If an agency has thirty client sites and checks them by hand, the arithmetic takes you somewhere uncomfortable. Opening thirty sites five days a week is six hundred page loads a month. None of those six hundred is recorded, none is compared against anything, and none produces evidence.
The result: if the person doing the checking did not look at a site that day, nothing that happened on it gets noticed. The client notices.
This post is about setting up monitoring across a portfolio: what actually needs watching, what a scalable approach looks like, and the criteria that matter when choosing a tool.
What "monitoring" actually means
For most agencies, monitoring means opening a site occasionally to check whether it still works. That is a visit, not a monitor.
A visit is a single observation. It tells you the page loaded. That is all it tells you.
Monitoring is the difference between two moments. It was like this yesterday, it is like this today, this is what changed.
The practical consequence: the question that matters is not "is the site up". It is "what changed" — and "the site is up" is the answer to a question nobody was asking.
What actually breaks on client sites
From years of watching the same failures:
Layout drift. A CMS update moves a blog excerpt, a hero image height, or a menu order. The site works. It just no longer looks the way you designed.
Broken conversion flows. The form renders, fields fill in, the submit button does nothing. Visitors read that as "this site is broken" and write to a competitor. Your status page can still say the site is healthy.
Price and copy errors. A number on the pricing page is wrong, "no credit card required" has disappeared, a phone number changed. Visible, but unnoticed.
Broken images and links. A banner fails to load, a PDF will not download, a phone number is no longer clickable.
Server-side problems. One page becomes slow, a form times out, an error page is served.
What these have in common: none of them is "the site is down". They all happen on a working site, somewhere invisible. Nobody catches them by eye, because nobody opens the pages you did not click.
A method that scales
Manual checking does not fail loudly. It gets quietly abandoned. The method that scales has three parts:
1. Define what to watch, per site. For each client, agree on the critical pages: home, pricing, contact, signup, checkout steps. This list should be agreed with the client — not what you think is critical, but what they think is critical. The gap between those two is the gap between monitoring and notification fatigue.
2. Compare, do not observe. The same page has to be compared across two moments: what changed has to be computed. This is visual regression testing performed from the outside — the same comparison, except you hold the reference image and the comparison runs on its own.
3. Send notifications through one channel. If every change sends its own email, nobody opens them three months later. Filter first, then notify. This is the subject of alert fatigue, and it matters even more at portfolio scale: if you cannot read the notifications from thirty sites, your monitoring is not running.
White label: the part that actually differentiates
Sending a client a report under your own brand is not the same as sending one under the tool's brand. A report that says "detected by Crawlens" transfers the value; one with your own logo is your service's output.
For agencies this matters twice over:
- The client sees that they bought something from you, not from a vendor.
- The report becomes an artefact of your service — useful in a price conversation.
In practice: the report must not carry the tool's name. It must carry yours. Ask any vendor whether they support this; many do not, and those that do usually put it on a paid tier.
For an agency, that single feature can decide the comparison, because it is the difference between a cost and a deliverable.
What to check when choosing a tool
Is the site limit real? "Unlimited sites" usually means the number of sites is unlimited, and something else is not: scan frequency, evidence retention, or pages per site. If your portfolio is forty sites, ask what happens to sites 26 through 40.
How much history is kept? When a client says "it was like this last month", the answer comes from your archive. Some tools keep a few days. At that length you cannot win the conversation, because you cannot show the evidence.
Is the report white label? See above — and check whether it is on the tier you are actually buying.
Can notifications be filtered? A tool that notifies on every change will be switched off within two months. Filtering is not optional.
Is the evidence presentable to a client? "A change was detected" is not an explanation. You need a screenshot and a sentence describing what changed.
What do you hand the client? The most valuable output for an agency is the moment the client can say "I did not do that". That requires history.
Where to start
A perfect setup beats no setup. So:
- Start with your three largest clients. Not the whole portfolio. On these you will learn what is genuinely critical.
- Five pages each. Home, pricing, contact, signup, contact form. More is noise.
- Change nothing for two weeks. This is the window you need to see what the tool really reports. Intervene early and you will never find out which notifications are worth anything.
- Then show a client. The first report you send should be one you are confident in. This is the hardest part of the work and it does more for sales than any pitch.
Plans and pricing are on the pricing page; setup steps are in the technical documentation.
Read next
-
What is visual regression testing? — how the comparison actually works
-
What is alert fatigue? — why notifications stop being read
-
The business cost of a visual bug — the price of a break nobody sees