Choosing a pricing model is choosing who bears uncertainty. Fixed price shifts risk to the vendor—if scope is knowable. Time and materials (T&M) shifts risk to the buyer—if governance is strong. Retainers buy capacity and responsiveness—if priorities are clear. At Smoother Development, we match contract type to discovery maturity, roadmap volatility, and stakeholder alignment. This guide gives CTOs and VPs a practical map: when each model works, typical fee structures in Europe, and failure modes to avoid.
Procurement teams often default to fixed price because spreadsheets love certainty. Engineering leaders know certainty is sometimes a mirage. The art is structuring contracts so both sides share truth early—discovery, prototypes, and written assumptions.
Fixed price (FP): certainty with guardrails
What it is: a defined scope, timeline, and price—often milestone-based payments tied to deliverables or dates.
Best for: well-specified modules after discovery—migration tranches, integration packages, MVP with frozen must-haves, or compliance features with objective acceptance tests.
Typical structure: 30/40/30 payments (kickoff / mid / acceptance) or monthly milestones against a statement of work (SOW) with exclusions spelled out.
Pros: predictable cash outflow; vendor incentive to finish; easier board communication.
Cons: change orders for anything outside SOW—can erode trust if discovery was shallow; vendor may pad risk premium (10–25% is common).
When it fails: ambiguous product goals, first-time founders without decision-makers, or emerging tech (some AI features) where unknown unknowns dominate.
Executive tip: insist on a shared backlog with Definition of Done, test strategy, and explicit assumptions (APIs available, design assets delivered, UAT participants named).
Time and materials (T&M): flexibility with discipline
What it is: you pay for hours at agreed role rates, usually monthly, with weekly or biweekly burn reviews.
Best for: 0→1 products, R&D, AI experiments, legacy modernization with surprises, or staff augmentation where your PM owns prioritization.
Typical rates (EU, illustrative): staff augmentation often starts from €45/hour (mid-level) and €60/hour (senior)—roughly ~€7,500–10,000/month full-time at ~168 hours; broader T&M for senior engineer EUR 95–140/hour depending on country and firm; principal/architect EUR 120–170/hour; PM/UX EUR 80–120/hour—always confirm blended vs role rates.
Pros: fast pivots; honest accounting of complexity; ideal when learning is the deliverable.
Cons: requires client-side governance—without it, budgets drift; finance may perceive open-ended risk.
When it fails: no empowered product owner, no sprint goals, or no code review/QA standards—then T&M becomes expensive motion.
Executive tip: pair T&M with not-to-exceed caps per phase, monthly budget alarms, and demo-based decisions. You keep flexibility without abdicating control.
Retainer: buy capacity, not tickets
What it is: a monthly fee for a reserved team slice or priority access—often 40–160 hours/month depending on scale.
Best for: continuous improvement, SLA-driven maintenance, fractional platform team, or post-launch iteration when roadmap is steady but not fully predictable.
Pros: predictable partner availability; faster response; strategic continuity—engineers learn your domain.
Cons: can feel expensive if utilization is low; needs mutual trust—governance still required.
Typical economics: monthly retainers for two senior engineers worth of capacity might land EUR 15,000–24,000/month in EU markets (in line with ~€10,000/month per full-time senior equivalent at from €60/hour)—before pass-through cloud or licenses—highly dependent on scope and seniority.
When it fails: unclear prioritization—retainer becomes busywork; or mismatch between promised capacity and actual staffing quality.
Milestone + hybrid: the pragmatic default
Many strong engagements blend:
- Discovery as T&M or small fixed package (EUR 8,000–25,000).
- Build as fixed per milestone with change order path.
- Hypercare post-launch as retainer or T&M with SLA.
This mirrors how risk actually changes over time—early uncertainty, later execution.
AI-era caveat: why pure fixed price is harder
Generative AI features have non-deterministic behavior and evaluation dependencies. Vendors may refuse fixed price without lab time, or will price in large buffers. Expect separate line items for:
- Baseline integration (deterministic).
- Model evaluation and prompt iteration (probabilistic).
- Ongoing monitoring and dataset updates (operational).
If a vendor offers cheap fixed AI without evaluation plans, question the model.
Contract mechanics that matter (Europe)
- IP assignment timing—upon payment is standard; work-for-hire clarity matters.
- Warranty window (30–90 days) and severity definitions for bugs.
- Liability caps (often fees paid in last 12 months)—understand what is excluded (often indirect damages).
- Data processing terms for GDPR—especially if vendor accesses production data.
- Exit assistance: repo transfer, docs, handover hours.
Budget legal review EUR 5,000–20,000 for meaningful first contracts—cheap relative to disputes.
How to choose: decision matrix
Choose fixed when scope is documented and stable, and you want budget certainty.
Choose T&M when learning dominates and you have strong product leadership.
Choose retainer when you need continuous partnership and predictable access—post-PMF scaling, platform evolution, or SLA coverage.
Red flags across all models
Vague SOWs with pretty decks. No references in your domain. Unrealistic timelines. No access to senior engineers pre-sale. Fixed price without assumptions list. T&M without weekly demos.
Also watch for misaligned incentives: vendors who maximize hours on T&M without outcomes, or fixed price shops that cut quality to protect margin. Adult contracts include quality gates: CI, coverage thresholds, security scans, and explicit defect SLAs for warranty periods.
Incentives: aligning vendor and client success
Outcome-aligned clauses are rare but valuable when measurable: bonus payments tied to latency SLOs, conversion uplift on a specific funnel step, or support ticket deflection rates—provided baselines and measurement ownership are clear. Beware metrics outside vendor control; pair bonuses with transparent analytics access.
Where outcomes are hard to isolate, prefer milestone payments tied to demos with acceptance tests—proxy outcomes that still force integration truth.
Change management: how scope evolves without chaos
Even with fixed price, reality changes—new APIs, security findings, customer feedback. Establish a single change control path: impact statement (time/cost/risk), approver, and updated acceptance criteria. Without this, fixed price becomes adversarial, and T&M becomes opaque.
For retainers, maintain a rolling three-sprint plan with WIP limits. Capacity without prioritization is how retainers underperform.
CFO-friendly reporting
Finance teams prefer predictable cash curves. Map milestones to invoice dates, and T&M to monthly caps with variance explanations tied to deliverables—not “engineering happened.” Outcome-based pricing is rare in custom software for good reason—outcomes depend on client inputs—but milestone payments approximate value delivery when written well.
Benchmarks: what “good” looks like
Healthy T&M engagements show stable velocity after 2–3 sprints, declining defect rates, and demo cadence every one to two weeks. Fixed projects show green CI, staging demos, and early UAT involvement—not a big-bang reveal in week twelve.
If you never see staging until the end, your pricing model is irrelevant—the process is broken.
Closing guidance
Pick the model that matches where uncertainty lives today—then renegotiate as you learn. The best vendors won’t optimize for your downside forever on fixed price, and won’t hide behind hours on T&M if outcomes are what you purchased. Transparency and shared metrics beat contract theology.
Summary table for executives
| Model | Predictability | Flexibility | Best owner on client side |
|---|---|---|---|
| Fixed price | High | Low | Strong PM + stable scope |
| T&M | Medium | High | Empowered product + governance |
| Retainer | Medium–high | Medium | Clear roadmap + backlog hygiene |
Use the table in vendor discussions to align on governance, not only price.
Appendix: what to ask vendors in the first call
Ask for two references with similar complexity, a sample statement of work (redacted), and how they handle defects after the warranty window. Ask who owns architecture decisions on day one and day ninety, and how they measure quality beyond hours logged.
If answers are vague, price is not your biggest risk—delivery is.
Suggested governance cadence (lightweight)
Weekly: demo, burn vs budget, risk log review. Biweekly: backlog refinement and scope change queue. Monthly: steering with KPIs tied to business outcomes (activation, revenue enablement, incident rate). This cadence works for both T&M and fixed-price programs—the contract changes who signs change orders, not whether you need adult supervision.
For retainers, add a quarterly capacity planning review: expected roadmap themes, upcoming compliance deadlines, and hiring gaps on your side that could starve the backlog. Retainers fail when both parties assume the other side will magically generate perfect tickets—proactive roadmap hygiene is the client’s job, not a vendor deliverable you can outsource without structured input.
When you inherit a brownfield codebase, T&M with a short discovery spike is usually smarter than fixed price—unknown coupling and hidden tech debt are not moral failures; they are information problems. Convert uncertainty into a phased plan: stabilize, measure, then commit fixed-price milestones for well-bounded modules once the terrain is mapped.
For AI initiatives, split contracts intentionally: a fixed discovery package for architecture and risk review, T&M for experimentation and evaluation, then fixed milestones only for integration surfaces that are deterministic (APIs, auth, data pipelines). This pattern keeps finance comfortable while preserving the empirical loop your engineers need to ship reliable models in production.
