case study-leaming ecosystem design
They didn't need more training. They needed us to ask better questions.
A B2B SaaS company brought me in to fix their onboarding. Completion
was at 31%. Satisfaction scores were high. Customers were still leaving. What we found beneath the surface – a confidence trap, a language barrier between two teams, an exhaustion problem hiding as a technical one, and a human conversation nobody knew how to have everything about how we designed the solution.
CLIENT
Claros (B2B SaaS)
MY ROLE
Lead Learning Experience Designer
DURATION
5 months
SCOPE
Discovery Facilitation Ecosystem design Digital learning experience
What we achieved
-31%
Reduction in 18-month
churn
-34%
Support ticket volume,
weeks 4-12
-5 wks
Time-to-proficiency: 11
→ 6 weeks
+41%
Feature adoption at 90-
day mark
case study-leaming ecosystem design
The numbers looked fine. Something was wrong.
I was brought in to fix the training. Before touching a single piece of content, I spent three weeks asking a different question: what do we actually know about why customers leave?
91% completion. 4.2/5 satisfaction. 68% churn. That's not a content problem — that's a mystery. And I was very, very curious.
High completion. High confidence. High churn.
The training data and the business data were telling completely different stories. That disconnect became the central question of the entire project.
Learning is working
91% module completion. 4.2/5 confidence score. Every learning metric pointed up.
Customers are leaving
68% churn within 18 months. Feature adoption dropping month on month. Support tickets rising.
Curiosity led to an uncomfortable truth.
I interviewed Product (n=4) and Customer Success (n=6) separately, asking the same question: “In your own words, why do customers leave?” I didn’t share the answers between teams. The divergence was striking.
Both teams were working hard. Neither was wrong exactly. They'd just never been in the same room — and I started thinking that gap might be the problem itself.
Upload your image here
Recommended: the “Product team said vs Customer Success said” two-column visual
Suggested size: 760 × 380px
The confidence trap
Users scored 4.2/5 confidence — but had only ever practised in clean demo environments. When they hit their own messy real-world data in production, they froze. They didn’t ask for help. They quietly stopped using the feature. The decision to churn was made weeks before any renewal conversation.
The translation gap
CS heard “it’s too complicated” and logged it as product feedback. Product simplified the UI. But customers meant: “I don’t know which of the 12 ways to do this fits my use case.” A decision-support problem dressed as a UI complaint — invisible because no one was translating between teams.
Before we design anything, we need to be in the same room.
No eLearning was going to fix a problem two teams couldn’t agree on. I recommended a full-day facilitated workshop before a single piece of content was scoped.
I had to pitch this internally. "You want to run a day-long session before building anything?" Yes. Designing without alignment isn't faster — it's just confidently wrong.
facilitated workshop — "See What We're Missing"
Full day
8 hrs
12 people
Product + CS + L&D
No slides
data only
09:00 Data walk — Both teams review the same LMS + usage data side by side. No commentary. Just observe.
10:30 Customer voice wall — Verbatim exit quotes, printed and anonymised. Both teams sort and theme.
12:00 Assumption clash — Each team presents their churn theory. Facilitator maps where they diverge.
13:30 Journey map — together — First 90 days on the wall, drawn jointly. Emotional state included.
15:30 Root cause convergence — 3 highest-impact intervention points. Vote together.
16:30 Solution space — Open ideation. Structural changes only — no new meetings.
The friction tax
- An 8-minute task was taking 40+ mins — customers were depleted before exploring anything else.
- Churn wasn’t frustration. It was exhaustion.
Nobody knew how to have the hard conversation
- CS Managers were avoiding capability conversations — they feared it would sound like blame.
- No one had ever taught them how to name a knowledge gap kindly.
A shared tagging taxonomy
- Shared tags across CS and Product’s existing ticketing system — same words, same definitions.
- No new meetings. No new tools. Just a shared language.
It was never just a training problem. Here's what it actually wa.
The workshop, interviews, and usage data together produced one clear picture. This was four problems layered on top of each other — and only one was addressable with learning content.
False confidence — high training scores, zero transfer to real-world data. The gap between demo environments and production was never bridged.
The friction tax — a core workflow taking 5× longer than it should, leaving users too depleted to explore the product's value.
The translation gap — two teams logging the same problem in different languages, both solving the wrong thing, neither knowing the other's theory existed.
The human gap — CS Managers lacking the language to have capability conversations at renewal without it feeling like blame. Unspoken. Unaddressed. Expensive.
Not a course. An ecosystem.
Once we had the journey map, I kept asking: where does a standalone eLearning actually show up in this picture? The honest answer: only at the very start. The rest of the journey was completely unaddressed.
The eLearning was about 20% of the solution. The rest was systems, process, and one very human conversation that nobody had planned for.
The difficult conversation workshop
Tooltips and decision prompts at the exact moments users freeze in production. Triggered by behaviour, not a timer.
90-second decision job aids
“Which path is right for my data?” — short guides for the 3 most common decision points. In the product, not the LMS.
Real-data scenario practice
Rebuilt around anonymised customer data profiles — not clean demo environments. Branching, Storyline 360.
Manager conversation guide
Signals to help managers spot the Confidence Trap early — what to watch for at weeks 2, 4, and 8.
Shared tagging taxonomy
Same words, same definitions across CS and Product’s ticketing system. No new meetings. Just a shared language.
The difficult conversation workshop
A short facilitated session for CS Managers: how to name a knowledge gap without blame. Not a course. A conversation that changed the renewal dynamic.
Three versions. Three theories about what would work.
The scenario module went through three distinct iterations, each tested with a cohort of 8–12 newly onboarded customers before scaling. Each failure taught me something the data alone couldn’t.
V1 made things worse. I'm putting that up front. Bad data from a test is still good data — it told me exactly what the real problem was.
More content, better structure
Extended the module from 20 to 45 minutes. Added two new workflow sections and a knowledge check every 10 minutes.
Branching scenario with demo data
Built a branching scenario in Articulate Storyline. Three diverging paths based on data type. Users loved the format.
Real-data profiles + contextual in-app prompts
Worked with Product to create anonymised profiles from actual customer data types. Combined with contextual prompts at the exact in-product decision points. The scenario and the product now spoke the same language.
Run the workshop in week one
The misalignment between Product and CS was obvious from the first interview. I’d propose it on day one, not after three weeks of discovery.
Instrument V1 from the start
Completion rates told me it wasn’t working. xAPI click-path data would have told me where and why — much faster.
Involve CS in content review
Their customer language would have made the V2 scenario feel more real from the first draft.
Build the feedback loop first
What I'd do differently
The numbers looked fine. Something was wrong.
My first instinct when I got this brief? Don't touch the content yet. 91% completion but 68% churn? That's not a content problem. That's a mystery. And I was very, very curious.
I was brought in with a clear mandate: “customers aren’t using the product — fix the training.” Before writing a single learning objective, I spent three weeks asking a different question: what do we actually know about why customers leave?
I reviewed LMS analytics, NPS data, exit surveys, and six months of support ticket logs. I interviewed five Customer Success Managers, sat in on two renewal calls, and watched three new users attempt their first real workflow in production. That last part was where things got interesting.
The numbers looked fine. Something was wrong.
Before we design anything, we need to be in the same room.
It was never just a training problem. Here's what it actually was.
Not a course. An ecosystem - and one hard conversation.
Three versions. Three theories about what would work.
Take this prototype for a spin.
The V2 branching scenario – the iteration that proved the format before I got the data right – is built and available. It gives you a feel for the decision architecture, the branching logic, and how I structure scenario-based learning for a technically complex product.
12 minutes – branching scenario. 3 diverging paths Articulate Storyline 360