The deploy went out at 16:40 on a Thursday. Nothing looked wrong. The homepage loaded, the ads kept spending, Ads Manager stayed green. What broke was the checkout for anyone arriving with a tracking string attached, which was every person the ads sent. Somebody caught it on Friday morning, after the account had spent a night buying clicks into an error page.

The tell was sitting in the reporting the whole time. Nobody was reading for it.

Short answer: Pause ads when your website is down, but pause the campaigns pointed at the broken URL, not the account. On Meta the pause itself carries a price: ad sets paused for seven days or more can re-enter the learning phase. For an outage measured in hours that price is nothing, and the wasted spend is real.

The takeaways

A destination outage has a signature: everything drops at the same minute. Creative decay is gradual and hits one ad. A dead URL takes every campaign, every ad set and both platforms together, while CTR and CPM sit exactly where they were.
The conversion rate is not low, it is absent. Zero purchases against four hundred clicks reads differently from a weak conversion rate, and it is the only number in Ads Manager with any view of your server.
Point the uptime monitor at the exact URL in the ad. A redirect can answer 200 on your bare domain while the version carrying the UTM parameters fails, so a homepage check stays green through the entire outage.

How do you tell a broken website from a broken ad?

The shape of the drop gives it away. A creative problem decays: one ad slides over several days while its neighbours hold. A destination problem is a cliff, and it takes everything at once, at the same minute, across campaigns that share nothing except where they send people. Auction metrics stay untouched, because the auction never finds out your site is broken. CTR is normal. CPM is normal. Click volume is normal.

That last distinction is the useful one. A conversion rate that fell from 3.1% to 1.4% is a marketing argument. A conversion rate of zero against four hundred clicks is plumbing, and no amount of creative work fixes plumbing. When you see the cliff, open a live ad's destination URL yourself, on mobile data, in a private window, before you touch a single budget.

What does pausing actually cost on Meta?

Less than most buyers fear, as long as the pause is short. Meta's own documentation puts the learning phase exit at roughly 50 optimization events per ad set per week, and names seven days of pause as the point where an ad set can re-enter that phase. A four-hour pause during an outage resets nothing. It costs you delivery you would have bought into a broken page.

The expensive version is the panic pause. Somebody switches a campaign off on Thursday evening, the weekend happens, and it comes back the following Wednesday. Now you have paid for a full reset over an outage that lasted ninety minutes. Write the pause down with a time on it, and make whoever paused responsible for turning it back on.

What if you leave Google Ads running instead?

Google finds out before you do. Its destination requirements policy disapproves ads whose landing page does not work, and "Destination not working" is a listed disapproval reason. That outlives the outage: your site comes back in an hour, the ad still needs a review cycle you do not control.

Meta will not disapprove you for a dead landing page, so on that side the damage is money and signal. You buy clicks that bounce, then hand the optimizer an evening of evidence that creatives which were fine are not converting.

Given the choice, pause Google first. It is the platform where waiting compounds into something you cannot undo by clicking a button.

Which URL should the uptime monitor check?

The exact one in the ad, tracking parameters and all. This is the detail that lets outages run for hours with a monitor sitting there reporting green. A check pointed at yourdomain.com asks the web server whether it can answer. It says nothing about whether /products/blue-hoodie?utm_source=facebook&utm_campaign=... resolves, whether the redirect chain still terminates, or whether the button behind it throws.

Copy the destination URL out of one live ad per campaign and paste that into the monitor. Add the checkout or the lead form as a second check: a shop can serve product pages beautifully while payment is dead. A synthetic transaction that walks add-to-cart beats ten homepage pings. If a link shortener or a click tracker sits in front of your site, monitor that too.

What can an ads rule catch on its own?

The money side, and only the money side. An ad tool reads ad performance. It cannot see your web server, your CDN or your checkout, so the half of this guardrail that watches the website is yours to build and no ads automation replaces it.

What the ad side catches is the consequence. Adscalr's rules fire on 8 metrics (CPI, CTR, hook rate, hold rate, ROAS, spend, frequency, CPM), and a dead destination shows up there as spend accumulating against a ROAS of zero. Actions are pause or kill only, every evaluation is logged, and a kill stays reversible for 30 minutes. Alerts dispatch every 5 minutes across 11 event types, which is how the Thursday version of this story ends at 17:05 instead of Friday at nine. Full-auto is opt-in. The default is a recommendation you approve.

The first ten minutes

Pause the Google campaigns pointing at the broken URL first, then the Meta ad sets, and put a reminder on both. Write down the minute it started and the minute it ended. Then unpause the same day if you possibly can.

That written window contaminates every test read you make for the next fortnight. An outage is one bad evening if you mark it and two bad weeks if you do not.

Adscalr does not monitor websites and will never tell you your checkout is down. What the automation side does is make the money half automatic: a rule on spend and ROAS that pauses before the evening runs out, with the alert attached.

If the drop turns out to sit on the ad platform rather than your own server, the Meta outage playbook covers that case. And when nothing is down anywhere and performance still fell, the four-layer drop diagnosis is the order worth checking things in.

Should you pause ads when your website is down?