Revenue Growth Can Conceal Weak Economics

City Shift Finance: Across your multi-brand ecommerce operation, what has caused revenue to remain healthy while margin or unit economics were weakening underneath it?

David Caruso: In a multi-brand operation, the honest answer is that healthy revenue is one number hiding hundreds of smaller ones. Group revenue can look fine while individual lines quietly rot underneath, and the causes are rarely dramatic: freight creep, a currency moving against you in two or three of your 32 markets, the product mix drifting toward lines that sell easily but carry thinner margin, and small discounts that become permanent. The protection isn't a monthly review of the topline, it's per-product economics set before launch. Every product we list carries its own ceiling on what we'll spend to win an order, worked out from its price before a single ad runs. When a line's real costs push past that ceiling, the line has a problem no matter what group revenue says. And I'll admit the bias openly: we've chosen to grow slower on purpose, because pace is where weak economics go to hide.

City Shift Finance: When a product or category underperforms, what evidence tells you where the problem is?

David Caruso: The evidence arrives in a sequence, and the order matters. First: is there traffic? If nobody visits the product page, it's a demand or visibility problem, and no amount of price-cutting fixes it. Second: if traffic is there but conversion sits below our bar (we cut any product that can't hold 2%), the problem lives on the page or in the offer: price, proof, the delivery promise, or photos that don't match what the buyer hoped. Third: if the product converts but the economics still don't work, the leak is in acquisition or fulfilment cost, and we check those against the ceiling the product was given at launch. Reviews arbitrate the last case: a product that sells and then generates returns and complaints isn't a marketing problem at all, it's a product problem, and we stop testing marketing fixes on it. The discipline is refusing to let the explanations blur: each cause has its own evidence, and its own fix.

City Shift Finance: Which decisions have had the greatest effect on inventory cash, markdown exposure, and fulfilment economics, and where can sales growth conceal deterioration in cash generation?

David Caruso: The biggest single decision is how much stock a new product earns before it's proven. We test around 50 new products a month, and a product sees a small first order of inventory until it holds our conversion bar. Most don't survive, and because the failed ones die small, markdown exposure stays small too. The cash discipline isn't in the markdown, it's in the moment before the reorder: the reorder is where inventory cash gets committed, so that's where the evidence has to be strongest. On fulfilment, holding our own warehouse and shipping everything through one carrier gave us predictable per-parcel economics instead of a blended mystery. Where growth conceals deterioration: any time sales growth is funded by inventory growth. The topline compounds, the warehouse fills, and cash quietly converts itself into cartons. If revenue is rising while stock turns are slowing, the growth is eating the cash it appears to be generating.

City Shift Finance: As AI and automation absorb more operating work, what evidence tells you a lower labour requirement is a durable productivity gain rather than risk transferred elsewhere?

David Caruso: The test I use: when the machine takes the work, who catches the mistake now? If the answer is nobody, the labour saving isn't productivity, it's risk quietly transferred to your customers. We learned this concretely. In the second week of automating our inbox triage, the system began throwing genuine sales enquiries out with the spam. The hours saved were real, and so was the risk that had moved onto customers who never got an answer. The fix kept the automation and rebuilt the boundary instead: the machine sorts and drafts, a person approves everything that leaves. Durable gains have a particular signature: volume grows, headcount doesn't, and the mistakes a human would have caught still get caught, because a human still sits at the point of consequence. There's an honest cost ledger too: replacing rented software with your own turns subscription fees into maintenance responsibility. We're six months into rebuilding the whole operation around AI, the early evidence is good, and I treat every saving as unproven until it has survived a failure.