Replatforming cost estimate: a 3‑year TCO framework

Replatforming cost estimate is the single decision that determines whether your technical debt becomes a competitive advantage or a balance‑sheet time bomb. You should expect a replatform to change your run rate by hundreds of thousands of dollars per year and to take six to 18 months of focused engineering time.

Direct answer: If you ask, “What is a reasonable replatforming cost estimate for my SaaS?” plan for three clear line items: engineering labor ($400k–$1.2M for a 4–8 month core delivery at $180k/engineer loaded), third‑party and cloud delta ($20k–$180k/year), and opportunity/downtime risk ($0–$500k depending on churn and lost sales). A defensible 3‑year TCO model will include all three and show net present value at a 8–12% discount rate.

The stakes are concrete. A five‑engineer product team at $180k loaded runs roughly $900k/yr; outsourcing the same scope to a specialized firm can cost $250k–$750k one‑time plus $150k/yr for maintenance. Meanwhile, a 99.9% to 99.99% availability improvement worth a single enterprise contract can exceed $300k/yr in revenue retention. Those are the numbers you model — not vague engineering effort.

Replatforming cost estimate: components and trade‑offs

Break every replatform into four accounting buckets: delivery labor, platform delta (cloud/SaaS), migration operations (data movement, egress, rollbacks), and deferred‑opportunity cost. Delivery labor is the largest and most certain line. Use a loaded engineer rate between $150k and $220k; we use $180k as a conservative median for US teams. A 6‑month focused delivery at four engineers is $360k; add bugfix and stabilization at two engineers for six months and you’re at $540k.

Platform delta can go either way. Moving from self‑hosted Postgres on EC2 to a serverless provider like Neon or PlanetScale might increase monthly costs by $1k–$8k, but reduce ops headcount by 0.2–0.6 FTE, worth $36k–$108k/yr. Conversely, splitting a monolith into microservices can raise total monthly cloud spend by 15–60% because of more egress, more containers, and duplicated baseline services. Estimate a conservative 25% cloud cost premium first year.

Migration operations are often underestimated. If you transfer 50 TB of active data, AWS egress at $0.09/GB costs $4,500 for a single month; a phased migration with backfills and validation can triple that. Add professional services for ETL and schema conversion — expect $40k–$150k for nontrivial data migrations. Include a rollback window cost: 2–4 weeks of parallel run at peak capacity, not just a dry‑run.

Deferred‑opportunity is the hardest number to model but the most important. A replatform that shortens time‑to‑ship for a new enterprise feature by three months can capture an extra $250k–$1M of ARR in the first year depending on sales cycle length. Conversely, an eighteen‑month rewrite that stalls launches can cost the business >$1M in lost deals and feature‑market fit erosion.

Treat a replatform as a capital allocation: model labor, platform delta, migration ops, and the revenue impact — one missing bucket will blow your TCO by 30%.

How to build a defendable 3‑year replatforming TCO

Start with a year‑by‑year cash flow table. Year 0 (the project) shows one‑time delivery labor and migration ops. Years 1–3 show recurring platform delta and maintenance. Example: a 9‑month replatform with 6 engineers core + 2 PMs costs $1.26M in Year 0 (8 people × $180k prorated). Add $60k professional services and $30k extra cloud for migration. Year 1 recurring delta might be +$90k, Year 2 +$60k as savings from reduced incident load kick in, Year 3 +$30k net. Discount at 10% to compute NPV.

Quantify risk buckets with scenarios: best (50% faster time‑to‑market, cloud cost −10%), base (no feature velocity change, cloud cost +15%), and worst (feature delays, churn +1.5%). Assign dollar values: a 1% churn uptick on $12M ARR is $120k/yr. An enterprise contract delayed by three months reduces net revenue by ~25% of that contract’s first‑year ARR. Modeling these as probabilities and expected values is how you get an honest replatforming ROI.

Include sunk vs amortizable costs. Tools and licenses purchased for migration (schema conversion tools, observability seats like Datadog or Grafana Cloud) are amortized over 24–36 months. If you buy $90k of tooling that reduces migration time by 30%, that is often cheaper than 1.5 FTE months of developer time (1.5 × $15k = $22.5k/month at $180k/yr, so $33.7k for 1.5 months). Do the math before buying.

Don’t forget operational lifting costs: incident runbooks, playbooks, and a 4–8 week post‑launch hypercare at 2–3 engineers. The hypercare period prevents a 0.5–2% revenue impact in the 12 weeks post‑launch. If your monthly MRR is $300k, that’s $1.5k–$6k/week in potential risk; insuring it with headcount is justified.

What this means for a CTO or technical founder

You decide with three questions: (1) Can you tolerate a 6–18 month delivery window? (2) Does the platform delta create recurring savings large enough to justify the one‑time cost? (3) Is the replatform the only way to capture critical revenue in that window? If you can answer yes to two of three, proceed with a phased plan and stop planning a big‑bang rewrite.

If you are budget constrained, prefer an incremental replatform: pick the highest‑value slice (API surface or heavy read path) and migrate it first. A 3‑month vertical extraction that reduces p95 latency from 600ms to 120ms on the invoicing path can directly increase conversion by 4–7% — that’s easier to justify to CFOs than a full‑monolith rewrite priced at $1.2M.

If you’re evaluating vendors, require three deliverables before signing: a migration runbook with explicit rollback criteria, a bounded SLA for the migration window (hours of allowable downtime), and a fixed scope for post‑migration support priced as a capped T&M. That moves the negotiation from vague promises to measurable commitments and reduces hidden costs.

3‑step checklist to produce a credible estimate

1) Inventory: list code, data, integrations, and their owners; estimate effort to move each (small = <2 engineer days, medium = 2–10 days, large = >10 days). 2) Price out line items: engineer months at loaded rate, migration tooling, egress, professional services, and hypercare. 3) Model three scenarios (best/base/worst) and compute NPV at 8–12%.

Use real comparables: a 9‑month external replatform engagement for a similar SaaS often quotes $350k–$850k; internal cost for the same scope is 0.6–1.4× that when you include opportunity cost and lost launches. Cite those ranges to your CFO and ask for a contingency line of at least 20%.

When presenting to the board, show the delta not the totals. Boards care about incremental cash and time. Show “do nothing” versus “incremental replatform” over 36 months and highlight the break‑even point in months and the probability‑weighted upside in ARR expansion.

Key takeaways

1. A defensible replatforming cost estimate models delivery labor, platform delta, migration ops, and deferred‑opportunity; use $150k–$220k as a loaded engineer range. 2. Expect 6–18 months of calendar time and $400k–$1.2M in delivery labor for a typical mid‑market replatform. 3. Quantify revenue risk and opportunity: a 1% churn or a three‑month enterprise delay are real dollar items. 4. Prefer phased extractions that deliver measurable business metrics (latency, conversion, uptime) before a full rewrite. 5. Always include a 20% contingency and amortize tooling over 24–36 months.

Replatforming cost estimate is a funding decision disguised as engineering planning. When you price every bucket — labor, cloud delta, migration ops, and lost or gained revenue — you turn an emotional bet into a manageable investment. Do the work up front, stay skeptical of vendor flat rates that omit rollback costs, and treat the first migrated slice as your financial experiment.