Yana Yantres
Yana Yantres
A reinvention of Semrush's Keyword Manager — from a passive place to store keywords into a tool that collects, clusters, and prioritizes them into a ready-to-execute content strategy.
Context
Semrush is an all-in-one digital marketing platform used by 10 million users. Its Keyword Manager existed as a storage utility, a place to save keywords collected elsewhere. It gave users data, but no direction.
The opportunity
Transform Keyword Manager from a passive list tool into an end-to-end keyword strategy experience. One that automates the hardest parts of the workflow and delivers a prioritized content plan, not just a pile of data.
My role
I took the project from initial discovery all the way through launch, managing research, product strategy, Alpha and Beta phases, UX design, validation, and delivery.
The Problem
SEO specialists had to leave Semrush to do their actual work (Keyword Cupid, WriterZen, Frase, Surfer, Ahrefs). Every one is a tool a Semrush user pays someone else for.
The painful loop
Collect
Pull keywords from many sources
Clean
Review and delete the irrelevant ones
Cluster
Group into topics manually or in a distrusted tool
Prioritize
Ranking it all in Excel by hand
Hand off
Turn it into content briefs
Our research revealed how SEO specialists work today and there's huge potential for automation. That's exactly what we're going to build.
The opportunity
Building a new tool that's intuitive and easy to use could drive significant growth, both in new traffic and in existing Semrush users adopting it.
Discovery
The first idea came from breaking down the standard keyword strategy workflow to see which steps could ideally be automated.
Breaking down the workflow showed which parts could be automated. That led to the idea of a Seed Keyword list: the user uploads keywords, and the tool collects the full keyword set and sorts it into groups.
Testing the first MVP
A first MVP was launched to test it with real users and real data. What we found:
Audience
The product was designed for middle-level SEO specialists, while providing enough transparency for senior specialists to trust the results.
Design Process
Through several rounds of testing where the data kept getting better it became clear that users in this audience wanted a strategy layer to make their research simpler.
The build was split into 4 parts, so new functionality could be added carefully and gradually while keeping a close eye on the metrics.
Final Solution
Users told us they wanted a ready-made strategy, not just a list to manage.
But 28% still wanted to build their list by hand. So we rolled it out in stages: the old "Create list" option stayed next to the new Strategy Builder, so users could switch at their own pace.
72% moved to the new flow within 1 year, conversion from "create list" to "open list" rose 43%.
I still want to be able to access the list I have just built up of keywords on the standart table.
User
The goal was to bring users into the new tool for whom this functionality would actually be relevant — not just any traffic. Mapping SEO workflows showed a natural fit: Keyword Overview, a neighboring Semrush tool where users already work closely with keywords.
The team partnered with that tool's owners to place the feature there. To avoid feeling like an ad in a dense report, "Keyword Clusters" became a native third column, styled like the existing ones, with a few real clusters visible and the rest blurred as a teaser.
The channel outperformed expectations
Impressions
Clicks on the widget (creating a list)
Why build it: our core audience is beginners, so we're moving away from Semrush's typical tables toward visuals that match what a newcomer expects.
The goal is for it to look good enough to go viral — shareable on social media. Looking at the Mind Map, a user should think: "This is cool, I want to drop it into a deck for my manager or client."
Users needed more than a flat keyword list — they needed to see how to actually organize content on their site.
So clusters were reframed as "Topics": each Topic has one Pillar page and several Subpages.
The list result now opens into an interactive structure overview instead of a table, so users immediately see how their keywords are already organized into topics and nested pages.
Success was validated with a prototype test across both single- and multi-seed-keyword cases, targeting an 80% task-completion rate among existing and new keyword-tool users.
Skeptical experts as test audience → treated as the toughest validation; designed for override, not perfection.
Users don't need less data — they need better-structured data with visible reasoning. The biggest adoption barrier for senior SEOs wasn't the interface. It was trust. Every design decision that made the logic transparent (editable clusters, metric context, named presets) directly contributed to retention.
Quality at the tail matters more than average quality. Even when 85–90% of clusters were correct, the remaining 10–15% drove disproportionate negative feedback. Fixing the algorithm was the highest-leverage design decision in the project — and it was made from data, not intuition.
Influence in action: when beta clustering feedback arrived, the default reaction was "tune and re-ship." I redirected the team to define and measure quality first — which surfaced the three-list-type insight and prevented us from over-optimizing for a number we could never hit.
Add a strategy for what to do and how
Add a new list type, by domain
