Headless CMS selection — a decision framework.

Last updated: April 2026

Quick answer

Pick a headless CMS by scoring five criteria against your real workload: content modeling flexibility, non-technical editor UX, localization support, integration surface, and total 3-year cost. Weight the criteria against who actually uses the system day-to-day — the CMS that looks best in a demo is rarely the one marketing editors will still be using in year three. We publish a scoring rubric with every selection engagement.

Why selection fails

Most headless CMS selections fail for one of three reasons: the team picks based on a 20-minute vendor demo, the evaluation ignores the people who'll actually use the system, or the pricing comparison stops at license cost and misses implementation plus maintenance.

This framework solves all three.

The five criteria

1. Content modeling flexibility

Can the CMS express your actual content model — not a sanitized version? Test with the hardest content type you have: a case study with nested media, a product page with variants, a localized event with per-region fields. If you have to flatten the model to fit the tool, keep looking.

2. Non-technical editor UX

The system's users are marketers, editors, and legal reviewers — not developers. Score editor UX with real editors, not stakeholders. Have them build the same page in each candidate, timed. Record blockers.

3. Localization support

If you publish in more than one language or market, localization is a first-class requirement, not a plugin. Look for field-level translation, fallback logic per locale, workflow-aware approval by locale, and a translation-memory integration or open API.

4. Integration surface

Every marketing site integrates with 10–20 systems: CRM, marketing automation, analytics, search, personalization, DAM, commerce, and increasingly an LLM layer. Score the CMS on webhook quality, API stability, and the realism of its third-party connector ecosystem. Open APIs beat proprietary plugin marketplaces for long-term flexibility.

5. Total cost of ownership over 3 years

Licensing + implementation (build + content migration + editor training) + ongoing maintenance. Anything shorter than a 3-year view will bias toward cheap-licensing tools that cost more to maintain.

A worked example — mid-market B2B

Client: 400-page marketing site, 3 locales, Salesforce + Marketo + Algolia + Cloudinary.

Shortlist: Sanity, Contentful, Storyblok.

Modeling: Sanity wins — code-first schema definition scales with complex models.
Editor UX: Storyblok wins — visual editor is the most forgiving for non-technical editors.
Localization: Contentful wins — field-level locale controls and workflow are best-in-class.
Integration surface: All three clear the bar; Sanity's GROQ query language pulls ahead for custom integration work.
3-year TCO: Storyblok lowest, Contentful highest; Sanity middle with the caveat that developer time shifts from configuration to schema management.

Decision in this case: Sanity, because the client's content model complexity (not licensing) dominated the selection. On a simpler site, the same framework would push toward Storyblok.

The rubric we publish

We publish the scoring rubric with every selection engagement. Clients can audit the weights, replay the evaluation, and fork the approach for future platform decisions. If a selection doesn't come with a rubric, you're buying an opinion, not a decision.

Sources & further reading

NUUN Software & Product Development service pillar
Marketing websites & headless CMS service page
Jamstack.org — https://jamstack.org/
Gartner Digital Experience Platform reports — https://www.gartner.com/
MACH Alliance — https://machalliance.org/

About the author

Feras Nasser

Principal, NUUN Digital

15+ years leading headless CMS selections and composable commerce replatforms across Contentful, Sanity, Payload, and commercetools; enterprise integration focus.

Frequently asked.

What is a headless CMS?

A content management system that stores structured content and exposes it via APIs, leaving the presentation layer to a separate front-end (Next.js, Astro, mobile apps, etc.). Compare to traditional CMSes like WordPress where content and rendering are coupled in one stack.

Which headless CMSes are worth evaluating in 2026?

For enterprise: Contentful, Contentstack, Sanity. For mid-market: Sanity, Storyblok, Hygraph, Payload. For teams that want self-hosted with code-first configuration: Payload, Strapi. The right answer depends on editor UX expectations and integration surface, not brand prestige.

How long does a headless CMS selection take?

Three to six weeks for mid-market. Week 1: workload discovery and scoring criteria. Weeks 2–3: shortlist of 3 vendors, side-by-side hands-on evaluation. Weeks 4–5: pricing, security, integration proofs. Week 6: recommendation and migration plan.

Should we pick the CMS before or after the front-end framework?

After. The front-end framework is a bigger decision (team skill set, performance envelope, ecosystem), and most modern headless CMSes work with most front-end frameworks. Lock the front end first, then pick the CMS.

What's the total cost of a headless CMS over 3 years?

Licensing ranges from $10k/year (entry tier) to $150k+/year (enterprise). Add implementation (50–200 hours at your blended rate), editor training (20–40 hours), and ongoing integration maintenance (0.2–0.5 FTE). Most mid-market deployments land at $60–180k over 3 years all-in.

How do we evaluate editor UX objectively?

Have two non-technical editors build the same content model and the same three pages across candidate CMSes, timed. Record the sessions. The CMS where both editors finish without asking for help is the one that will survive year three.

Can we migrate from WordPress to headless without a re-platform project?

Partially. Most teams move content first (migrate posts and pages to the new CMS), run WordPress for legacy templates for a quarter, then rebuild the front-end. A clean cutover in under 12 weeks is only realistic when the existing site has fewer than ~200 templated pages.