Nearly four years on a retail floor taught me the part that is hard to shortcut: what earns trust, what makes a recommendation credible, and what brings someone back a second time. Most of it is listening longer than feels comfortable.
Then I built a product on my own, which meant finding external partners, negotiating with them, managing them across time zones and correcting my own assumptions when the evidence disagreed with me.
VESSEL — developed independently, manufactured through partners in China
MECCA Double Bay is high volume and highly product-led. You get a few minutes with someone who may not know what they want and may not want to say so. The job is to work out what they are actually asking for.
I built a loyal client base across skincare, makeup, fragrance and home. Not by knowing more products than anyone else, but by asking about routine, budget and what had failed for them before recommending anything. Customers came back and asked for me specifically. That is the only retention metric I had access to, and it is the one I trust.
Over $10,000, in one consultation. It was not a pitch. It was a long conversation about what the person was trying to achieve, and a set of recommendations they believed because I had already told them what not to buy.
I was asked to train new team members on product knowledge, consultation technique and customer experience. This is closer to teaching someone an unfamiliar system than walking them through a manual — the range moves constantly, so what you are really handing over is a way of reading a customer and a way of finding the answer when you do not have it.
I helped plan and run in-store brand activations and campaigns, representing MECCA and partner brands directly to customers through live selling and public engagement. Standing in front of a room and holding attention is a separate skill from selling one to one, and I had to learn it in public.
My customer-care approach was recognised internally by MECCA founder Jo Horgan, and I was selected for founder-led entrepreneurial workshops focused on innovation and customer experience. Worth being precise: it was recognition for how customers were treated, not a sales award.
VESSEL is a premium travel skincare system I built independently, from a problem I heard repeatedly on the floor. The part relevant here is not the product. It is what happens when your entire supply chain is a group of people you have never met, in another country, who have no reason to prioritise you.
I had no experience in manufacturing, technical specification or supplier systems when I started. I taught myself supplier communication, specification writing and AI-assisted workflows from zero, fast enough to run development independently. I mention this because it is the honest version: I was not qualified for this at the start and I got there by doing it badly first.
I sourced and compared manufacturers in China, evaluated them on responsiveness and precision rather than quoted price alone, negotiated requirements and MOQs, and briefed samples. Choosing a partner is mostly a judgement about how they behave when something goes wrong, and you have to form that view before anything has gone wrong.
Every question has a half-day latency and no shared context. I learned to write briefs that survive being read by someone tired, at the end of their day, in a second language — one decision per message, ambiguity removed before sending, and the cost of a misread stated explicitly.
Original customer research and a 10-competitor audit came before I decided what the product was. I wanted a reason to believe the demand existed that was not just my own experience behind a counter.
The first physical sample came back and did not match the specification. The easy conclusion — the one I wanted — was manufacturer error.
I measured it against the spec properly instead. Some deviations were genuine non-conformance. Others were normal manufacturing variation that my specification had never told them was unacceptable, because I had written every dimension as if all of them were critical. That was my failure, not theirs.
I rebuilt the methodology to separate function-critical dimensions from tolerable variation, and reissued. It changed the relationship: the supplier was being held to a standard that was finally legible.
Before committing to a price, I modelled the unit economics and then tried to break them. Every figure here is planning work on quoted and researched inputs, not booked revenue.
My first version assumed a landed cost I had not re-quoted and freight I had guessed. It made the business look better than it was. I re-listed every input against an actual quote, rebuilt it, and the margin came down. The useful output was not the number — it was knowing which single input the whole thing hangs on, and therefore what to negotiate hardest.
To be clear about what this is: founder-modelled commercial reasoning built in a spreadsheet. It is not formal data analytics training, and I do not write SQL or work in analytics tooling.
Two working prototypes, built to find out whether an idea held up. Listed briefly because they evidence speed and independent execution, not because either needs a case study here.
A working marketplace prototype designed around real-world social connection: venue check-ins, mutual intent, trust and safety controls, and behavioural incentives.
A working AI prototype that turns spoken conversations into structured, reusable outputs.
Paying attention to behaviour rather than to what people say they will do. On the floor, the gap between the two is visible within about a minute, and it is where most of the useful information is.
Earning trust before offering expertise. Telling someone the cheaper option is the right one, or that we do not have what they need, is what makes everything said afterwards worth listening to. It works the same way with suppliers.
Holding three things at once with no playbook: what the customer needs, what an external partner can realistically deliver, and what the commercial reality allows. Nobody hands you the answer to that, and it usually cannot be optimised for all three. Most of the work is deciding which one gives.
I am comfortable saying I do not know yet, and then going and finding out. Nearly everything I can do now, I could not do two years ago.