There is a version of this advice that is really about saving money, and this is not that version. A good deal of website work is genuinely yours to do, and doing it yourself is often the right call for reasons that have nothing to do with an invoice. You know your business. You can describe it faster than anyone you could hire to describe it for you. A site you are comfortable editing is a site that stays current, and a site that stays current outperforms a better-looking one that nobody has touched since launch.
So the useful question is not what you are allowed to touch. It is where the nature of the risk changes, and that line sits in a different place than most owners assume. It has very little to do with how technical the task looks.
What is genuinely yours
These are the jobs I would rather a client keep, because they go better in the hands of the person who knows the business.
Words and pictures. Page copy, service descriptions, staff bios, the answers to the questions your customers actually ask on the phone. Nobody writes this better than you do. Real photographs of your own work and your own people beat anything that can be bought, every time.
Routine content edits. New hours, a price list change, a seasonal notice, a staff update, a new blog post. If your site was built properly, you can make these from the editor without help, and you should.
Your own accounts. Domain registration, hosting, the website login, the email platform, the payment processor. These should be in your name with your billing details, and checking that is a job only you can do. I covered how to go through them in the five accounts every owner should hold.
Watching the things that quietly break. Submit your own contact form once a month and confirm the message arrives. It takes two minutes,s and it catches a failure most businesses hear about from a customer instead, as our walkthrough for testing a form covers. Load your own site on your own phone now and then.
Updates, carefully, on a simple site. If your site is small, if you have a staging copy, and if you have restored a backup at least once so you know the backup works, running your own updates is reasonable. Those three conditions matter more than the task itself, and the order you run them in is most of the skill.
None of that is the shallow end of real work. It is the part of the work where being close to the business is the qualification.
Where it stops: four signals
I am not going to give you a list of forbidden tasks, because the same task is fine on one site and a bad idea on another. What travels better is a set of signals. When one of these is true, carrying on alone tends to turn a small problem into a bigger one.
One: you cannot describe what undoing it would involve
This is the single most reliable test I know. Before a change, say out loud how you would put things back. For a content edit, the answer is easy, because the previous version is sitting in the editor’s revision history. For a plugin change, a theme file edit, a redirect, or anything that writes to the database, the honest answer is often that you do not know, and that is the signal. Not the difficulty of the change. The absence of a way back.
The corollary is that a change you cannot undo becomes a change you can undo as soon as there is a staging copy and a tested backup behind it. The line moves when you build the net, which is why the net is worth building.
Two: you are treating a symptom without knowing the cause
Something is broken, so you deactivate a plugin. Nothing. You deactivate another one. You switch a setting, clear a cache, install a tool that promises to fix the category of thing that is wrong. Somewhere in that sequence the original problem gets tangled up with three new ones, and the site is now in a state nobody can describe.
I get called into this one often, and the repair is almost always longer than the original fault would have been, because the first job is working out what was changed and in what order. If you have made more than two changes without a clear theory of the cause, stop and write down what you have already tried. That list is genuinely valuable to whoever picks it up, including you tomorrow.
Three: the thing you are changing creates records or takes money
Checkout, booking, intake forms, anything that writes a customer record, anything connected to a payment processor. The failure mode here is not a broken page you can see. It is a quiet one: orders that do not arrive, bookings that double, a form that submits into nothing while showing a cheerful confirmation. Nobody complains, because nobody knows. You find out when the month’s numbers come in lower than the month felt.
Work on these paths needs testing before and after, with a real transaction on a staging copy, by somebody whose job is to be suspicious. That is a different activity from editing a page, and it is the clearest stopping point on this list.
Four: accessibility, and anything with an obligation attached
Accessibility is a build quality, not a setting. Whether a page works for somebody using a keyboard or a screen reader depends on how the markup is put together, so it has to be checked on the rendered page. Adding a plugin on top of a page does not change the markup underneath it, and anything that promises compliance without changing the page is promising something it cannot deliver.
If you are a non-profit, a public sector office, a school, or a practice of any kind, this probably sits inside an obligation you already carry. That is a reason to get it verified rather than assumed, and it is not a job to improvise.
The time signal, and one honest exception
Here is the practical rule underneath all four. Give yourself a box: an hour, say, on a work day, with a staging copy if the change is anything more than content. If the hour runs out and you are no closer, you have learned the useful thing, which is that this one is not what you thought it was. Owners who hold that line stay out of trouble. The trouble comes from hour five, when quitting feels like a waste, and the site is already three changes deep.
The exception is when you want to learn the thing rather than just get it done. That is a fair goal and a different activity, and it belongs on a staging copy on a quiet afternoon, not on the live site at four o’clock on a Friday.
Where Modus Medium sits on this
Modus Medium is a principal-led studio in Philadelphia, and a fair amount of the work is drawing this line with people rather than taking the whole site away from them. Some clients want the keys and a short handover so they can run their own content and updates. Some want the routine checks covered on a care plan and a person to call when something stops. Both are fine, and the handover and the training are part of the work either way, because a client who can run their own site is the point rather than a lost opportunity. The services page sets out what is included at each level.
One specific next step. Make a short list of the three things on your site you have been meaning to change and have not, then write beside each one how you would undo it if it went wrong. The ones where you can answer are yours this week. Bring the ones where you cannot to the online business audit on the services page, and we will sort out which actually need a developer and which only need showing once. If it is a single question rather than a list, contact is quicker.