NexTrend already had strong customer satisfaction, recognised clients and founder-led demand. But it was a small team with many possible customers, channels and growth initiatives. My job became to identify which ICP deserved the company's limited resources, build the data needed to validate that choice and concentrate marketing around the opportunities most likely to generate revenue.
The business situation
NexTrend is a French B2B training company helping creative, marketing and production teams integrate generative AI into their workflows.
There was evidence that the business worked:
- strong client satisfaction;
- recognised customers;
- demand generated through the founder's network and LinkedIn;
- successful webinars;
- substantial subject-matter expertise.
But there was not yet enough structure to know where to double down.
The company faced several possible markets and many possible marketing activities. At the same time:
- ICP definition remained broad;
- customer and learner intelligence was scattered;
- CRM data was incomplete;
- pricing and revenue information were inconsistent;
- attribution was weak;
- webinars, content, website, lead capture and sales were disconnected.
For a small team, that creates a resource-allocation problem: which segment actually deserves more time, budget and attention?
Decision 1: Don't scale acquisition before the commercial data can be trusted
My first question was not:
Which channel should we add?
It was:
What does the existing business tell us about where revenue is already coming from?
I reviewed customers, revenue records, acquisition sources, sales conversations, satisfaction data, marketing activity and pricing.
The audit surfaced a material measurement problem. A bottom-up reconstruction from client records came out at roughly 2× the revenue figure being used internally, leaving a discrepancy close to six figures.
That was not a growth result. It was evidence that commercial data needed to be cleaned before the company could confidently decide where to invest.
Decision 2: Don't define the ICP as a branding exercise
The objective was not to create a persona document.
We needed to determine which customers represented the strongest combination of:
- pain;
- willingness to pay;
- product fit;
- satisfaction;
- repurchase potential;
- strategic value.
I combined historical customer evidence with new market hypotheses.
Segments including agencies, larger corporate teams and audiovisual / production organisations were then carried through the actual operating system.
ICP became something captured in:
- CRM properties;
- forms;
- lists;
- website messaging;
- webinars;
- prospecting;
- lead magnets;
- sales qualification.
That created a feedback loop:
- hypothesis
- audience
- campaign
- conversation
- opportunity
- revenue
- learning
Every interaction could now help narrow the company's focus.
Decision 3: Don't confuse audience volume with commercial traction
NexTrend's webinars could already generate significant demand. But the logs made an important distinction visible: the largest audience was not necessarily the most commercially useful audience.
For example, one webinar attracted 400+ registrations but skewed heavily toward freelancers and produced only a small number of people showing immediate buying intent. A later webinar attracted 300+ registrations, but the audience contained a much stronger concentration of agencies, large accounts and audiovisual-production companies, with several dozen people already comparing options or discussing the topic internally.
That is the kind of signal the system needed to surface.
400+ registrations can be a weaker commercial outcome than 300+ registrations.
So webinar performance could no longer be judged mainly through registrations.
The questions became:
- How many registrants match the ICP?
- Which companies are represented?
- What pain brought them in?
- Which attendees progress into conversations?
- Which conversations create pipeline?
- Which segment ultimately creates revenue?
Decision 4: Don't bolt more campaigns onto unreliable infrastructure
To make those decisions possible, HubSpot needed to become part of the GTM system rather than simply a contact repository.
I rebuilt the commercial data layer across:
- contacts;
- companies;
- deals;
- lifecycle stages;
- ICP properties;
- meeting flows;
- forms;
- sales views;
- source and discovery attribution.
The model eventually captured eight different acquisition and discovery sources.
That allowed marketing and sales activity to increasingly connect to the same commercial picture.
Decision 5: Don't leave customer knowledge trapped inside conversations
NexTrend already possessed valuable proprietary market intelligence.
It existed in customer feedback, training evaluations, sales calls, meeting transcripts, webinars and surveys.
But scattered information does not automatically become strategy.
I created a customer-intelligence layer that could answer:
- Which pains repeat by segment?
- Which use cases generate urgency?
- Which objections recur?
- How do customers describe the value themselves?
- Which proof resonates?
- Which topics should become content?
- Which signals suggest a segment deserves more attention?
That intelligence fed back into positioning, campaigns, sales conversations, website copy, webinars, content and offer decisions.
Decision 6: Don't stop at marketing when the offer creates the friction
Some constraints eventually appeared inside the product and delivery journey itself.
The training journey needed redesign work, including adjustments required for Qualiopi compliance and learner experience.
So the scope expanded into training journey and offer architecture.
Generating more leads into a weak or confusing journey would not solve the underlying growth problem.
What the system started to reveal
At the measurement point, marketing-origin activity was connected to roughly 10 active deals, including opportunities with well-known media and consumer brands. Individual opportunities ranged from smaller four-figure proposals to deals above €20K, and the confirmed marketing-origin pipeline had reached the mid-five figures.
The website also recorded roughly 2.8K traffic at the tracked point and the first five identified leads from LLM-driven discovery within three months of the rebrand.
What changed
| Before | After |
|---|---|
| Broad set of possible customers | ICP hypotheses tested through commercial evidence |
| Founder-led demand | Multiple connected demand motions |
| Revenue information difficult to reconcile | More structured commercial visibility |
| Webinar success measured heavily through registrations | ICP quality and buying-intent signals tracked |
| Customer insight scattered | Reusable customer-intelligence layer |
| Marketing and sales signals disconnected | Shared CRM and attribution structure |
| Limited marketing-origin pipeline visibility | Roughly 10 active deals / mid-five-figure confirmed pipeline |
What this case demonstrates
NexTrend shows how I approach an early-stage B2B business with limited resources.
The question is not:
How can we do more marketing?
It is:
Where is the strongest commercial signal, and what should we stop spreading resources across so we can concentrate there?
Marketing becomes both a distribution system and a learning system for the business: customer data, campaign response, conversations and revenue progressively tell the company where its winning ICP is likely to be.