Benchmarks · Preliminary protocol

WordPress monolith vs headless hosting: 5-year TCO

Status: cost model published, audited 5-year projection scheduled for Q4 2026.

Preliminary

I do not publish projections I cannot defend. This page ships the cost model and the assumptions so peers can audit them and so the eventual numbers carry provenance. Marketing claims of "edge saves 70 percent" deserve scrutiny; this protocol exposes the inputs that let any reader rebuild the projection on their own traffic.

Cost components

  • Hosting (managed PHP host or VPS for monolith; Workers + Pages + small origin for headless)
  • CDN (bundled with managed host or separate Cloudflare on monolith; bundled in Workers on headless)
  • Image pipeline (WordPress plugin or third-party for monolith; Cloudflare Images or R2 + Worker on headless)
  • Sysadmin time (estimated hours/month for updates, security, scaling)

Traffic scenarios

  • Scenario A: 100k monthly pageviews (small business, low peak)
  • Scenario B: 1M monthly pageviews (mid-market publisher, moderate peak)
  • Scenario C: 10M monthly pageviews (large publisher, frequent traffic spikes)

Hosting options compared

Monolith

  • WP Engine Standard plan with included CDN
  • Kinsta Pro plan with bundled Cloudflare
  • Self-managed VPS on Hetzner CCX series with Cloudflare on top

Headless

  • Cloudflare Workers Paid plan + Cloudflare Pages + R2 storage
  • Small WordPress origin (Hetzner CCX12 or equivalent budget PHP host)
  • Cloudflare Images for image pipeline (or R2 + Worker for self-pipeline)

Assumptions exposed

  • Edge cache hit rate: 60-85% (production WordPress observed range, not 95%+ marketing claim)
  • Peak-to-average ratio: 5x for scenario A, 10x for B, 20x for C
  • Image traffic: 60% of total bytes
  • 5-year horizon, 3% annual price escalator on all line items

Methodology file

The public methodology contains assumptions only, not audited spend: open the methodology and result template. The audited projection should only be added after the 12-month billing checkpoint described in this protocol.

The full benchmark methodology cluster is available at /en/benchmarks/methodology/.

Expected findings (predicted, not measured)

Headless tends to win on the high-traffic scenario because edge cache absorbs spikes that would force a monolith to scale up its origin. Monolith tends to win on the low-traffic scenario because the small managed PHP host bundles enough capacity that the parametric Worker cost adds friction without payoff. The crossover point is near scenario B, give or take 50% depending on the cache-hit rate.

I will only publish the projected numbers after auditing 12 months of real billing from at least one production migration. Until then, treat the predicted shape above as a hypothesis under test.

Cite this protocol

WPPoland. WordPress monolith vs headless hosting: 5-year TCO (preliminary protocol). Published 2026-04-28. URL: https://wppoland.com/en/benchmarks/wordpress-monolith-vs-headless-hosting-cost/

Frequently asked questions

What does total cost of ownership cover here?

Hosting + CDN + image pipeline + caching layer + sysadmin time. I exclude content production cost, design cost, and engineering hours, since those are constant across the two architectures. The benchmark is purely the infrastructure spend over a 5-year window.

What traffic profile is assumed?

Three scenarios: 100k monthly pageviews, 1M monthly pageviews, 10M monthly pageviews. Each scenario has a defined peak-to-average ratio and a defined cache-hit-rate distribution. The cost model is parametric so any reader can plug in their own traffic.

What hosting options are compared?

Monolith side: WP Engine Standard, Kinsta Pro, and a self-managed VPS (Hetzner CCX series). Headless side: Cloudflare Workers Paid + Pages + R2, with a small WordPress origin on a budget PHP host (Vercel/Netlify deliberately omitted as a separate axis tracked elsewhere).

How are cost-saving 'edge' claims handled?

Conservatively. I assume the edge cache hit rate observed in production WordPress sites (60-85%, not the 95%+ marketing claim). Cache misses pay full origin cost. The model exposes the assumption so readers can adjust.

When will the numbers ship?

The reproducible cost model and the 5-year projection are scheduled for Q4 2026, after I have at least one full year of live cost data from a production migration. Until then, this page ships the model structure and assumptions, not the projection.

Want to reproduce this cost model?

Get in touch. I share the spreadsheet harness with peers under CC BY 4.0.

Contact me