eCosentino
Auditing a B2B platform in an industrial giant
The context: accidental digitalization
Cosentino is an industrial surfaces manufacturer turning over around 1.4B euros globally. Over the years, the business expanded rapidly in the physical world, but digital growth happened through isolated initiatives rather than a centralized product vision. Most digital touchpoints lived as separate portals built on Salesforce Communities, each answering to a different business unit.
At the end of last year, the company kicked off a migration to move these private portals to a custom web app. We started auditing peripheral services like Wecommerce for kitchen studios and C-Top for interior designers, and we quickly found a major bottleneck: teams were maintaining over 14 different implementations of the exact same basic patterns (wizards, file uploaders, order trackers, table filters). We were developing everything three times over, duplicating bugs and piling on technical debt.
At that point, management and the UX team paused. It made no sense to define a unified sales experience starting with the smaller tools while ignoring eCosentino, which handles the vast majority of digital orders and over 60M euros in revenue. We shifted focus to audit eCosentino from the ground up to establish the foundation of a single, scalable B2B platform.
Field discovery vs. digital telemetry
We paired empirical telemetry (VWO session recordings and Google Analytics) with direct on-site visits to stonemason workshops across Spain and the US to contrast what users say with what they actually do.
89% Search reliance (used 5x more than manual filters)
26% Average homepage scroll depth
1.5% Interaction on merchandising carousels
Spending time on factory floors made the reality obvious. A fabricator covered in stone dust switching between a wet CNC saw and an office laptop operates under completely different mental models than a retail consumer.
The data backed this up: users bypassed the catalog almost entirely to use the search bar. Yet the search engine was rigid, demanding exact text matches and breaking whenever browser auto-translators kicked in (like translating the finish "Suede" into "Gamuza"). The homepage was treated as a dead transit screen: people interacted almost exclusively with the top navigation bar to jump straight to their task.
"In the search bar, as soon as you type 'Enzo 12mm' together, it fails. You have to clear filters, check if it's meshed or not... you lose so much time."
— Fabricator
1st click heatmap
What happens when the system breaks trust
To understand where the platform failed backstage, we documented 20 core use cases into as-is service blueprints across seven operational layers: user actions, front-end views, back-end logic, core ERP and pricing systems (SAP, PCM), commercial teams, automated communications, and pain points.
We uncovered 66 distinct problems across the board (34 technical/UX bugs and 32 structural failures). The core friction came from data reliability and operational rigidity:
1. Catalog SKU fatigue
Every combination of format, thickness, and finish was listed as a separate product card. Users had to scan dozens of visually identical thumbnails, causing a 64% drop-off before reaching product sheets. Critical attributes like whether a slab was unmeshed ("Sin Malla") went unnoticed, leading to wrong cuts on CNC machines and expensive returns.
Product cards and Meshed atribute
2. The cart as a workaround calculator
Around 32% of users abandoned their cart because the catalog displayed list prices rather than contracted net rates. Fabricators added items to the cart simply to discover their real price, left without buying, and reconciled invoices manually on paper at month-end.
3. The "rule of 50" stock heuristic
Inventory numbers suffered from a few minutes of synchronization delay with SAP, meaning recent orders didn't reflect right away. Fabricators had internalized an informal rule: if the screen showed fewer than 50 slabs, they assumed it was zero because a single bulk container order from the US could wipe out that stock instantly. Instead of checking out online, they picked up the phone to call their local center rep.
"I don't look at screen prices anymore; I wait for the invoice. And if I see fewer than 50 slabs, I call Cristina at the Center to verify physically if the truck is coming."
— Fabricator
4. The dimensioned PDF workaround
Because the 3D configurator was rigidly parametric, lacked custom cutouts, and forced arbitrary structural supports across sinks, fabricators placed dummy orders just to download the dimensioned PDF drawing. They immediately deleted the order from their cart, modified the dimensions and cutouts manually by hand on the printed sheet, and sent it to the technical office to process manually.
Blueprint as-is slab bought
The strategic proposal
We presented the audit to the Executive Committee, proposing four architectural pillars to unify the sales experience:
Dual-mode checkout: A toggle between direct purchasing and formal quotations, allowing fabricators to save project estimates without clogging warehouse queues with fake orders.
Showroom privacy mode: A point-of-sale switch to mask trade discounts and display retail markups in front of end-clients.
Parent-child catalog structure: Consolidating scattered cards into material families with thickness and finish selectors directly on the card.
Core data alignment: Pushing for real-time inventory validation with SAP and dynamic regional proximity rules rather than rigid warehouse pickers that cleared user carts on center changes.
Outcomes and my personal delivery
Executive leadership valued the audit and recognized that the primary blocker was data integrity and operational friction, not visual styling. Because core improvements in pricing logic, SAP middleware, and payment automation required deep coordination across finance, logistics, and IT, the full e-commerce re-platforming was scheduled for the following roadmap cycle.
Effort-impact matrix/ Quick Wins
In the meantime, I took direct ownership of two high-leverage initiatives:
The relational measurement plan: I designed an end-to-end telemetry specification across Google Analytics and VWO. We mapped user funnels, micro-interactions, and context dimensions so our data partners could track reliable behavioral data across the catalog, slab inspection tools, and configurators.
Catalog & PDP quick wins: I designed and deployed immediate usability improvements across the live catalog and product detail pages for the other e-commerces, refining search tolerance, filter layouts, and card readability to reduce daily friction for kitchen studios and designers while backend refactoring progressed.