Guides

In-House or Agency? The Honest Magento Maintenance Decision

September 15, 2026 - Alexandru-Manuel Carabus

The in-house versus agency question is really five smaller questions about skills, workload, coverage and cost shape. Here is the honest decision matrix, when each model wins, and the hybrid most serious stores actually end up running.

On an off-white surface, a small train of matte precision gears meshed together and driving one another, with one muted-orange gear at its heart, beside a single lone gear resting by itself and turning nothing, symbolizing a genuine agency plugged into a wider system of experience, community and better methods versus an in-house team working alone from scratch.

Here is the question that actually decides it, and it is not about budget. It is Friday, 17:45. Checkout just broke. Who owns that? Not "who could look at it," not "who do we usually call." Whose phone rings, what have they committed to, and what happens if they are on vacation?

If you have a crisp answer, you have already made the in-house versus agency decision, whether you formalized it or not. If you do not, the store is running on the third model nobody chooses on purpose: unowned. Most Magento stores that get hurt are not on the wrong model. They are on no model, with maintenance done by whoever has a free afternoon, which works right up until the afternoon it does not. This piece is the honest comparison, including the costs each side prefers not to mention, and the hybrid that most serious stores quietly converge on.

What "maintaining Magento" actually demands in skills

The full maintenance checklist looks like a list of tasks. Staff it, and it turns out to be a list of professions:

Backend Magento engineering. Patches, module conflicts, upgrade paths, the parts of the checklist that touch code.
Frontend engineering. In 2026, realistically Hyvä: Tailwind, Alpine.js, and the discipline to keep Core Web Vitals green after every change.
DevOps. Varnish, Redis, OpenSearch, CDN, deployment pipelines, the restore tests nobody enjoys and everybody needs.
Security operations. Reading advisories the day they publish, knowing which ones apply, moving in hours when it is a SessionReaper and not a footnote.
Release management. Staging discipline, testing, the boring process that separates an update from an outage.

That is five competencies. The awkward truth of the hiring market: they rarely live in one person, and the people who do cover three or four of them are senior, expensive and heavily recruited. The in-house question is never "can we hire a Magento developer." It is "how many of these five do we get per hire, and who covers the rest."

The cost of in-house that the job ad does not show

A senior Magento engineer's salary is public knowledge on any job board, and in Romania it alone exceeds what most stores spend on a full external maintenance program in a year. But the salary is the visible part. The full in-house bill includes recruitment (three to six months to find someone genuinely senior, in a market where they are the most fought-over profile), the coverage problem (one person means no vacations without risk, no sick days without risk, and an interview elsewhere is a business continuity event), the bus factor (one resignation walks out with the entire operational memory of your store), and skill drift (one person alone rarely stays current across all five competencies; training time is real time).

None of this argues in-house is wrong. It argues in-house is a real program with real overhead, not a salary line. Stores that budget the salary and skip the rest end up with the worst of both models: in-house cost, unowned outcomes.

When in-house genuinely wins

Ecommerce is the business, not a channel. When the store is the company, platform knowledge is strategic capital worth accumulating in-house, an argument we have made elsewhere about owning your stack in general.
There is 40-plus hours a week of real platform work. Daily feature development, constant integrations, a living roadmap. At that workload an external hourly model gets expensive and an internal team gets efficient.
You can staff at least two or three people. Two seniors, or a senior plus mid plus DevOps support, cover the five competencies with vacation coverage and a survivable bus factor. One person is not a team; it is a single point of failure with a payroll number.
Accumulated domain knowledge pays daily. Complex B2B catalogs, custom pricing engines, deep ERP coupling: contexts where re-explaining the store to an outsider costs more than employing an insider.

When an agency genuinely wins

Maintenance workload is real but under one full-time job. Most sub-10M EUR stores need serious maintenance but not 40 hours of it weekly. You cannot hire 30 percent of a senior engineer, but you can contract exactly that capacity.
You need a guarantee, not an intention. An employee owes you working hours; an SLA owes you a response time. Ours commits to two working hours on critical incidents, in the contract. No single hire can promise that across their own vacations, weekends and notice periods; teams absorb absences that individuals cannot.
You need all five competencies at once. A retainer with a real team puts backend, frontend, DevOps and security under one agreement, at a fraction of one senior salary. The math of what maintenance actually costs makes the comparison concrete.
The workload is spiky. Quiet months and then a peak-season emergency is the exact shape salaried staffing handles worst and a retainer with response guarantees handles best.

The hybrid most serious stores end up running

Between two and ten million EUR in annual revenue, the honest answer is usually not a side in the debate. It is a split: an internal person or small team owning product development, the features and roadmap where accumulated domain knowledge compounds, with an external retainer owning the maintenance calendar, the patch cadence, the SLA coverage and the DevOps depth. The internal side does not burn its week on patch testing; the external side does not need to learn the pricing engine's history.

It also solves the coverage problem that kills pure in-house at this size: vacations, spikes and 17:45-on-Friday stop being staffing crises, while strategic knowledge stays in the building. Most of our own retainer relationships have exactly this shape: an internal developer or team we work alongside, with the calendar, the SLA and the escalations on our side of the fence. The Temanovelart group, twenty years in the market and running three storefronts including Kit&Kin, works with us on exactly this model; the Kit&Kin case study shows what the arrangement looks like in practice.

The decision matrix

Answer these five honestly and the model picks itself:

Weekly platform workload. Under 20 hours: agency. Over 40, sustained: in-house core. Between: hybrid.
The Friday 17:45 question. Need a contractual response time: agency or hybrid. Genuinely fine with best-effort until Monday: in-house can carry it.
Patch discipline. Someone applies security patches within days, on the payroll, even in August: in-house passes. Otherwise the calendar needs an external owner. The SessionReaper numbers show what this single row is worth.
Team depth. Can you staff two-plus people covering the five competencies with real coverage? If not, in-house is one resignation from zero.
Cost shape. Salaries are fixed and step-shaped; retainers are flat and adjustable quarterly. Match to your revenue curve, not to taste.

Three or more answers pointing the same way is your model. A split verdict is the hybrid telling you it is the answer.

Red flags on both sides

Going in-house, worry when: the whole plan is one hire, however good; patches wait because the sprint is full (product work outranking maintenance daily is unowned maintenance); vacations mean crossed fingers.

Hiring an agency, worry when: there is no written SLA, only adjectives; there is no monthly report, so you cannot see whether the checklist actually ran; everything is billed per incident, which pays your provider best when your store breaks most; nobody maintains a staging environment. Our three self-tests are also a fast audit of whichever provider or person owns your store today: run them and you will see how well the current model is working.

Where to go from here

Write down your five matrix answers. If they say in-house, hire for the calendar and the coverage, not just the CV, and give maintenance its own protected hours. If they say agency or hybrid, hold every candidate against the five questions from the cost teardown, SLA in writing first. What our own program includes is on the support and development page. And if the matrix came back split and you want a second opinion on your specific shape, book a free strategy call. We will tell you honestly which model fits, including the versions that do not involve us.