Integrator, consultancy, hire or freelancer: what each model really costs

Key takeaways

• Four ways to fund a Kubernetes platform — T&M with an integrator, fixed scope with a specialist, hiring in-house, a freelancer — and none of them compare on day rate.
• The right calculation is the full cost of the deliverable, elapsed time included: a cheaper profile who takes three times as long costs more and loses you a quarter.
• Four situations where the large integrator is objectively right are listed here. If one is your binding constraint, the rest of this article does not apply to you.

Most integrator-versus-specialist decisions come down to a table of day rates. It is the only immediately comparable number, so it is the one that gets compared — and it is also the least informative.

A day rate only means something paired with a speed. A €600 profile who takes nine months where an €1,100 profile takes three costs more in absolute terms, and additionally costs you two quarters of time to market. Conversely, an expensive senior on work that never needed seniority is pure waste.

The meaningful unit is not the day. It is the full cost of the deliverable, milestones and elapsed time included. This article compares the four funding models on that basis.

Disclosure: Edixos is a specialist consultancy, one of the four models described. We wrote the “when the large integrator is right” section with the same care as the rest, because we say it regularly in meetings.

The four funding models

Time and materials with a generalist integrator

You buy person-days. The integrator places consultants in your teams, you direct the work, billing follows time spent.

What it solves: capacity, immediately, across a broad catalogue. And procurement compliance: if your organisation requires a pre-approved supplier, this is often the only workable route.

What it actually costs: the staffing structure is a pyramid, and margin comes from the gap between the profile sold in pre-sales and the profile delivered. On deep work — writing CRDs, designing a control plane, settling an abstraction boundary — that gap is paid in months of drift. And because billing follows time, nothing in the contract rewards finishing.

Fixed scope with a specialist consultancy

You buy an outcome on a defined scope: a control plane, a migration, an internal platform.

What it solves: depth and alignment. The vendor is paid to deliver, not to persist. The engineers who show up are the ones you met, because there is no bench behind them.

What it actually costs: rigidity. Fixed scope assumes a stable perimeter; if your requirements move monthly you will spend your time on change requests. And a ten-person firm will not staff forty roles, will not cover your security workstream, and probably will not clear your procurement process.

Hiring in-house

You build the platform team.

What it solves: everything, eventually. A platform is a permanent asset; over the long run nobody will run it better than people whose job it is, inside your company.

What it actually costs: time, and a threshold. Hiring a senior platform engineer routinely takes six months. More importantly, a credible platform team starts around six people: below that you cannot hold an on-call rotation and you have created a single point of failure carrying your production. The lowest-risk sequence is often to outsource the build, then hire for operations while the platform already exists.

The freelancer

You buy one precise skill with no overhead.

What it solves: the best value in the market for an audit, an unblocking, or reinforcement on a team that already exists.

What it actually costs: bus factor. Nobody to review, no continuity, and a control plane designed by one person without review becomes an asset you can neither evolve nor hand on.

The decision matrix

When the large integrator is the right call

Four situations, and they are common:

1. Volume. Forty people next quarter. No specialist consultancy can answer that, and pretending otherwise would be dishonest.
2. Process. Supplier pre-approval, framework agreements, three years of minimum revenue, multi-million liability insurance. These criteria mechanically exclude small firms regardless of competence.
3. Coverage. You need cloud, data, security and mobile under one contract with one accountable counterparty.
4. Signal. Your board needs to recognise the name. That is a real requirement, not a weakness, and it has value.

If your binding constraint is on that list, the rest does not apply to you, and you will save time by calling a large integrator directly.

When the specialist is the right call

1. The problem is deep rather than broad. A control plane, CRDs that model your domain, an exit from Terraform sprawl, multi-tenancy, bare metal. This is the only dimension where a five-person team can be objectively better than anyone.
2. Your team must inherit the result. If autonomy is the goal, the commercial model has to fund it. A contract that bills time will not.
3. The scope is definable. You know what you want at the end, even if you do not yet know how to get there.
4. You want to talk to the engineer, not a salesperson. In a small firm they are the same person. That is the one structural advantage a small company has, and it is real.

The part nobody else will tell you

A vendor who also sells the operation of your platform has an economic interest in you never taking it back. That is not bad faith, it is recurring revenue — and for many organisations it is exactly the right purchase.

But ask explicitly before signing: “contractually and technically, what happens if we want to take over operations in two years?” The quality of that answer will teach you more than any reversibility clause.

What Edixos is

Model 2: fixed scope, defined perimeter, senior team, and a platform your engineers then operate. Our engineers are named on the team page, our work is public — a Crossplane provider for OVHcloud with 15,000+ downloads — and our practice goes back to Kubernetes 1.2, in 2015.

The alignment question above applies to us: we also offer managed cloud operations. What you can verify in the contract rather than take on trust: it is a separately contracted service, and the repository, documentation and runbooks are yours from day one.

We are not an answer to a capacity problem, and we say so in the first meeting when that is the case.

To work out which kind of provider fits your situation before you get to budget, see Kubernetes expertise: how to choose a consulting partner — the five models, the seven questions to ask, and the red flags.

The lowest-risk next step is a two-to-five-day architecture review.

Book a conversation · cloud-native consulting · platform engineering · Crossplane