The problem

Liberty's B2B brands (Flow Business, +Móvil, C&W Business and Enterprise, MasMóvil Negocios, BTC and Liberty Networks) ran a sprawl of websites across 15 countries on a non-modular architecture. Every change was slow and costly, the experience shifted from one market to the next, and the team spent its time firefighting instead of building on past work. The business needed to ship more, and faster, without giving up quality or control.

What I did

• Built and led a distributed, cross-functional team across UX, UI, content strategy, analytics, BA/QA and engineering, with full accountability for experience strategy, roadmap and delivery.
• Introduced a modular (headless-ready) architecture and an atomic design system with a shared component library, so new sites and updates assembled from reusable parts instead of being rebuilt each time.
• Redesigned the operating model across the three players involved, market stakeholders, the digital team and external agencies, clarifying ownership, speeding up the content matrix and making better use of capacity.
• Stood up a B2B Digital Center of Excellence and a digital-maturity program that moved from assessment to new capabilities to a repeatable design process, to keep raising the bar across markets.

The results

The modular system turned a maintenance burden into a delivery engine. Engagement compounded year over year: organic sessions rose about 25%, total sessions about 21%, pageviews about 22% and leads about 18% in the measured window. The same foundation then scaled cleanly into new Eastern Caribbean markets, starting with St. Maarten and continuing into Montserrat, Anguilla and Antigua.

"Natalia manages the team responsible for our websites across the Caribbean for Enterprise, SMB and Wholesale. She is a skilled UX leader who guides her team to deliver clear, user-centric digital experiences, and I recommend her to any team that wants thoughtful, customer-oriented solutions."

William Andrés Suárez OrtizCustomer Experience Executive, Liberty Latin America

Program-level financials (CAPEX / OPEX) are confidential and intentionally excluded. Metrics above are drawn from the team's own quarterly reporting and shown as percentages and counts only.

The problem

Marketing could see sessions and forms. Sales could see opportunities in Salesforce. But nobody could draw a line between the two: which traffic source, which page, which interaction actually produced a qualified lead and, in the end, revenue. Investment decisions were being made on vanity traffic numbers rather than on what really converted.

What I did

• Designed a LeadID data model and a DataLayer implemented across every page and every form, capturing context from the moment a user arrived to the moment a lead was created.
• Unified measurement across Google Analytics, Pardot and Salesforce so a single lead could be traced from traffic source, to on-site behavior, to form, to Salesforce opportunity.
• Created the custom dimensions in analytics and the corresponding field in Salesforce, and governed the model so the two data layers stayed consistent across markets.

The results

This turned the websites from a brand surface into a measurable revenue channel. Stakeholders could finally see "lead converted vs. opportunity" by source, the kind of decision-grade data that changed where the team invested next. It is the same idea I write about as decision intelligence: the value is not the dashboard, it is the decision the dashboard makes possible.

Figures shown as percentages; underlying revenue and budget data are confidential and excluded.

The problem

The enterprise and wholesale web properties (C&W Enterprise, Liberty Networks) were dated, inconsistent and hard to maintain, and they didn't support the new brand or the company's regional expansion. A surface-level reskin wouldn't last, so the architecture itself had to change.

What I did

• Led the redesign on a modular approach built to seed a future headless architecture across C&W Networks regions and Enterprise LatAm, a reusable baseline rather than a one-off site.
• Drove the rebrand of C&W Enterprise into Flow Enterprise: new technology, stronger site security, an "Expert Voice" resources section, and a solutions experience designed to guide customers to the right product.
• Delivered the Liberty Networks wholesale overhaul, then a second phase adding a Spanish version and improved search and news/events sections.

The results

The rebrand was an architecture decision, not a paint job. By building the modular baseline first, each new market launched faster and stayed consistent, while the engagement program of interactive banners, a stronger knowledge section and entrepreneur-focused offers targeted a 50% lift in conversion on the core Flow Business funnel.

CAPEX / OPEX figures from the source plans are confidential and deliberately omitted.

The problem

AWRA started from a consumer hypothesis: that loneliness was a large enough, urgent enough problem that people would pay for an app to help with it. The pain was easy to find evidence for. Whether it was a business was a different question, and the honest answer was that I didn't know.

Conventional discovery would have answered it in four to six weeks, between recruiting, scheduling, moderating, transcribing and synthesising. That timeline has a hidden cost: it is slow enough that teams start building in parallel, and by the time the research lands the decision has already been made by momentum. I wanted the answer before that happened.

What I did

• Defined the decision first, not the study. One question had to be answerable at the end: is this a business, yes or no. Everything else in the design served that.
• Built an AI interview agent that conducts the session conversationally and generates its own follow-up questions based on what the participant actually says, rather than reading a fixed script. Two days from first line to working product.
• Designed the synthesis layer, not just the interview. The agent codes every conversation into structured dimensions, the type of loneliness described, the underlying pain, its urgency, and willingness to pay, so output arrives as filterable evidence rather than a pile of transcripts to read.
• Ran the study to saturation and made the call. Twenty participants, past the twelve-interview threshold commonly cited for thematic saturation in qualitative discovery.
• Wrote the decision down, including what would have to be true for the answer to change.

The results

The finding was specific, and it was not the one I wanted. The pain was real and participants described it clearly. But in the target age range it sat well below the threshold that drives people to act: they were not looking to solve it with an app, and they were not willing to pay for one. A real problem and a viable business turned out to be two different things.

What the week bought was not a study. It was the months of engineering that would have gone into a product nobody was going to pay for, and the confidence to say so out loud with evidence behind it instead of a hunch.

The instrument outlived the hypothesis. Turning scattered conversation into structured, decision-ready insight is a problem enterprises have too, and the model now sits in AWRA's product architecture as a governed, human-in-the-loop agent rather than a consumer feature. The research tool built to test one business became a component of another.

What I'd carry into a team

• Discovery is worth accelerating precisely because slow research gets overtaken by momentum. The value of a one-week cycle is that the decision is still open when the evidence arrives.
• AI moderation earns its place in breadth, consistency and speed. It does not replace a skilled researcher where rapport, reading hesitation, or following genuine surprise decides the session. Knowing which study is which is the actual skill.
• An agent that only asks questions saves you very little. The leverage is in the synthesis layer, where conversation becomes something a team can filter, compare and act on.
• The most useful outcome a discovery process can produce is permission to stop.

AWRA is in development and not publicly available, so no live link is included. Participants took part under a confidentiality commitment: findings here are aggregate only, with no names, quotes or identifying detail. Commercial projections for AWRA are excluded throughout. Unlike the Liberty programs above, this case documents product definition and a decision, not delivered business results.