Written by: Michał Sobieraj, Operations Manager, Digital Colliers
The SaaStr numbers from this quarter are worth staring at for a minute. Shopify grew 34%. Toast grew 23%. Samsara grew 30%. Meanwhile a lot of pure-play software names got beaten down hard. These three don't share a category or a customer profile. What they share is a data position that generic SaaS can't replicate.
Shopify sees every checkout, every abandoned cart, every fulfilment event across millions of merchants. Toast sees every plate that leaves a kitchen. Samsara sees every truck, every driver hour, every cold-chain deviation. That data is welded to the physical world. You can't scrape it, you can't synthesise it, and a competitor with a better UI can't take it from you.
The market is pricing data gravity, not features
For the last two years the pitch deck template said "AI-powered" and the multiple followed. That trade is over. Investors have watched roughly 95% of enterprise AI projects fail to reach production or ROI, and they've drawn the obvious conclusion. Features are cheap now. Any competent team can wire GPT into a workflow in a weekend.
What's not cheap is the pipe carrying real operational signal. Order-level margin. Return reason codes tied to SKU and customer cohort. Delivery exceptions matched to carrier and postcode. That's the layer the public markets are willing to pay for, because that's the layer AI actually needs to do useful work.
What this means if you sell physical goods
UK eCommerce grew about 3% in 2024 over 2023. Single-digit growth is the new baseline, and it's not going back. In that environment the operators who win are the ones who know, per SKU and per channel, what actually makes money. Most brands don't.
A rough benchmark from Profitero: at a typical multi-channel brand, around 30% of SKUs lose money per order once you subtract returns and ad spend. That's a stunning number if you sit with it. A third of what you ship is destroying enterprise value, and if your P&L rolls up at the category level you literally cannot see it.
The reason you can't see it is almost always that the data lives in five systems that don't talk. Shopify has the order. The 3PL has the return. Meta has the ad cost. Klaviyo has the acquisition source. NetSuite has the COGS. Nobody has the joined view.
Data you should own, data you can rent
Here's the split that seems to separate operators pulling ahead from ones that are stuck.
Own:
- Order-level economics with true landed cost, ad-attributed CAC, and returns netted in
- Return reason codes at SKU and variant granularity, not "didn't fit"
- Customer LTV cohorts broken by acquisition channel and first-product
- Inventory position and sell-through by location, updated hourly not weekly
Rent (use vendor tools, don't rebuild):
- Ad platform bidding
- Email and SMS orchestration
- Reviews and UGC collection
- Tax and compliance calculation
The pattern is that the owned data is the stuff a vendor can't give you back if you switch. The rented layer is commodity, and paying a SaaS bill for it is fine.
The left-behind risk is quieter than it looks
Customer acquisition cost across DTC brands has risen roughly 40% since 2023, and Meta CPMs kept climbing through 2024 and 2025. Online return rates sit around 19 to 20% of gross sales, higher in apparel. If those three numbers are all trending against you and you can't see per-order margin, you're not going to feel the problem as a crisis. You'll feel it as a slow squeeze on cash conversion that nobody can quite explain in the Monday meeting.
The brands that are going to look untouchable in 2027 are the ones building this operational data layer now, quietly, while everyone else is buying another AI copilot. They're not doing anything glamorous. They're doing the plumbing. A warehouse that joins order, return, ad, and cost data. Dashboards their merchandising team actually opens. Alerts when a SKU tips into loss-making territory.
When an AI project finally does land inside one of these businesses, it works, because the data underneath it is real. That's the actual lesson from Shopify, Toast, and Samsara. The AI story only works when the operational story already does.