We turned Nebius Academy's website into a lead engine — and its funnel into something sales could trust.
Nebius Academy sells corporate AI-adoption training across three regions. We built the full go-to-market lead engine on HubSpot — capture, ICP, scoring, and nurture — so a website visitor becomes a known, qualified, routable opportunity.
Client
For a considered B2B purchase — the kind a company weighs for weeks and signs off by committee — the distance between a website visitor and a sales conversation is where growth quietly leaks. Most of that traffic stays anonymous, unscored, and unrouted; the sales team hears about the good ones too late, or not at all.
Here is how we closed that gap for Nebius Academy: a go-to-market lead engine, built on the HubSpot they already run, that turns strangers into a pipeline sales actually trusts — across three very different regions.
Traffic they could see. Pipeline they couldn't.
Nebius Academy runs corporate training programs on AI adoption — practical AI education for teams putting the technology to work. Its buyers are businesses across LATAM, the US, and Europe, and its website is the front door: the place interest begins. But interest isn't pipeline. Marketing couldn't see who was engaging, sales couldn't tell which leads were worth a call, and no single number connected the two.
The brief was a classic go-to-market (GTM) problem, made harder by three regions that buy at different speeds and in different languages: turn website demand into a measurable, qualified, sales-ready pipeline on HubSpot — and make the whole funnel legible enough that marketing and sales finally agree on what a good lead is.
Make the funnel legible. Make the leads trustworthy.
A go-to-market engine, on the stack they already run
We're a senior CRM collective, and on a GTM build we do three things at once: get more out of the platform a client already runs — here, HubSpot — and coach the team to own it; instrument the entire funnel so nothing is invisible; and use Advanced Jobs-to-be-Done (AJTBD) to find who actually buys, and why.
The engine we built for Nebius has four moving parts, in the order a stranger travels through them: capture the intent, understand who it belongs to, score it honestly, and hand it off to sales with context.
Turn anonymous traffic into known, tracked intent
You can't qualify a stranger. So the first job was to give people good reasons to raise their hand — and to capture what they did next, not just their email. We built a set of lead magnets matched to where a buyer is in their thinking, upgraded the site's forms to progressively profile without adding friction, and instrumented custom web events so high-intent behavior became a signal we could act on.
Lead magnets
Practical assets — readiness guides, syllabi, ROI tools — matched to early, mid, and late intent, so raising a hand feels worth it.
Progressive forms
Forms that ask for a little at a time and remember the rest, capturing role, company, and need without the wall of fields that kills completion.
Custom web events
Pricing views, syllabus downloads, and repeat visits tracked as first-class signals — the behavior that separates a browser from a buyer.
All of it inside HubSpot. We built the capture layer on the platform Nebius already ran and coached the team to extend it — new magnets and events are theirs to add, not a ticket to us.
Who Nebius is really for — found with Advanced JTBD
A lead engine that treats everyone the same is just a louder version of the old problem. So before scoring or nurturing anyone, we asked who Nebius is truly for — using Advanced Jobs-to-be-Done to move past job titles and firmographics to the actual job a company hires AI training to do. We drafted several ideal-customer hypotheses, put them in front of the market, and let engagement and sales conversations narrow the field to the segments that genuinely convert.
Several hypotheses tested; the field narrowed to the highest-fit segments below.
Why it matters: a defined ICP is the backbone everything downstream leans on — the scoring model, the nurture tracks, and the language sales uses on the call. Get this wrong and you optimize the funnel for the wrong people.
A lead score both marketing and sales believe
With the ICP defined and behavior tracked, we built a lead scoring model that turns scattered signals into a single, trustworthy number — and, crucially, a shared definition of when a lead becomes a marketing-qualified lead (MQL) ready to route to sales. The model reads three kinds of signal, and it keeps learning from the one that matters most: what sales actually does with the leads it receives.
Declared data
Role, company size, region, and stated need from the progressive forms — fit against the ICP.
Behavioral signals
The custom web events — pricing views, downloads, return visits — weighted by how much intent they really carry.
Sales feedback loop
Which MQLs sales accepted, worked, or rejected — fed back to retune the weights so the score tracks reality, not theory.
We didn't set the model once and walk away. Reading the reports beside the sales team, we recalibrated the thresholds against accepted-lead outcomes until an MQL reliably meant a lead worth a rep's time.
Segmented nurture, and a funnel everyone can read
A qualified lead still needs to arrive warm. We built nurture campaigns per segment, each mapped to that buyer's job-to-be-done — the enterprise track leads with governance and proof; the tech-team track leads with depth — so leads mature toward a conversation instead of going cold. And we made the whole thing legible with end-to-end reporting: Subscribers → Leads → MQL → sales-qualified lead (SQL), with the conversion between every stage visible to both teams.
One shared view of the journey — every stage, and the conversion between them.
The point: one funnel, one language. When marketing and sales read the same stages and the same conversion rates, the argument about "lead quality" turns into a shared number they can move together.
A funnel that converts — and that sales stopped arguing with
Measured over the first two quarters on HubSpot, as the scoring model and nurture tracks matured. The headline isn't any single number — it's that the same journey now produces qualified, accepted pipeline predictably, in three regions at once.
Read this honestly. These are program-level figures from the first two quarters, and the biggest driver was the sales-feedback loop tightening the scoring model — so the sharpest gains are in lead quality (acceptance, MQL→SQL), not raw volume. Regional pace differs: the US matured fastest, the EU close behind, LATAM building through partners.
A go-to-market motion Nebius owns — and can repeat
- Capture — anonymous traffic became known, tracked intent, with lead magnets, progressive forms, and custom events feeding every downstream decision.
- Understand — an Advanced-JTBD ICP replaced guesswork, so the engine is tuned for the buyers who actually convert — not the ones who merely fill out forms.
- Score — a lead model, recalibrated against sales feedback, turned "is this a good lead?" from an argument into a number both teams trust.
- Hand off — segmented nurture and an end-to-end funnel report (Subscribers → Leads → MQL → SQL) gave marketing and sales one shared, legible view.
- How — every layer built on the HubSpot stack Nebius already owned, with the team coached to run and extend it across all three markets.
Kept honest: the strongest, most defensible gains are in lead quality — MQL→SQL conversion and sales acceptance — driven by the scoring feedback loop. Volume and revenue figures move with market conditions and sales capacity, so we scope this case to the funnel mechanics we built and can attribute.
If your website has traffic but your pipeline has fog
You just rolled out HubSpot
The platform is in, but the pipeline isn't flowing. You have the tool — you need the engine that makes it produce.
You're scaling into new markets
Growing across regions and languages, and you need one GTM motion that travels — instrumented the same way everywhere.
Sales & marketing disagree on leads
The "lead quality" argument never ends. You need a scoring model and a funnel both teams can actually trust.
Your traffic already knows who's ready
It's just not telling sales yet. Book a 20-minute look and we'll show you the pipeline hiding in the demand you already have.
Book a 20-minute look