedding.
• What you learn: the gap between where your portfolio is invested and where your coverage actually sits, and which competitors own the segment you are trying to enter.
• What you do with it: take a coverage gap to a market team with a number attached to it.

🔀 The combinations worth running

A single dimension tells you what happened. Crossing two or three is where the analysis is.

• Product Name × Event. Which products genuinely carry an event, and which are simply present in the room.
• Category × Gender × market. Where your portfolio is visible market by market, against where you are actually investing.
• Mentioned Voices × Category. Which Voices are associated with which part of your range. This is also what powers Voice Echo.
• Collection × time. How long a collection keeps generating coverage after launch, and which ones fade within a week.
• Corporate = No × Universe. Product storytelling read by context: Ready-to-Wear against Haute Couture, Sportswear or Wedding.

✨ What makes these enrichments different

• Cross-channel. The same taxonomy across Online, Print and Social, so a category means the same thing wherever the coverage came from.
• Multilingual. Applied consistently across every language you are covered in, including non-Latin alphabets. Nothing to translate or maintain per market.
• Granular, at brand level. Fine-grained enough for real product and collection-level analysis, and attached to the brand rather than to the document.
• Benchmarkable. Standardized values, so your data is comparable to your competitors' and stays relevant as your portfolio evolves.
• At scale. Automated AI enrichment across Online and Social, with no manual tagging. On Print, the taxonomy is applied the way it always has been, by the expert team.
• Quality-first. Every enrichment has to clear a defined accuracy threshold before it is released, and human cross-checks continue afterwards.

📚 The enrichments, one by one

Product Category & Sub-category

Classifies each brand mention into one or more valid product categories and, where applicable, sub-categories.

• Values: 49 categories (Clothing, Skincare, Necklaces and so on) and 121 sub-categories (Tops, Face Care, and so on).
• How it's applied: a fixed two-level taxonomy. Levels are not all mandatory, so tagging stops at the highest level that can be detected with confidence. A lower level always implies the level above it.
• Good to know: a small number of Print values aren't recorded the same way on Online and Social, so they aren't yet fully cross-channel mapped and selecting them returns no documents. Affected values such as Tops, Bottoms and Dresses are flagged with an icon in the filters. Closing this gap is on the roadmap.

Corporate (Yes / No)

Flags whether a mention is primarily about corporate news, whether financial, institutional or strategic, so you can separate institutional noise from product editorial.

• Values: Yes, No.
• How it's applied: a simple binary classification on the mention, applied to all documents, so every mention should carry a value.
• Good to know: a mention can be both Corporate = Yes and carry an Industry.

Corporate Categories

For corporate coverage, classifies the type of corporate topic.

• Values (13): Financial news, Social responsibility, Governance news, Brand narrative, Brand partnership, Celebrity endorsement, Ambassador, Event, Sponsorship, Store news, Store opening, Mention, Other news.
• How it's applied: only when Corporate = Yes.
• Good to know: a mention can carry both a Corporate Category and an Industry.

Collection & Product Name

Detects specific collections and product names mentioned in association with a brand, which powers collection-level and product-level coverage tracking.

• Values: defined per brand. Collections such as LOVE by Cartier or Van Cleef & Arpels Alhambra, products such as Love Bracelet or Alhambra Pendant.
• How it's applied: Product Name is the editorial name as it appears in content, not an internal reference, and excluding attributes. A collection can span several sub-categories.
• Released brand by brand: each brand's list is human-validated and has to clear the accuracy bar before that brand's enrichment goes live.
• Self-evolving taxonomy: the list starts from a product base enriched by the Launchmetrics team, then keeps growing on its own. Any new product or collection the model extracts but can't match is flagged as a candidate for review, then added. Always complete, always current.

Gender

Classifies the gender target of the brand mention, for dimension-based filtering and reporting.

• Values (5): Men, Women, Unisex, Kids, Unknown.
• How it's applied: based on contextual signals, meaning product information and how the item is described.

Universe

Classifies which world the mention lives in, meaning a lifestyle, occasion or wardrobe context, rather than what the product physically is.

• Values (10): Ready-to-Wear, Sportswear, Wedding, Maternity, Haute Couture, Sleepwear & Loungewear, Costume Jewelry, Fine Jewelry, High Jewelry, Unknown.
• How it's applied: flagged independently of the product category.

Events

Detects events mentioned in association with a specific brand, which powers event-linked coverage tracking.

• Values: the Launchmetrics global events list, roughly 94 standard events such as Watches & Wonders or Paris Fashion Week.
• How it's applied: only when the content explicitly links the brand to the event. This is distinct from the existing Events feature.
• Scope: custom or single-brand events, such as a specific store opening or a one-off campaign, are out of scope and belong in Smart Tags.

Mentioned Voices

Detects Voices, meaning celebrities and influencers, mentioned in association with a specific brand. Powers brand-level Voice reporting and the Voice Echo feature.

• Values: celebrities and influencers from the official monitored list.
• How it's applied: only when content explicitly links the brand to a Voice on the official list.

🏷️ Smart Tags: the custom layer

Smart Tags are the enrichment layer specific to your organization. They classify Discover placements with an AI model and write into the existing Topics & Tags system, so results appear in feeds, filters, dashboards and reports exactly like any other tag.

• Values: anything. Campaign tracking, brand themes, key messages, spokespeople rosters, one-off events, custom sentiment rules.
• How a tag is built: you describe the need in your own words and your guidelines are used verbatim. A classification rule is drafted and automatically tested twice against 15 real placements, then a Launchmetrics reviewer validates it and can override any classification before it is saved.
• Time to results: a nightly batch run means results are live in your feed within 24 to 36 hours.
• Quality gate: every classification is tested against real placements before going live and reviewed at a human checkpoint. Prolonging an existing tag preserves the prior human review.

Three cases that show why keyword automation isn't enough:

• A campaign line such as "Designer for the Planet" never matches a sustainability keyword, so it is missed entirely.
• Themes such as Effortless, Surprise or Delight don't exist in keyword form.
• A brand's own key-message framework is a messaging structure, not a keyword pattern.

🧭 Advanced enrichment or Smart Tag?

An advanced enrichment covers it when the need is already in the universal taxonomy: Industry, Category & Sub-category, Collection & Product Name, Corporate, Corporate Categories, Gender, Universe, Events and Mentioned Voices. This route is standardized, benchmarkable and maintained for you, so it is always the preferred one.

A Smart Tag is the right answer when:

• It is specific to your organization: your campaign, your calendar, your spokespeople roster, your own themes.
• It is adjacent to the taxonomy but not covered yet, or is a variant that doesn't quite match the standard values.
• It is a one-off event that isn't on the standard events list, such as a specific store opening or a campaign.
• It is a product type not yet covered.
• It is sentiment.
• It is affiliate content detection.
• It is a theme identified in your own coverage that you want to compare against competitors or over time.

✅ How quality is held

Every enrichment has a production accuracy threshold it has to clear before release. Below the threshold, the enrichment is not released. Thresholds are set per enrichment, higher for simple, high-volume classifications and slightly lower for fine-grained multi-class ones.

Passing the bar isn't the end of it. Human cross-checks per brand continue after release, both to hold quality over time and to keep improving the models.

Two limits worth stating plainly:

• No image recognition. The models work from text, so visual-only signals aren't captured.
• Partial cross-channel gaps on Category. A small number of Print values aren't yet mapped to Online and Social. They are clearly flagged in the filters, and closing the gap is on the roadmap.

📅 Availability and setup

• Smart Tags: live since 22 July 2026.
• Advanced enrichments: live since 16 September 2026 on new coverage. Re-enriching your existing 2026 coverage is possible on request and is not applied by default, so ask your Customer Success Manager to scope it if you want a full year of history enriched.
• Setup: Launchmetrics configures the scope for you. Talk to your Customer Success Manager to define which enrichments and which brands you want activated.

➡️ How to get it activated

Launchmetrics sets the scope up for you. To make that conversation quick, come to your Customer Success Manager with three things:

• Which brands. Your own brands, and the competitors you benchmark against.
• Which enrichments. Start from the questions you want to answer rather than from the list of dimensions.
• Whether you want your 2026 history re-enriched, so you can analyze a full year straight away.