Lola Casademunt’s “total look” involves work that a photograph alone does not complete: linking every piece worn by the model to the store’s catalogue. Múltiplo describes a system based on GPT-4o and GPT-4o-mini to classify images and establish those relationships. The resulting descriptions also help produce product-page text and translations into the ecommerce store’s languages.
The consultancy includes the case in a presentation of its projects published on Oct. 19, 2026. The separate project account explains how the connection between image, recommendation and content is built. The economic interest lies in turning a fashion composition into specific routes to other items, while maintaining coherent product descriptions.
Category demand and brand searches follow different rhythms
Spanish searches show why a category’s attraction should be distinguished from interest in a particular label. In a comparison of the exact expressions “Lola Casademunt,” “vestidos de fiesta,” “cazadora” and “moda mujer,” the index for “cazadora,” meaning jacket, rises from 49 in the week of Oct. 4-10, 2026, to 68 during Oct. 11-17. “Lola Casademunt” remains at 9 in both weeks; “moda mujer,” at 16. “Vestidos de fiesta” moves from 9 to 8.
Search interest in Spain: brand and fashion categories
Compared weekly series for the exact expressions “Lola Casademunt,” “vestidos de fiesta,” “cazadora” and “moda mujer.”
interest index (0–100)
Tap or hover for details. Select a series above to show or hide it.
Collapse notes
Explore the data table
| Category | Lola Casademunt | vestidos de fiesta | cazadora | moda mujer |
|---|---|---|---|---|
| 2025-10-26/2025-11-01 | 9 interest index (0–100) | 8 interest index (0–100) | 56 interest index (0–100) | 16 interest index (0–100) |
| 2025-11-02/2025-11-08 | 10 interest index (0–100) | 9 interest index (0–100) | 65 interest index (0–100) | 16 interest index (0–100) |
| 2025-11-09/2025-11-15 | 9 interest index (0–100) | 8 interest index (0–100) | 67 interest index (0–100) | 17 interest index (0–100) |
| 2025-11-16/2025-11-22 | 12 interest index (0–100) | 8 interest index (0–100) | 75 interest index (0–100) | 18 interest index (0–100) |
| 2025-11-23/2025-11-29 | 18 interest index (0–100) | 9 interest index (0–100) | 93 interest index (0–100) | 21 interest index (0–100) |
| 2025-11-30/2025-12-06 | 11 interest index (0–100) | 9 interest index (0–100) | 74 interest index (0–100) | 16 interest index (0–100) |
| 2025-12-07/2025-12-13 | 13 interest index (0–100) | 9 interest index (0–100) | 70 interest index (0–100) | 17 interest index (0–100) |
| 2025-12-14/2025-12-20 | 11 interest index (0–100) | 9 interest index (0–100) | 66 interest index (0–100) | 15 interest index (0–100) |
| 2025-12-21/2025-12-27 | 13 interest index (0–100) | 8 interest index (0–100) | 81 interest index (0–100) | 16 interest index (0–100) |
| 2025-12-28/2026-01-03 | 13 interest index (0–100) | 8 interest index (0–100) | 78 interest index (0–100) | 17 interest index (0–100) |
| 2026-01-04/2026-01-10 | 21 interest index (0–100) | 7 interest index (0–100) | 100 interest index (0–100) | 19 interest index (0–100) |
| 2026-01-11/2026-01-17 | 14 interest index (0–100) | 10 interest index (0–100) | 63 interest index (0–100) | 16 interest index (0–100) |
| 2026-01-18/2026-01-24 | 12 interest index (0–100) | 10 interest index (0–100) | 59 interest index (0–100) | 15 interest index (0–100) |
| 2026-01-25/2026-01-31 | 13 interest index (0–100) | 11 interest index (0–100) | 55 interest index (0–100) | 14 interest index (0–100) |
| 2026-02-01/2026-02-07 | 11 interest index (0–100) | 12 interest index (0–100) | 51 interest index (0–100) | 15 interest index (0–100) |
| 2026-02-08/2026-02-14 | 9 interest index (0–100) | 11 interest index (0–100) | 45 interest index (0–100) | 16 interest index (0–100) |
| 2026-02-15/2026-02-21 | 8 interest index (0–100) | 13 interest index (0–100) | 43 interest index (0–100) | 18 interest index (0–100) |
| 2026-02-22/2026-02-28 | 10 interest index (0–100) | 12 interest index (0–100) | 46 interest index (0–100) | 18 interest index (0–100) |
| 2026-03-01/2026-03-07 | 9 interest index (0–100) | 15 interest index (0–100) | 43 interest index (0–100) | 16 interest index (0–100) |
| 2026-03-08/2026-03-14 | 9 interest index (0–100) | 16 interest index (0–100) | 43 interest index (0–100) | 17 interest index (0–100) |
| 2026-03-15/2026-03-21 | 11 interest index (0–100) | 15 interest index (0–100) | 45 interest index (0–100) | 18 interest index (0–100) |
| 2026-03-22/2026-03-28 | 11 interest index (0–100) | 14 interest index (0–100) | 46 interest index (0–100) | 19 interest index (0–100) |
| 2026-03-29/2026-04-04 | 9 interest index (0–100) | 15 interest index (0–100) | 47 interest index (0–100) | 17 interest index (0–100) |
| 2026-04-05/2026-04-11 | 9 interest index (0–100) | 15 interest index (0–100) | 35 interest index (0–100) | 16 interest index (0–100) |
| 2026-04-12/2026-04-18 | 10 interest index (0–100) | 15 interest index (0–100) | 34 interest index (0–100) | 17 interest index (0–100) |
| 2026-04-19/2026-04-25 | 10 interest index (0–100) | 14 interest index (0–100) | 31 interest index (0–100) | 18 interest index (0–100) |
| 2026-04-26/2026-05-02 | 9 interest index (0–100) | 14 interest index (0–100) | 31 interest index (0–100) | 18 interest index (0–100) |
| 2026-05-03/2026-05-09 | 10 interest index (0–100) | 15 interest index (0–100) | 33 interest index (0–100) | 17 interest index (0–100) |
| 2026-05-10/2026-05-16 | 8 interest index (0–100) | 14 interest index (0–100) | 34 interest index (0–100) | 15 interest index (0–100) |
| 2026-05-17/2026-05-23 | 9 interest index (0–100) | 12 interest index (0–100) | 28 interest index (0–100) | 15 interest index (0–100) |
| 2026-05-24/2026-05-30 | 9 interest index (0–100) | 12 interest index (0–100) | 21 interest index (0–100) | 16 interest index (0–100) |
| 2026-05-31/2026-06-06 | 11 interest index (0–100) | 10 interest index (0–100) | 20 interest index (0–100) | 15 interest index (0–100) |
| 2026-06-07/2026-06-13 | 12 interest index (0–100) | 11 interest index (0–100) | 20 interest index (0–100) | 15 interest index (0–100) |
| 2026-06-14/2026-06-20 | 12 interest index (0–100) | 9 interest index (0–100) | 17 interest index (0–100) | 15 interest index (0–100) |
| 2026-06-21/2026-06-27 | 12 interest index (0–100) | 8 interest index (0–100) | 17 interest index (0–100) | 13 interest index (0–100) |
| 2026-06-28/2026-07-04 | 12 interest index (0–100) | 9 interest index (0–100) | 17 interest index (0–100) | 12 interest index (0–100) |
| 2026-07-05/2026-07-11 | 13 interest index (0–100) | 10 interest index (0–100) | 18 interest index (0–100) | 12 interest index (0–100) |
| 2026-07-12/2026-07-18 | 12 interest index (0–100) | 9 interest index (0–100) | 17 interest index (0–100) | 12 interest index (0–100) |
| 2026-07-19/2026-07-25 | 13 interest index (0–100) | 9 interest index (0–100) | 21 interest index (0–100) | 12 interest index (0–100) |
| 2026-07-26/2026-08-01 | 11 interest index (0–100) | 9 interest index (0–100) | 20 interest index (0–100) | 13 interest index (0–100) |
| 2026-08-02/2026-08-08 | 10 interest index (0–100) | 9 interest index (0–100) | 21 interest index (0–100) | 13 interest index (0–100) |
| 2026-08-09/2026-08-15 | 10 interest index (0–100) | 9 interest index (0–100) | 23 interest index (0–100) | 11 interest index (0–100) |
| 2026-08-16/2026-08-22 | 11 interest index (0–100) | 11 interest index (0–100) | 27 interest index (0–100) | 13 interest index (0–100) |
| 2026-08-23/2026-08-29 | 10 interest index (0–100) | 11 interest index (0–100) | 39 interest index (0–100) | 15 interest index (0–100) |
| 2026-08-30/2026-09-05 | 10 interest index (0–100) | 11 interest index (0–100) | 33 interest index (0–100) | 14 interest index (0–100) |
| 2026-09-06/2026-09-12 | 8 interest index (0–100) | 11 interest index (0–100) | 36 interest index (0–100) | 16 interest index (0–100) |
| 2026-09-13/2026-09-19 | 8 interest index (0–100) | 11 interest index (0–100) | 36 interest index (0–100) | 14 interest index (0–100) |
| 2026-09-20/2026-09-26 | 8 interest index (0–100) | 8 interest index (0–100) | 39 interest index (0–100) | 15 interest index (0–100) |
| 2026-09-27/2026-10-03 | 9 interest index (0–100) | 8 interest index (0–100) | 41 interest index (0–100) | 15 interest index (0–100) |
| 2026-10-04/2026-10-10 | 9 interest index (0–100) | 9 interest index (0–100) | 49 interest index (0–100) | 16 interest index (0–100) |
| 2026-10-11/2026-10-17 | 9 interest index (0–100) | 8 interest index (0–100) | 68 interest index (0–100) | 16 interest index (0–100) |
The annual series puts that movement in context: “cazadora” reached 100 in the week of Jan. 4-10 and recorded 17 during July 12-18. The comparison uses one common relative-interest scale and retains complete weeks from Oct. 26, 2025, to Oct. 17, 2026. The figures express the relative importance of those searches, rather than sales or percentages of shoppers.
For an online store, a category query can provide a different entry point from a brand-name search. A product page has to serve someone looking for a garment as well as someone already familiar with the collection. Lola Casademunt’s mechanism operates within the next stage of that journey: once a person has an image or an item, it proposes other products with a specific relationship to it. The search series supplies demand context; the case explains how the store organizes further exploration.
From recognizing a garment to finding the outfit
Difficulty locating products is a recognizable ecommerce problem. A Coveo and Arlington Research survey of 4,000 shoppers in the United States and United Kingdom, conducted Dec. 14-19, 2025, finds that 31% encounter too many choices or unhelpful filters. Difficulty searching or browsing is reported by 30%, and irrelevant recommendations by 28%. This is international context about reported online shopping problems, with multiple responses possible.
Finding the product remains a shopping friction
Reported product-discovery problems in US and UK online stores, December 2025.
Tap or hover for details. All values are also available in the data table.
Collapse notes
Coveo and Arlington Research online survey of 4,000 shoppers aged 18+ in the United States and United Kingdom, December 14–19, 2025. Multiple responses: categories can overlap. Context from other markets, not Lola Casademunt results.
Explore the data table
| Category | Respondents |
|---|---|
| Too many choices or unhelpful filters | 31% |
| Difficult search or browsing | 30% |
| Irrelevant product recommendations | 28% |
| Slow site or errors | 28% |
| No stock-availability or availability-by-store selection | 27% |
| Not mobile friendly | 15% |
| None of these problems | 18% |
Múltiplo’s proposal for Lola Casademunt starts with a visible relationship: the main product and its accessories appear in the same photograph. The module identifies and describes the items in the image to offer the complete outfit. It also establishes similar alternatives by style or color and recommendations for complementary products. The commercial relationship can therefore address both looking for another option and completing a combination.
Daniel Cuenca, identified as Múltiplo’s CTO in the project account, explains that the team first examined computer-vision libraries to detect objects and find candidate products. According to his account, those tools were fast but did not provide the depth required for the client’s objective. The described solution uses GPT-4o and GPT-4o-mini to classify and relate products at a scale of thousands of items.
The technical decision is tied to a specific commercial task. Detecting a garment in a photograph is the beginning; a useful recommendation has to lead to related products in the catalogue. The consultancy presents PLai Related Products, built on PLai Framework, as the module organizing those relationships for online stores. The “total look” gives an outfit structure, while alternatives keep the search open when the first item is not suitable.
A contextual recommendation helps at a particular moment
The value of shopping guidance also depends on the help a person expects. An EAE Business School study published in February 2024, involving 1,000 consumers in Spain, finds that 56.9% value advice from a real person and 28.9% appreciate notifications about products that interest them. Some 9.1% say they do not mind AI recording their preferences to receive advertising. These are different attitudes, expressed through statements that can apply to the same person.
Human advice and personalization address different expectations
Attitudes toward brand interactions in Spain, reported in a study published in February 2024.
Tap or hover for details. All values are also available in the data table.
Collapse notes
EAE Business School report involving 1,000 consumers in Spain; its original profile includes ages 18–70. Published in February 2024, without exact fieldwork dates. Statements can overlap and do not form exclusive segments. Accepting preference recording for advertising differs from valuing a contextual recommendation.
Explore the data table
| Category | Respondents |
|---|---|
| Likes advice from a real person | 56.9% |
| Likes notifications about products of interest | 28.9% |
| Dislikes personalized interactions | 12.9% |
| Does not mind shopping history or name being known | 10.8% |
| Does not mind AI recording preferences for advertising | 9.1% |
| Does not know | 7.1% |
An outfit recommendation works with context the customer is already viewing: the photograph and the selected garment. The description of Lola Casademunt’s project focuses on identifying items and their visual relationships. That information explains the proposition without equating a product combination with storing personal preferences for advertising.
The catalogue described by Múltiplo includes clothing for special occasions and everyday wear, alongside costume jewelry, handbags, footwear and accessories. That variety gives a recommendation crossing categories a purpose. A product page can show how a garment fits with other elements in the same image and offer alternatives by style or color. Its usefulness lies in making the relationship understandable while leaving customers to decide which pieces to explore or choose.
EAE’s finding about the preference for human advice adds another distinction: assistance takes different forms. Contextual product relationships address a specific part of navigation, while personal advice responds to broader questions. The 2024 study helps distinguish those expectations; the documented module addresses the connection between image and catalogue.
Images and product pages need to speak the same language
Relating products at scale requires content that remains usable. In a Canto and Ascend2 survey of 434 content and creative professionals in the United States and United Kingdom, conducted in November 2025, 40% identify growing content demand across channels as a challenge. Maintaining brand consistency accounts for 31%, and locating or accessing assets for 30%.
Content work combines volume pressure and consistency
Challenges reported by US and UK content and creative professionals, November 2025.
Tap or hover for details. All values are also available in the data table.
Collapse notes
Canto and Ascend2 online survey of 434 content and creative professionals, November 2025; United States 51% and United Kingdom 49%. Values from the challenges question in the report published in January 2026, page 10. Multiple responses. This context does not measure Lola Casademunt processes or results.
Explore the data table
| Category | Respondents |
|---|---|
| Meet content demand across channels | 40% |
| Manage growing content volume amplified by generative AI | 36% |
| Find or implement the right technology | 36% |
| Align teams and workflows | 34% |
| Allocate budget and resources | 32% |
| Maintain brand consistency | 31% |
| Find and access assets | 30% |
| Deliver at the speed of business | 22% |
| Other | 1% |
For Lola Casademunt’s project, Múltiplo identifies object-detection precision, costs and resources, and running processes across thousands of items in multiple categories among the challenges. The image therefore enters two related tasks: identifying products to combine and producing information describing them. A recommendation can display a visual relationship; the product page has to explain each item in the brand’s language.
The consultancy says descriptions generated from images support product-text creation and translation into the online store’s languages. Lola Casademunt’s tone and voice are part of the brief. Coherence extends beyond whether garments work together in a photograph to how they are presented in writing across product pages and languages.
That connection matters to the cost of maintaining a changing catalogue. Manually establishing relationships, describing items and adapting text multiplies with each category. The solution brings those operations together around visual product information. The economic analysis concerns that organization of work as well as the recommendation customers eventually see.
SEO supplies the commercial direction for generated text
The same Canto and Ascend2 research asks a different question about the tasks for which teams use AI. Optimizing content for SEO or performance records 40%; automated tagging and organization, 39%; and improving discovery or recommendations, 38%. Automated content creation also accounts for 38%. The categories overlap and describe reported uses among the professionals surveyed.
AI uses span SEO, organization and recommendations
Reported AI applications in a US and UK content and creative survey, November 2025.
Tap or hover for details. All values are also available in the data table.
Collapse notes
Current-AI-use question in the original Canto and Ascend2 report, page 25. Online survey in November 2025; full study sample of 434 US and UK content and creative professionals. The report does not print a separate response base for this question here. Categories can overlap and describe reported uses, not performance improvements.
Explore the data table
| Category | Respondents |
|---|---|
| Analyze content performance and insights | 41% |
| Optimize content for SEO or performance | 40% |
| Organize through automated tagging | 39% |
| Create content | 38% |
| Improve content quality | 38% |
| Enhance discovery and recommendations | 38% |
| Automate repetitive tasks | 36% |
| Personalize for audience segments | 34% |
| Streamline content distribution | 33% |
| Other | 1% |
Múltiplo’s project account specifies where the positioning input for Lola Casademunt’s descriptions comes from: keywords identified in analyses produced by the client. The image supplies product information; the brand supplies its voice and those positioning analyses. The system combines those inputs to prepare descriptions the consultancy presents as detailed and SEO-oriented, while also supporting their translation.
This division directs text generation toward commercial needs identified by the client. Product relationships let people continue exploring inside the store; descriptions carry garment attributes into product pages and the search terms the brand works on.
Isabel Gómez, identified as Lola Casademunt’s CRM Manager in the supplier’s published testimonial, attributes improved shopping experience and processes to the module. She describes conversion in terms of an expectation, saying the team hopes it will increase the conversion rate. Her assessment places the work within store experience and operations, without turning that expectation into measured uplift.
The case shows an AI application defined by the journey it enables: moving from an image to its constituent products, finding another option and accessing descriptions adapted to the catalogue. The commercial value of the “total look” depends on that continuity between inspiration, choice and product content. Technology provides capacity to construct the relationships; the brand retains direction over presentation and the search needs it wants to address.
In this article
Marian Vasile
Marian Vasile is a MarTech and growth professional. His work spans product marketing, SEO, paid advertising and deep online tracking, including behavioral analytics and attribution.
