A B2B growth idea that made sense on paper, failed on contact with the catalogue data — and the rescope that captured the demand anyway.
Digikala Business is the B2B arm of MENA's largest eCommerce platform (40M MAU). I owned the backlog and roadmap for seller tools, product catalogue and inventory management, working across commercial and engineering teams.
Business customers were buying supplies through the consumer site instead of the B2B platform — so we were missing B2B margin, B2B account data, and the chance to build a real relationship with business buyers, on purchases we were already winning.
If we put a B2B purchase call-to-action directly on the consumer product pages business customers were already buying from, we could catch that demand at the moment of intent and move those buyers onto the B2B platform.
Before committing engineering time, I checked how many of the products business customers bought on the consumer site were actually live on the B2B platform.
Only 7% were. Shipping the CTA as designed would have sent 93% of the people who clicked it to an empty result, on a page they'd been happily buying from a second earlier.
I stopped the page-level rollout. A CTA that fails 93% of the time doesn't just underperform — it teaches business buyers that the B2B platform doesn't have what they need. The catalogue gap was the real problem; the CTA was treating a symptom.
The hypothesis was reasonable and the validation still killed it — which is the point. The cheap version of this lesson cost an afternoon of catalogue analysis. The expensive version would have been a sprint of engineering and a trust problem with our most valuable buyers.
Predictive maintenance for rail operators — where a missed fault is a safety incident and a false alarm is a cancelled train.
PANTOhealth builds AI-driven predictive maintenance for overhead rail lines, sold to government and enterprise operators across EU and Asian markets. I owned the product roadmap end-to-end, working with engineering across AI, simulation and hardware workstreams.
Operators maintain overhead lines on fixed inspection schedules — expensive, and still reactive when something degrades between inspections. The value of prediction is obvious. The obstacle is trust: an operator won't act on a model's output unless they can see why it flagged what it flagged.
The predictive-maintenance dashboard wasn't meeting users' needs. Instead of iterating on guesses, I documented a specific hypothesis about why, and validated it before any engineering time was spent:
Three groups building very different things, with no shared delivery process. I introduced Scrum ceremonies and PI planning, and built a QA workflow with the engineers — a test-case repository, a release checklist and a regression cadence.
I led the product showcase at InnoTrans 2024 in Berlin — the world's largest rail-technology trade fair — coordinating the live demo suite, collateral and client engagement on the floor.
The product on the EIT Urban Mobility marketplace ↗
Working in a marketing team at an Australian-listed consumer tech company, as the standing bridge to Product.
Spacetalk (Australian Securities Exchange: SPA) sells kids' smartwatches, a family-safety app and MVNO mobile plans through national retail. I was the Marketing Project Manager, working day-to-day with Product on feature launches, onboarding flows and customer-journey work tied to app releases.
After a new app release, the same questions kept appearing in social media comments — users asking how to do things the app already did.
My hypothesis: new users weren't failing at the app, they were failing to understand it. The features existed; nothing taught them. If that was true, a dedicated onboarding surface should reduce the repetitive questions without changing the product at all.
I led the build of an onboarding page in Sanity CMS from concept to launch — video walkthroughs plus written guidance, aligned with product messaging.
A behavioural measure, not a vanity one: the same users, the same product, asking fewer questions because the answers were finally findable.
