Website maintenance: what it covers and what it should cost

Most small businesses either pay nothing for maintenance and discover the problem during an outage, or pay every month for something nobody can describe. Both are common and both are avoidable with one conversation.

The useful frame is equipment. You would not buy a walk-in and never service it, and you also would not pay a monthly fee to a company that never opens the door.

What is actually in real maintenance?

Four categories. Ask which of them your provider does.

Updates. Platform, plugins, themes, certificates. On WordPress this is genuine, ongoing work with genuine risk, because an update can break a page. On a hosted builder most of it is handled by the platform and you should not be paying much for it.

Backups that have been restored. Every host claims backups. Very few providers have ever tested a restore. A backup nobody has restored is a belief, not a safeguard.

Monitoring that includes the things customers use. Uptime monitoring is table stakes and it is nearly free. The valuable version watches whether your contact form still delivers and whether your conversion tracking still fires, because those break silently and can stay broken for months.

Content edits. A stated number of small changes: hours, staff, a new service, swapping a photo. This is the part owners actually value and the part most agreements are vaguest about.

If the agreement does not name a response time for edits, there is no response time.

What does a silently broken form actually cost?

How do I tell if I am overpaying?

Look at what changed on your site in the last six months, then look at what you paid.

If nothing changed, no edits were requested, and no incidents occurred, a modest monthly fee is fine and is essentially insurance. A large one is not. The question is whether the price scales with the work or with the platform's reputation for being complicated.

The second test is what happens when you ask for something. Send a small edit request and time the reply. A provider who takes three weeks to change your hours is not maintaining anything. Hours are worth more than they look: Whitespark's Local Search Ranking Factors report, published 6 November 2025 with 47 local search experts scoring 187 factors, has "business is open at time of search" entering at fifth overall with 189 points, above review quantity and above every on-page factor. That is expert opinion rather than a controlled test, and Whitespark says so. It is still enough to treat stale hours as an operating problem, and wrong business hours and missing holiday hours have a direct ranking and revenue cost that has nothing to do with your website design.

The third test is whether the fee includes hosting and whether that hosting is any good. A lot of maintenance pricing is really a resold hosting plan with a markup, and knowing what hosting decisions actually matter for a small business site makes it easy to see which half of the invoice is which.

Is there a benchmark for what maintenance should cost?

No, and the absence is the answer.

Every "maintenance should run this much a month" figure in circulation traces to agency pricing pages quoting other agency pricing pages. There is no survey with a disclosed sample, a stated method and a published field date behind any of them, and the same is true of build and hosting ranges. A figure claiming most small business buyers spend under a specific amount is attributed through third-party pages to a 2026 Clutch survey whose underlying report and sample have not been produced.

A benchmark you cannot trace cannot tell you whether you are overpaying, but it can be used to make an unusual invoice look normal. So when a provider says their retainer is in line with industry standard, ask which survey, how many respondents, and what date. A provider quoting their own pricing as their own pricing is being straight with you.

What you can do instead is price the retainer against the four categories at the top of this article and against step four of the calculation above. That is a real comparison and it belongs to your business rather than to an invented average.

What about the monthly performance score?

Treat it as a diagnostic and never as a report card, because the metric underneath it is known to be broken.

Kohavi, Deng, Longbotham and Xu, in a peer reviewed KDD 2014 paper drawing on thousands of controlled experiments at Microsoft, Amazon, Booking.com and LinkedIn, spend part of their speed chapter explaining why page load time is a poor measure. Citing Steve Souders, they note that an Amazon page can render in 2.0 seconds above the fold while the window load event fires at 5.2 seconds, and that Gmail is the inverse, with the load event at 3.3 seconds and above-the-fold content at 4.8. Two pages, two opposite readings, depending on which number you report. Bing's own key timing metric is time to success, not page load.

They also found that where an element sits on the page decides whether milliseconds matter at all. A controlled delay of 250 milliseconds applied to page elements outside the main content, across almost 20 million users, produced no detectable impact on key metrics.

So a maintenance report showing a score out of 100 moving from 71 to 78 is not evidence that anything improved for a customer. Ask instead for the three Core Web Vitals against Google's published thresholds: Largest Contentful Paint good at 2.5 seconds or below, Interaction to Next Paint at 200 milliseconds or below, Cumulative Layout Shift at 0.1 or below, each at the 75th percentile of visits, with all three needing to pass. That is a defined standard rather than a number a tool made up, and it grades the slow quarter of your visitors rather than the fast one somebody tested from an office.

What should I never accept in a maintenance agreement?

Three things, and they all come down to control.

Hosting that only they can access. If the site lives somewhere you cannot log into, the maintenance agreement is also a lock. Ask for the credentials. The answer tells you what kind of relationship this is.

Content you cannot edit. A proprietary content system that only the agency can operate is a business model, not a technical necessity. Standard tools exist and they work.

Termination terms that do not say what you keep. The agreement should name what you get on the way out, in what format, and by when.

None of these are exotic requests and none of them offend a good vendor. They are the same questions worth asking before you sign anything at all, and the questions to ask a web designer before you sign covers the rest of them.

What does maintenance not do?

It does not improve your results. This is where the category oversells itself.

A maintained site is a site that keeps working. It is not a site that gets better. If your inquiries are flat, adding maintenance changes nothing, because patched software has never persuaded anybody to call.

Improvement is a different activity with a different budget. It looks like rewriting the pages that get traffic and do not convert, replacing stock images with real photographs of your work, and fixing whatever single thing is losing the most people. Deciding what to fix first when conversion is bad everywhere is a prioritization exercise, and it does not belong inside a maintenance retainer where it will always lose to routine work.

The one maintenance item that does affect results is content freshness, and it is usually the item nobody performs. New photos of recent work, current staff, current services. Photography for a small business website ages faster than anything else on the site, and a site showing a team that left three years ago is doing quiet damage that no plugin update will repair.

Where does this break down?

Two places.

When the site carries real money. If bookings, ordering or payments run through it, the calculation above is much larger and a serious agreement is cheap by comparison. Do not benchmark a booking system against a brochure site.

When nobody is accountable. Splitting hosting, maintenance and edits across three vendors produces a site where every outage belongs to somebody else. One name, one number, one response time. That is worth paying for even when the technical work is minimal.

What to do this week

Find your current maintenance agreement and read what it actually promises. If you cannot find it, that is the finding.

List what has changed on your site in the past six months. Compare that to what you paid over the same period. Decide whether the number makes sense as insurance, as work, or as neither.

Test the response time. Send one small, real edit request and note how long it takes.

Test the form. Submit your own contact form from your phone and confirm the message arrives where you think it goes. Do the same for your quote form. Then run the six step calculation on what a broken window would have cost you.

Then ask two direct questions in writing: when was the last backup restore tested, and what do I keep if we stop working together. Both answers should be immediate.

Questions about your maintenance invoice?

Email me at eric@seod.com with what you pay each month and what the agreement says you get. I will tell you which line items are real work, which are resold hosting, and which are being charged for something the platform already does. It is a quick reply and it has ended a few unnecessary retainers.

More on ongoing costs in the web design and branding library.