What does it actually cost to build it yourself?
A build estimate and a vendor quote are not the same number. This prices what shows up after the decision: the maintenance, the orphaning risk, the months before anything works.
1Start from a company like yours
2Adjust anything that doesn’t match
Shared assumptions
Applied to bothBuild it in-house
$327,400Engineer-weeks × fully loaded salary. Count design, review and rollout time, not just the build — the estimate people give is almost always the coding alone.
Bug fixes, upgrades, migrations and the rest of the tedium of owning a project. Benchmark: ~42% of a developer’s week for homegrown or unmanaged systems (Stripe).
Measured against the fastest option on the table, not from zero. If a platform is live in one month and your build in five, the delay that counts is four.
The system becomes orphaned — the engineer who owned it leaves, gets promoted, or moves on. Estimate 20–40% for single-engineer builds; lower for anything with documented, funded ownership.
What it costs to rebuild or replace it if it is abandoned. Often ≈ C₀ again, sometimes more once other systems have grown to depend on it.
The one term that can make building win. Assume $0 unless this is a core competency: if what you are building is a commodity or a cost center, there is no net-new value to count.
Buy a platform
$308,000The quoted list price, less whatever you actually negotiate. Include the support tier you would realistically buy, not the cheapest one on the sheet.
Any one-off onboarding, migration or professional-services line on the quote. Zero for most self-serve platforms.
Configuration, integration and upgrades do not disappear when you buy. McKinsey puts upkeep at 10–20% of budget even for purchased tools carrying unresolved debt.
Compare actual onboarding timelines from the vendor’s existing customers, not the sales estimate.
The chance you replace the vendor before the horizon is up — acquisition, price rises, or a capability gap. Usually well below the abandonment risk on a homegrown build.
Data export, re-integration and retraining if you move off. Lower than a from-scratch rebuild, but rarely zero.
Where the money actually goes
- Build
- Buy
Directional cost, not a budget-grade forecast. Profile figures are published benchmarks for a company of that shape, not a claim about yours. Your inputs stay in this browser.
The equation behind it
Quantifying a build-or-buy decision is always part art, part science. It is easy to calculate a subscription fee; it is much harder to know whether that fee goes up in six months. Better to have a number and be slightly wrong than to never have had a number at all.
TCO = C₀ + n·(M·S + L) + Pₐ·C_r + TTV·OC − n·V_d
A mid-size engineering org decides whether to build its own software factory in-house. Three engineers for six weeks is about $70K of loaded time; maintenance runs at 42% of one engineer; there is a real chance the person who owns it moves on.
TCO = 70,000 + (0.42 × 140,000 × 3) + (0.30 × 70,000) + (4 × 15,000) − 0
TCO = $327,400 over 3 years
Compare that to a managed platform at $90K a year over three years, with minimal time-to-value delay, and you have a defensible case for buying the point solution instead of maintaining someone’s science project.
Where to get a starting number
| Var. | Meaning | Where to get a starting number |
|---|---|---|
| C₀ | Initial cost — build hours × loaded cost, or the license and setup fee. | An actual quote, or engineer-weeks × fully loaded salary. |
| M | Ongoing maintenance burden, as a share of engineering time: bug fixes, upgrades, migrations, and the rest of the tedium inherent to owning a project. | Benchmark: ~42% of a developer's week for homegrown or unmanaged systems (Stripe); 10–20% of budget even for purchased tools carrying unresolved debt (McKinsey). |
| S | Fully loaded annual cost of one engineer — salary plus everything around it. | Your own comp bands, loaded for benefits and overhead. |
| L | Annual license or subscription fee. | The quoted list price, less whatever you actually negotiate. |
| n | Time horizon — the years the system will realistically be in use. | Usually 3–5 years for planning purposes. |
| Pₐ | Probability of abandonment — the system becomes orphaned because the engineer who owned it leaves, gets promoted, or moves on. | Estimate 20–40% for single-engineer DIY builds; lower for anything with documented ownership. |
| C_r | Cost to rebuild or replace it if that happens. | Often ≈ C₀ again, sometimes more once dependencies have grown around it. |
| TTV | Time-to-value delay, in months, relative to the fastest option on the table. | Compare actual onboarding timelines. |
| OC | Opportunity cost per month of delay — revenue, competitive ground, or engineering hours not spent on differentiated work. | The monthly value of the capability once it is live. |
| V_d | Annual differentiation value. Arguably the most important term, because it assumes some value is actually derived from the decision. If what you are building is a commodity or a cost center, there is no net-new value to count. | Assume $0 unless this is a core competency. You can also assume a point solution decreases this number by some percentage over time. |