Why a paid invoice does not mean control
A typical scene: after three years, a business decides to change agency. The new agency asks for access to the hosting, the domain management and the analytics account. Nobody in-house knows where these logins are. They run through an email address belonging to the old agency, and so do the invoices for servers and licences.
Legally, the business may be entitled to a great deal. In practice, everything depends on the old agency cooperating. Usually it does. But even with goodwill, it takes weeks before every account has been transferred. Some cannot be transferred at all, because they are tied to an agency account.
The problem rarely stems from bad intent. Agencies set up accounts themselves because it is quicker during the project. Once the project ends, nobody asks whose name they are in. The dependency is therefore not a contractual detail but a by-product of convenience. Nowhere is this clearer than with the domain.
Domain and DNS are what everything else hangs on
The domain is the address under which the website, and usually email, can be reached. Whoever controls it controls both. What matters is who is registered as the holder, which for .de domains means the record held by DENIC. If the agency or a former employee is listed there, the domain formally does not belong to the business.
A separate matter is the account with the registrar, the provider through which the domain is managed. That is where you retrieve the auth code needed to move the domain to another provider. It is often also where DNS is managed, which determines where website traffic and emails are routed. A business can be the holder and still have no access to this account.
This can be checked in a few minutes. Who is the holder according to the registrar, which email address the account uses, and who pays for renewal. If the invoice goes through the agency, a missed payment can cost the domain. It is one of the few mistakes no backup can fix.
Accounts belong to the business, not the agency
Beyond the domain, a website depends on many other services. These include hosting, the CMS, analytics, Search Console and a tag manager. Then come form and newsletter services, payment providers and plugin licences. The same question applies to every one of these accounts: who is the owner, and who has merely been invited.
Here lies an insight that runs against intuition. The agency having administrator rights is not a risk. It needs those rights to do its work. The risk is the agency being the only owner. Many services explicitly distinguish between owners and users with extensive permissions. Only the owner can remove users, change billing and delete the account.
The right arrangement is therefore: the business is the owner, the agency is an authorised user. If the business changes agency, it revokes access itself without having to ask for cooperation. The other way round, the transfer often takes a long time. Some services do not allow it at all, and the account has to be recreated, losing its history.
A second trap is personal accounts. If the analytics account runs through the marketing manager’s private address, it is effectively lost when she leaves. Accounts belong on role-based addresses such as web@ or it@, which more than one person can access. Two-factor codes should not live on a single company phone. They are better kept in a shared password manager or with two responsible people.
Source code and usage rights: what a website project handover must put in writing
With source code, many people confuse two things: possessing the files and having the right to use them. Under German law, copyright stays with its author. What is transferred are usage rights, and their scope follows from the contract. Where the contract is unclear, the default covers only what the purpose of the agreement requires.
For a business, this means the contract should state that it may use and modify code and design without time limits. It should also allow third parties to develop the website further. That right to have another agency work on the code is missing from many agreements. Without it, changing agency is legally uncertain, even when all the files are available.
Technically, the code belongs in a repository, a versioned store such as GitHub or GitLab, held in the business’s name. A ZIP file from the end of the project is not enough, because it contains neither history nor later changes. Platforms such as Shopify or Webflow often have no code in the classic sense. There, what counts is ownership of the shop or project and a complete export of the content.
A special case is components an agency reuses across many projects, such as its own theme framework. It is legitimate for the agency to keep the rights here. But it must then be agreed that the business may use the component permanently and have others maintain it. Purchased licences for fonts, images and plugins should likewise be registered to the business.
Documentation must make sense without the agency
Access and rights are of little use if nobody knows how the system works. The usual documentation is an editor’s manual: how to create a page or swap an image. For a handover to a third party, that is not enough. What is needed is a technical description an unfamiliar developer can follow within a day.
Structure. Which systems are involved and how they communicate with each other. This includes where the website runs and which data flows where.
Operation. How to install the website locally and how an update is deployed. Also where backups are kept and how to restore one.
Dependencies. Which external services, interfaces and keys are involved. And what stops working if one of them disappears or its contract ends.
Decisions. Why something was built the way it was. Without this note, the next agency rebuilds a seemingly awkward solution. In doing so, it often breaks something that was designed exactly that way.
Where the documentation lives matters too. It belongs in the repository or on the business’s own drive, not in the agency’s project tool. Project tools are archived when a project ends, and the knowledge disappears with them.
How to check whether the website project handover is complete
Ticking off a checklist is easy. Only a practical test shows whether the handover holds. It takes about half a day and answers a simple question: what could the business do if the agency were unreachable tomorrow.
In concrete terms, someone in-house logs into every account themselves, without help from the agency. They check whether they have owner rights and whether the invoice goes to the business. They also test whether password recovery works through a company address. Wherever a login fails, a gap has been found.
The same thinking applies to code and operations. A backup only counts as existing once it has been restored as a trial. To go further, have an outside developer read the documentation and start the website locally. Whatever goes wrong shows which knowledge exists only in the agency’s heads.
The result belongs in an overview of all accounts, listing owner, internal responsibility and notice periods. This overview is not a password store; the credentials themselves stay in the password manager. It is updated with every staff change and every new service. Otherwise it is out of date within a year.
Where full independence is not worth it
Independence has a price, and not every form of it pays off. A business with a small website and no in-house technical skills gains little from managing servers itself. Managed hosting makes more sense in that case. All that matters is that the account is in the business’s name.
Documentation can be overdone as well. A hundred-page manual nobody maintains is wrong after two updates. That makes it worse than a short, current note. The aim is not to explain every line, but to give a capable third party a way in.
Finally, no handover prevents a change of agency from taking effort. A new agency needs time to get up to speed, even with very good documentation. The handover reduces that effort from a rebuild to an induction. It promises nothing more.
A good handover makes the relationship voluntary
A complete project handover is not a sign of mistrust. It changes the relationship: the business stays with its agency because the work is good, not because leaving would be too risky. For an agency that takes its work seriously, that is the better footing.
The simplest approach is to raise the handover when choosing the agency and to put it in the contract. It can be done afterwards, just with more effort. What remains is the ongoing task of keeping accounts, licences and documentation up to date.
How we look after systems so that control and knowledge stay with the business is described under AI Operations Care.