The arithmetic nobody runs
Take a $15M DTC beauty brand, a real composite, not a hypothetical. Here's what the analytics stack actually costs per year:
Direct licence costs. The BI platform, the customer data platform, the attribution tool, the email analytics layer, the inventory planning module. Individually, each one looks reasonable. Collectively, it's $80K–$140K per year. At 15M revenue, that's approaching 1% of top line before anyone has made a single better decision because of the data.
Integration maintenance. Every platform connects to every other platform, in theory. In practice, someone on your team (or an agency) spends 10–15 hours per month keeping the integrations running, debugging data mismatches, reconciling numbers between systems that should agree but don't. That's another $30K–$50K in salary time.
Time-to-insight lag. The question gets asked on Monday. The analyst pulls data from two platforms on Tuesday. The numbers don't match. Wednesday is spent reconciling. Thursday the insight reaches the commercial team. Friday it gets discussed. The following Monday, someone decides it needs a different cut. Repeat. The average time from question to decision in a mid-market DTC brand with a standard SaaS stack: 2–3 weeks. In retail, where the window for action is often measured in days, that latency isn't inconvenient; it's expensive.
Decision meetings. Every tool generates reports. Reports generate meetings. Meetings generate requests for more reports. A mid-market brand with five analytics tools has at least three recurring meetings per week dedicated to interpreting what the tools are saying. That's senior commercial time, the most valuable and scarce resource in the business, consumed not by deciding but by preparing to decide.
Add it up: $150K–$250K per year in total cost of ownership. Against a $15M top line, that's 1–1.7% of revenue. Against a typical DTC net margin of 8–12%, it's 10–15% of your profit, paid to a collection of tools that were supposed to create clarity but instead created a second job.
What you get for the money
Here's the uncomfortable question: what decisions changed because of these tools?
In most $10M–$30M DTC brands, the answer is: fewer than you'd expect. The customer data platform reveals segments that the marketing team already intuited. The attribution tool provides numbers that nobody fully trusts. The BI platform generates 130 pre-built reports, of which the team regularly consults maybe eight.
The tools aren't bad. They're generic. They were designed for "any business" and sold to yours with a pitch deck that showed your logo on a template. They can tell you what happened. They struggle to tell you why it happened in the specific context of your business. And they cannot tell you what to do about it.
This is the fundamental problem with bundled SaaS for mid-market retail: the tool's architecture assumes that your business looks like every other business. Past $10M in revenue, that assumption starts breaking down. Past $20M, it's actively harmful.
The bespoke threshold
There's a revenue range (roughly $10M to $50M) where the economics of analytics flip. Below $10M, the SaaS stack makes sense. You need basic infrastructure, standard reporting, the fundamentals. The cost is proportional to the value and the business isn't complex enough to need precision.
Above $10M, the complexity of the business outpaces what generic tools can handle. Channel interactions become non-linear. Customer economics vary by acquisition source in ways that standard cohort analysis can't capture. Margin isn't a single number; it's a function of channel, geography, return rate, fulfilment method, and timing. The SaaS stack can show you pieces of this picture. It cannot show you the picture.
This is where bespoke engineering starts to make economic sense. Not bespoke in the sense of building everything from scratch; that's wasteful. Bespoke in the sense of engineering the specific analytical tool, model, or system that answers the specific commercial question your business is actually facing.
What the alternative looks like
Instead of five platforms and three weekly meetings, imagine this: a single, purpose-built analytical model that answers the three most important commercial questions your business faces right now. Built on your data. Validated against your commercial reality. Operated by someone who understands retail deeply enough to know which questions matter.
The cost is comparable to the annual SaaS stack. The difference is that instead of 130 reports and eight dashboards, you get three answers that directly inform three decisions. And those decisions move.
This isn't an argument against software. It's an argument against paying a margin tax for generality when your business has passed the point where generality is useful.
The question worth asking
Next time a SaaS vendor pitches you an analytics platform, ask them one question: "Which specific commercial decision will my team make differently because of this tool, and how much is that decision worth?"
If the answer is vague, the tool is generic. And if the tool is generic, you're paying the tax.
Mid-market DTC brands past $10M deserve better than a collection of logins and a hope that someone will find the insight. They deserve something built for the specific shape of their business. That's what we mean by engineering the solution instead of buying the platform.