The "you broke my site" problem

Every agency that's ever managed a client site has had this call. You haven't touched the homepage in three weeks. The client insists you've moved their logo, dropped their phone number, or rewritten the hero copy. You explain. They double down. Eventually you load up the Wayback Machine, find nothing useful for the date in question, and either concede the point or stand your ground without proof. Either way, you lose a little trust.

The frustrating part isn't being blamed. It's that you genuinely don't know whether something changed, when, or by whom — and neither does the client. Memory is the worst possible substitute for evidence. So you spend an hour you can't bill, half-apologise to keep the relationship intact, and quietly file the conversation under "things that erode my margin every quarter."

This is the "you broke my site" problem. It's almost universal in agency work, and almost completely solvable. Most agencies just haven't seen the solution framed as a workflow rather than a tool.

It's not really an accusation problem

Modern client websites have more contributors than any one person tracks. On a typical small-business site under agency management, you'll have:

Four to six people with admin access — your team, the client's marketing person, sometimes a content writer, sometimes the client themselves.
Auto-updating plugins (in WordPress, Shopify, Webflow, anything with extensions) that quietly rewrite meta tags, swap out blocks, or break layouts on minor releases.
Theme or framework updates that subtly shift typography, colours, or spacing without anyone touching content.
The client's own staff making edits they forget about within a week.
CDN and caching layers serving slightly different versions to different visitors.

Without a record of what the site looked like at fixed points in time, you're flying blind. The client is too. So when something feels different, the conversation defaults to whoever feels more strongly about it — and the client always feels more strongly about their own site.

The hidden tax

Each "you broke my site" call costs more than the time you spend on it. Count it honestly:

30 minutes investigating the change — checking git, asking your team, pulling up the page.
30 to 90 minutes drafting and sending the explanation, often across multiple emails as the client pushes back.
Sometimes hours rolling back work that was actually fine, because the client is convinced and you'd rather concede than fight.
Trust erosion, slower invoice payments, and elevated churn risk over the following quarter.

Two or three of these per quarter on a single retainer client is a percentage point of margin, gone. Spread across a portfolio of ten clients, that's a meaningful chunk of an account manager's time, and a quiet drag on your reputation that doesn't show up on any dashboard.

Why the Wayback Machine isn't the answer

Most agencies have tried this at some point. Wayback Machine is a remarkable public good. It just isn't built for billable work. For a specific client site on a specific date, it usually:

Crawled when it crawled — sometimes months apart for small business sites — so the date the client cares about is rarely captured.
Skipped authenticated pages, dynamic content, or anything behind staging auth.
Doesn't give you a diff between captures, so even when you have two snapshots you're squinting at them side-by-side.
Doesn't notify you when something changes — you only think to check after the client calls.
Has no SLA when you actually need it.

Agencies need a record on their cadence, of their chosen pages, with diffs and alerts. Wayback is a backstop, not a workflow.

What "having receipts" looks like

Three things in order of value, where each one wins more arguments than the last:

A scheduled archive. Every one to two weeks, every page that matters on each client site, captured as a replayable copy. Fonts, layout, copy, scripts — all of it, browseable like the live site. So when the client says "you changed our homepage on March 14", you click March 14 and the page opens like a time machine.
Automatic diffs. When something does change between captures, compute what's different and surface it in plain English. "The phone number on /contact changed from 020-1234 to 020-5678." So you walk into the call already knowing what they're talking about.
Daily monitoring with triaged alerts. An email when a meaningful change happens — not every CSS class swap, just the ones a human would care about. You spot regressions before the client does. Better still, you forward the alert with one line of context, and now you're the one who told them about it.

The first one alone wins most arguments. The second and third change the dynamic from "defending against accusations" to "running the conversation."

What it actually costs to set up

You don't need a huge ops project. The pattern that works is unglamorous:

Pick the ten to thirty pages on each client site that genuinely matter — homepage, pricing, T&Cs, privacy, contact, top SEO landing pages, top product or service pages.
Configure a tool to auto-capture them every one or two weeks.
Route alerts to a shared inbox or Slack channel that someone on the team scans once a day.
When changes are flagged, decide in three seconds whether they're worth surfacing to the client. Most aren't. The ones that are, forward with one sentence of context.

The tool can be a custom script using Wayback Machine's "Save Page Now" API (free, fragile, no diffs), ArchiveBox self-hosted (free, takes a weekend to set up), or a SaaS that handles all of it (paid, takes about five minutes per site). Pick whichever you'll actually maintain six months from now.

The before-and-after

Before: client claims you changed something on a date neither of you can verify. You spend an hour. The conversation drags across three emails. Either you concede ground you didn't need to concede, or you stand firm and feel them mark a small black mark against you. Trust is a wasting asset and you've just used some.

After: client claims you changed something. You forward two timestamped archive links — one from before they say it changed, one from after — and either show conclusively that you didn't, or show the client themselves signed off on it. Conversation closed in ninety seconds, and the client walks away thinking these people are organised in ways I'm not. Trust banks rather than erodes.

The arithmetic is simple. A handful of pounds a month per site, against an hour of your time and an unmeasurable amount of trust per dispute. Most agencies will pay back the cost on the first call.

One specific recommendation

If you want to test this workflow without committing to it, the cheapest move is to pick one client site — ideally one of your harder ones, where you've had at least one of these calls in the last year — and run an archive on it for thirty days. Don't try to roll this out across your whole portfolio in week one. One site. One month. See whether having the receipts actually shifts the dynamic next time something comes up.

That's exactly what our free 30-day pilot is built for. We set it up for you on one client site, capture everything, and at the end of the month you've either had a moment where the archive saved you or you haven't. If you have, you keep going. If you haven't, you keep the report regardless and we go away.

The thing nobody mentions about the "you broke my site" problem is that it doesn't disappear. The clients who do it once usually do it again. Better to walk into the second call already holding the answer.