UX Research & Product Design
Delfi Monitor
Designing a B2B media monitoring SaaS from zero, against an incumbent that had owned the market for 25 years.
Delfi Monitor mockup with desktop and mobile screens.
Timeline
2024, research through launch
Constraint
Ship fast
My Role
I lead design at Delfi, and I took this project on personally rather than delegating it: research, interaction design, design system, and launch. For a 0-to-1 product where every design decision depended on what the research said, I wanted no translation layer between the user interviews and the screens. Development was Texta, an agency partner whose machine learning engine the product is built on. The project was led by Delfi's Head of Business Development, with executive sponsorship from group level.
OUTCOMES: Product 0 to 1 in half a year · Research-led: every major decision traced to an interview finding · Part of the B2B bundle that closed out the 2025 year-end subscription goal.
Context
Delfi Meedia is Estonia's largest digital media company, with over 20 portals and more than a million monthly readers. For decades, the PR agencies and communications teams who needed to track media coverage all used the same tool: Station, a 25-year-old monitoring platform with no real competitor in the market. A product people tolerate because there is no alternative.
In 2024, Delfi decided to build the alternative. The structural advantage was obvious once stated: Delfi owns the content. That means it can show how many people actually read a specific article, data Station has never had access to.
The research
At first I ran six interviews: two Delfi journalists and four external users from PR agencies and corporate communications roles. I wrote the script, recruited the participants, ran every session, and synthesised the findings myself. Three of them shaped the entire product.
People would switch for one thing: real read counts. Every external user raised it unprompted. Station shows publication-level audience figures, so a PR consultant can tell a client "Delfi has 90,000 readers" but not whether their press release got 200 reads or 20,000. Several users said this single data point justified switching providers.
Users couldn't change what they were tracking without asking permission. In Station, keywords and filters are fixed - set up by their team, and changing them means contacting product support and waiting. One consequence users described is an anomaly where relevant coverage never reaches them at all, because it falls outside the keywords someone else configured months ago. So we made monitoring self-serve. Users add, edit, and save their own keywords and filters instantly, and build their own saved searches without a support ticket in the loop.
The bar was clarity, not features. One consultant had used Station for twelve years and told me he still couldn't fully use it. There is a 22-page manual; his team navigated by tribal knowledge instead. When he saw the Monitor prototype, his reaction wasn't "this is better." It was "this is usable." That reframed the whole project. We weren't building for power users who would master a complex system. We were building for people who need an answer in 30 seconds before their first meeting.
Delfi Monitor product in use on desktop.
What I cut
Three things came out of scope v1, all to protect the launch date. Social media monitoring was the most requested feature across all six interviews; I made the argument to stakeholders directly that the development complexity would have pushed launch by months, and the core job of tracking media coverage had to work perfectly before we widened the scope. The reporting module - branded PDF exports for clients, went the same way: real demand, but not core to v1. And mention importance: users couldn't tell a headline from a buried stock-ticker line, so I designed a ranking by where a keyword landed (headline, lead, or deeper), but we scoped the ranking itself out and shipped a lighter answer, a toggle to jump straight to where a keyword appears. A media monitor product with core features shipped on time and in budget beats a broad one shipped late. The argument held, and we launched on schedule.
The product
Delfi Monitor is search-first. The morning workflow it's built around: search a client's name, filter to the last 24 hours and the publications that matter, scan results with live view counts and mention importance indicators, save the search for tomorrow. The interface is a three-panel layout: navigation, results, article preview.
Delfi Monitor product in use on desktop.
Every major design decision traces back to the research:
Newest first by default. Sounds obvious. Station doesn't do it, and every user in testing expected it.
Clickable article titles. Station requires a separate icon click. Five users in testing clicked the title and were confused when nothing happened.
Self-serve keyword management. Changing monitored keywords in Station means contacting their editorial team. In Monitor, users edit their own keywords instantly.
View counts inline on every result. The switching feature, shown as standard data rather than an upsell.
Keyword Suggester. Built on Texta's ML engine, it proposes related terms the user hadn't thought to monitor, cutting the manual work of building a complete search setup.
Saved searches with scheduled email alerts. Users described checking Station the way they check email. We brought the results to their inbox instead.
A "show only keyword locations" toggle, requested explicitly in usability testing, so users can scan where their keyword appears without reading the full article.
The UI is in Estonian, matching the product's market. The design system was built from scratch within Delfi's brand guidelines.
The design system of Delfi Monitor.
Outcome
Delfi Monitor launched on schedule and became a core component of Äripakett, Delfi's B2B subscription bundle, which drove a significant share of new subscriptions and helped Delfi hit its annual target.
To be precise about attribution: the bundle closed those subscriptions, not Monitor alone. What Monitor did was make the bundle credible — clients could see they were buying a real tool, not a placeholder feature.
What I'd do differently
Of those three cuts, two I stand by and one I'd reverse.
Social media monitoring was the right call, though it stays the gap between a compelling alternative and a full replacement for high-value agency clients. It's the obvious next chapter.
Mention importance is the one I got wrong. I treated the ranking as a nice-to-have and protected the launch date instead. But it was arguably the single feature that best proved Monitor understood the job better than the incumbent, turning a wall of equal-weight mentions into a usable signal. The toggle we shipped helped, but it wasn't the ranking. If I ran it again I'd push it in.
