IDU
AD BUDGET WAS LEAKING TO BOTS, NOT BUYERS
IDU's Google Ads account looked healthy — spending steadily, logging conversions every week. The problem: most of those conversions were bots. This is the three-phase reset that rebuilt the account on clean, human signal.
AN ACCOUNT THAT LOOKED ALIVE
IDU weren't happy with their paid media, so they asked us to take a proper look. The audit was blunt: the account was bleeding money — spending steadily, logging conversions every week, but optimising for bots, not buyers. Its conversions had been built on unverified, automated form-fills, so Google's algorithm had spent a long time learning to chase exactly the wrong audience. By the time we opened it up, over 90% of the budget was going to bot traffic. (Bots aren't a fringe problem: automated traffic overtook humans online in 2024 and now makes up more than half of all web traffic, with financial services among the most-targeted sectors — Imperva Bad Bot Report, 2025.) Three things had let it happen: the website's native tracking was a black box that couldn't tell a human from a script, the keyword list had bloated past 500 low-intent terms, and the targeting was wide open to regions that sent traffic but never sent a buyer. We ran the whole account through our own agentic paid-media software — the deep analysis no one could do by hand at that speed — turning what would normally be days of forensic digging into decisions we could make in hours.
MOST ACCOUNTS OPTIMISE FOR THE WRONG THING
Google Ads runs on what you tell it to value. Flag something as a conversion and the algorithm will chase more of it — relentlessly, and better every day. That's the engine working as designed. It's also where most accounts quietly come off the rails. Someone set the account up, someone moved on, and somewhere along the line a button click or a page view got marked as a "conversion" — an action with no real intent behind it. The algorithm can't tell the difference; it just optimises toward whatever it's been pointed at. So an account can look busy and healthy while it quietly pours budget into the wrong thing — a low-intent click, or in IDU's case, traffic that wasn't even human. So we didn't start by scaling, or rewriting ads, or chasing a cheaper cost per lead. We started by working out what the account was really chasing — and pointing it back at something real. Clean signal first. Everything else second.
Plenty of accounts can show you a big conversion number. Far fewer can tell you how many of those conversions were human.
We'd rather report the smaller real figure than the bigger hollow one — because the real one is the only one you can build on.
STOP THE BLEED
First, we stopped the account learning from bots. We moved bidding off conversion-optimisation — you can't let the machine chase conversions you don't trust — and demoted every existing conversion goal so the algorithm would stop counting bot form-fills as wins. We locked targeting to people actually present in IDU's markets and shut out the regions that had only ever sent noise. And we paused 239 dead keywords — terms that hadn't earned a single real engagement in two months.
REBUILD THE TRACKING
Then we rebuilt the measurement from the ground up, inside IDU's own Google Tag Manager. What we found there was thin: a handful of tracked button clicks, no clean way to tell a human from a script, and the all-important demo action scattered across so many pages you couldn't tell which one was doing the work. We fixed what actually counted as a conversion — so the ads would optimise toward real actions, demo requests and form submits, not vanity clicks — and built tracking that finally gave visibility across the whole website. Then one simple, ruthless rule: if a form is filled and submitted in under three seconds, it isn't a person, and it doesn't count. Bots are fast. Humans aren't.
REFINE AND RELAUNCH
With clean data finally flowing, we rebuilt the account properly — and the structure was the strategy. Instead of one catch-all campaign, we split searchers by intent into purpose-built ad groups, so each one meets an ad written for exactly where they are. We ring-fenced IDU's own brand searches into a separate campaign, so the main budget chases new demand instead of paying to reach people already looking for IDU. We locked match types so Google couldn't wander off-brief, and leaned the bids hard toward mobile — where every recorded conversion was actually coming from. We fed in IDU's own customer list as a match audience, so the algorithm could learn what a real IDU customer looks like instead of what a bot looks like. And we put the full array of Google Ads assets to work — responsive search ads, sitelinks, callouts, structured snippets — so every ad claimed as much of the results page as Google would give it. Then we relaunched on fresh copy, retiring ad text that hadn't changed in two years.
WHAT'S NEXT
Three phases. Eighteen hours. One clean foundation. We didn't hand IDU a triumphant spike on a chart — we handed them an account that had stopped lying to itself: leads from real people, data built to compound instead of decay, and a measurement layer they could actually make decisions on. If an account is "converting" but the pipeline never shows up, check the conversions before you touch the budget — the fix is rarely a bigger budget or a cleverer bid strategy, it's clean signal. For IDU, the reset moved the account out of recovery and into a position to grow. The scaling chapter comes next.