Composable by Default: SaaS for Commodity, Custom for Competitive Advantage

Thirty-five percent of teams have already replaced the functionality of at least one SaaS tool with something they built themselves, and 78% expect to build more in 2026 — per Retool's 2026 Build vs. Buy report (opens in new tab). The categories leading the wave are not exotic: workflow automations, internal admin tools, BI, CRMs.

We run a custom software shop, and "build vs. buy" is still the wrong framing — binary, one-shot. Every team we work with runs both. The useful question is: which functions should be SaaS, which should be custom, and where does the seam live?

The old framing is dead because the economics flipped

First, SaaS got expensive fast. Vertice's SaaS Inflation Index (opens in new tab) clocks pricing growth at more than four times the US CPI — peaking at 14% in May 2025 and still running 11–12% annualized into 2026, with cost per employee around $9,100 by year-end 2025.

Second, build got cheaper. Retool finds 51% have shipped production software using AI tooling, and about half save six or more hours a week. Internal tools that needed a five-person squad can now be carried by one senior engineer with good taste and a CI pipeline.

Meanwhile, the regulatory ground shifted. The EU Data Act (opens in new tab), in force since 12 September 2025, forces SaaS providers to allow termination with two months' notice, forbids exit fees, and requires data exports in a "structured, commonly used, machine-readable format." The default direction of gravity is finally away from lock-in.

What "commodity" actually means for software

A function is commodity when three things are true: customers would not notice if you switched providers tomorrow; the logic is industry-standard; and the rate of change matches the vendor's release cadence. Email, identity, payroll, sales-tax, e-signature, log aggregation, cloud infrastructure — commodity for 99% of companies.

The test we give clients: if the function were a truck, would you lease or buy? If every logistics company drives the same truck for the same job, you lease. If your truck has a hydraulic arm nobody else has, you buy. Most CRMs, helpdesks, and accounting systems are trucks. Order orchestration, pricing logic, compliance workflows — those are hydraulic arms.

What competitive advantage looks like in code

The opposite of commodity is anything that encodes a decision you made differently from competitors. A German Maschinenbau client of ours prices spare parts using machine vintage, regional service density, framework-contract discounts, and take-back credits. No SaaS CPQ fits without extensive customization — and the moment you customize SaaS you lose its main advantage.

The heuristic: if you can describe the workflow to an SAP consultant in a sentence, it's off-the-shelf. If it takes three whiteboards, it's yours to own. Gartner's composability research found high-composability companies projected ~7.7% revenue growth versus 3.4% for low-composability peers; a related forecast (opens in new tab) projected 30% higher revenue for composable finance organizations.

A five-point framework for drawing the line

Score each function on five axes. None is decisive alone; the pattern matters.

1. Differentiation

Does this function change how a customer experiences us versus a competitor? A generic "send an invoice" flow does not. A payment plan that adapts to seasonal cash flow of German landscaping firms does. If no, SaaS. If yes, custom — or custom orchestration on SaaS primitives.

2. Data ownership and gravity

The EU Data Act helps on paper; in practice, exporting ten years of nested CRM relationships is still a two-quarter project. For data feeding analytics, AI models, or regulatory reporting, default to custom storage with SaaS as a thin integration layer. Veeam's vendor lock-in analysis (opens in new tab) notes egress fees alone can cost five figures before engineering starts.

3. Integration depth

One or two systems, SaaS usually wins. Five or more — ERP, three product lines, a warehouse system, two payment processors, a compliance database — the integration is the system, and you'll want to own the orchestrator.

4. Change velocity

If your workflow changes faster than your vendor ships, you are the vendor's feature request queue, and you will lose.

5. Regulatory and residency fit

For German and EU Mittelstand, this is the tie-breaker. NIS2, Data Act, AI Act, GoBD all push toward systems you can audit end-to-end. The Bitkom Digitalisierung der Wirtschaft 2025 (opens in new tab) report found 53% of German companies report problems managing digitalization itself — much of that is compliance overhead.

Run those five for every function. The resulting heatmap is your composable architecture. Rarely pretty, almost always defensible.

Where SaaS quietly costs more than the sticker

The subscription is the smallest line item in TCO. Gartner's research (opens in new tab) holds that personnel and maintenance dwarf licence fees, and benchmarks put annual maintenance at 15–20% of build cost (opens in new tab) for custom. Three recurring hidden SaaS costs: integration drift (vendor APIs change ~every 18 months and you re-pay across the stack); workflow drift (by year three half your team has a spreadsheet doing the "real" work — Retool found 60% of builders admit to shadow IT, 64% of them senior managers); and exit cost (3–6 months of engineering even with the Data Act).

Worked example: BookMe vs. Calendly

If you are a consultant, small clinic, or recruiter where the booking flow is pick-a-person, pick-a-slot, send-an-invite — use Calendly. Differentiation zero, integration depth low, change velocity low. We tell people this and we mean it.

If you are a regional physiotherapy group with forty therapists, three locations, a state-reimbursement schedule limiting sessions per patient per quarter, a clinical-priority waiting list, and a monthly Kassenärztliche Vereinigung CSV obligation — Calendly will hurt you. BookMe competes with the spreadsheet plus three coordinators plus the compliance risk. Break-even is usually 18–24 months, tracking German cost benchmarks (opens in new tab) for Mittelstand projects. The same logic applies across our line: ZahlFlow above Stripe, Commersio for B2B catalog ordering Shopify can't model, Ordrino for order orchestration when the POS is one of ten systems involved.

When we tell clients to stay on SaaS

The workflow is genuinely generic. A 30-person professional services firm asking for a custom CRM is better served by HubSpot or Pipedrive. We cannot beat their feature cadence.

The team cannot carry maintenance. A €120K build wants ~€20K/year to stay healthy. No engineering leadership, no CI, no plan for who owns the codebase in year three? Buy SaaS and hire a good integrator.

The regulatory surface is already covered. If every competitor uses the same two vendors, their compliance is a moat working in your favor. Payment card compliance is canonical — we'll build your checkout logic, but we will not operate the card vault.

The timeline is too short. Custom work is 9–15 months whiteboard to production. Six-month regulatory deadline? SaaS-plus-integration now, custom module as Phase 2.

A software shop that builds everything every client asks for is eventually a graveyard of half-loved systems. The right answer is almost always composable — commodity on SaaS rails, differentiation in code we own together, a clean seam, and a plan for what happens when either side shifts. That plan is the deliverable. Not the code.