Buy vs. Build Calculator

Every number is editable — start from the defaults, then plug in your lab's reality. Nothing is saved or sent anywhere; this runs entirely in your browser.

Time horizon 3 yrs

● Buy commercial software

Validated, supported, working in days.

Annual license / subscription $ per year One-time onboarding / training $ total Time to productive use weeks

Advanced

Extra annual support / modules $ per year

● Build on open source

Hire someone to implement, validate & maintain it.

Typical open-source building blocks: napari, Fiji/ImageJ, DeconvolutionLab2, BigStitcher, QuPath, CellProfiler, elastix.

Who builds it? Annual salary / stipend (base) $ per year Benefits / fringe rate % of salary Their effort spent building it % of their time Time to a working, validated pipeline months

Includes data/format plumbing (CZI, ND2, OME-TIFF…) and QC against ground truth — the parts labs routinely underestimate.

Ongoing upkeep after launch % of their time

Advanced

Re-build when they leave every N years (0 = never) Data & GPU infrastructure, build-side $ per year

Storage + compute to engineer TB-scale pipelines. Leave 0 if shared with the commercial path.

PI / staff oversight $ per year

Commercial — total

Build — total

Time to first results

—

Cumulative cost over time

Commercial Build on open source — the crossover point, if any, is where one becomes cheaper than the other.

When building your own genuinely wins

The method is the science. If you're developing a novel imaging or analysis technique that will itself be published, the salary isn't overhead — it's the research. Count it differently.
No commercial tool fits. Your modality, probe, or sample is unusual enough that off-the-shelf can't do it, and building is the only path.
In-house expertise + a long, stable horizon — a developer who won't graduate mid-project and will actually maintain the pipeline as formats and OSes change.
Open source is often the right base either way (napari, Fiji, BigStitcher, elastix, QuPath, CellProfiler) — the question is whether your lab should be the one integrating, validating, and maintaining it.

The usual hidden costs in imaging: getting a pipeline to handle real TB-scale data and a zoo of proprietary formats robustly; validating it well enough to trust in a figure; and the day the person who built it graduates and the undocumented code stops running. Months spent building are months before any images get analyzed.

How these numbers are calculated (full transparency)

Loaded cost of the person — salary plus benefits:

loaded = base_salary × (1 + fringe%)

Commercial — cumulative cost at year t:

commercial(t) = onboarding + (license + support) × t

Build — cumulative cost at year t. Effort is split into a build phase then a maintenance phase, repeating each time the person turns over (every N years):

build phase (first build_months of each cycle):
    loaded × build_effort% × (build_months ÷ 12)

maintenance phase (rest of each cycle):
    loaded × upkeep% × (years spent maintaining)

build(t) = Σ (build + maintenance over all cycles up to t)
           + (data/GPU infrastructure + oversight) × t

Totals shown in the cards are each curve evaluated at t = your time horizon. The break-even point on the chart is the year where the two cumulative curves cross (if they cross within the horizon).

Time to first results is compared directly: commercial in weeks, build in months.

Not modeled (and noted below): indirect / F&A costs, and any inflation or discounting. Both tend to widen the gap in favor of buying; leaving them out keeps the estimate conservative.

Estimates only, for planning discussion — not a quote or financial advice. Default postdoc salary uses the NIH NRSA FY2026 Level-0 stipend ($63,480); adjust to your institution's fully-loaded figures. Indirect (F&A) costs, which often apply to personnel but not to software licenses, are not modeled and typically widen the gap further.

Imaging buy-vs-build TCO calculator · runs locally · adjust every input to your lab.