← Back to journal

You Trained Your Algorithm on the Wrong Answer. It Learned Perfectly.

Dream Outcome · JournalFig. YOU-TR

You Trained Your Algorithm on the Wrong Answer. It Learned Perfectly.

Here's a scenario we see every month.

A business spends $5,000 on Google Ads. The dashboard shows 80 leads. Cost per lead: $62.50. The marketing report goes out. Everyone nods.

Then reality shows up. Of those 80 leads, 12 were spam bots. Another 20 were price shoppers who ghosted after the first call. 15 more were "just looking" and never returned a message. The sales team actually closed 4 customers.

That's not a $62.50 cost per lead. That's a $1,250 cost per customer. And nobody in the building knows that number because the dashboard doesn't track it.

Here's what happened: the business told Google's algorithm that a form submission equals success. The algorithm went and found more people who submit forms. It did exactly what it was told. The algorithm isn't broken. The instruction is.

This is the most expensive decision in your marketing, and most businesses make it once, by default, and never revisit it.

What You See Is All There Is (And Your Algorithm Sees Even Less)

Daniel Kahneman coined the term WYSIATI in Thinking, Fast and Slow. It stands for What You See Is All There Is. Our brains build confident stories from incomplete information. We don't notice what's missing. We work with what's in front of us and treat it as the whole picture.

Your Google Ads account has the same problem, except it's worse.

Kahneman's System 1 doesn't pause to ask "what information might be missing." Neither does your bidding algorithm. It optimises based on the signals you give it. If the only signal is "form submitted," that's the entire universe of information it uses to decide who sees your ads, what you pay per click, and when you show up.

Sam Tomlinson puts it bluntly in his analysis of ad account audits: "99% of ad account audits fail because they obsess over tactical levers (CTR, Quality Score) instead of strategic drivers." The fix, he argues, starts with aligning your targets to actual profit profiles. Not blended averages.

The tactical levers are the things most businesses tinker with: ad copy, keywords, bid adjustments. The strategic driver they ignore is the foundational question: what are we actually optimising for?

Think of it this way. You hire a recruiter and tell them "find me people who are good at interviews." They will. You'll get a team of charming communicators who can't do the job. The recruiter did exactly what you asked. You asked for the wrong thing.

Your algorithm is that recruiter. And "form fill" is your job interview.

The Cascade Nobody Audits

The wrong conversion definition doesn't just produce bad leads. It poisons everything downstream. One wrong input, and the entire system optimises toward the wrong outcome with increasing confidence.

What gets affectedHow it goes wrong
Smart BiddingAlgorithm learns to find form-fillers, not buyers
Budget allocationMoney flows to campaigns generating the most forms, not the most revenue
ReportingDashboard shows "success" while revenue stagnates
Team focusMarketing optimises for lead volume. Sales blames lead quality. Nobody changes the input.
Strategic decisions"Google Ads isn't working" becomes the conclusion, when the channel was never given accurate success criteria

This cascade is invisible because every individual metric looks fine. CTR is healthy. Conversion rate is above benchmark. Cost per lead is within target. The dashboard is green.

But the dashboard is measuring the wrong event. We've written before about why your marketing dashboard lies to you. This is the foundational reason. It's not that the numbers are wrong. It's that the numbers are answering the wrong question.

The result is what Tomlinson calls optimising for "blended averages." Your $62.50 cost per lead is a blend of the $15 spam submissions, the $40 tyre-kickers, and the $400 genuine customers. Blended together, they look acceptable. Broken apart, they reveal a business subsidising its own waste.

Meaningful Conversions vs. Mechanical Ones

Dan Ariely ran a famous experiment where participants built Bionicle Lego models. One group had their completed models set aside on a shelf. The other group watched their models get disassembled immediately after completion.

Same pay. Same task. Same Lego pieces.

The "meaningful" group built 11 models on average. The "meaningless" group built 7 before quitting.

Here's the part that matters: the decline was strongest among people who actually enjoyed building Lego. Meaninglessness didn't just reduce output. It killed the motivation of the people who cared the most.

Your sales team is living this experiment. When marketing "conversions" arrive as an undifferentiated stream of unqualified form fills and price shoppers, the people who care most about closing real business get demoralised fastest. They stop following up quickly. They stop treating leads as opportunities. And the business haemorrhages the thing it paid to generate.

A mechanical conversion is a technical event: form submitted, button clicked, page loaded. A meaningful conversion is a business event: qualified prospect identified, appointment booked, proposal requested.

Mechanical ConversionMeaningful Conversion
What it measuresA technical action happenedA business outcome moved forward
Algorithm learns"Find more people who click buttons""Find more people who become customers"
Team responseProcess the leadPrioritise the opportunity
Revenue correlationWeakStrong
ExampleContact form submittedDiscovery call booked and attended

Alex Hormozi's Value Equation frames this from the customer's side: Value = (Dream Outcome x Perceived Likelihood) / (Time Delay x Effort & Sacrifice). Your customer's dream outcome isn't submitting a form. It's solving their problem. If your marketing system is built around form fills instead of problem-solving, you've optimised the denominator (made it easy to submit) while ignoring the numerator entirely.

The most expensive leads aren't the high-cost ones. They're the cheap ones that waste your sales team's time and teach your algorithm to find more of the same.

Close the Loop or Keep Guessing

The fix is straightforward. But it requires effort most businesses aren't willing to invest. You need to tell your algorithm what actually happened after the lead came in.

Google calls this offline conversion imports. You feed back into the system which leads became customers, and what those customers were worth. When you do this, the performance shift is measurable:

These aren't marginal improvements. A 16% lift in qualified leads at the same ad spend is the difference between a profitable channel and a break-even one. And the gains compound. Once the algorithm learns what a real customer looks like, it finds patterns you'd never spot manually. It adjusts bidding to prioritise the audiences, times, and queries that produce actual revenue.

This is what closing the gap between ad spend and revenue looks like in practice. Most businesses assume that gap is a sales problem. Often, it's a signal problem.

And there's a multiplier hiding in plain sight: speed to lead.

Research consistently shows that 78% of customers buy from the first business to respond. Responding within one minute lifts conversion rates by 391% compared to waiting a day. Yet the median B2B team takes 42 hours to respond, and only 7% answer within five minutes.

Hormozi's Value Equation explains why. Time Delay is the denominator. Every minute of wait divides the perceived value of your offer. A qualified lead at 9:01am who gets a callback at 9:03am is a different prospect than the same person called back at 2pm. The intent hasn't changed. The perceived value of engaging with you has.

This is Tomlinson's "optimisation outside the ad account" in its purest form. The highest-leverage change to your marketing performance often has nothing to do with your ads. It's what happens between the click and the customer. And if you're measuring your ads without measuring what comes after, you're seeing half the picture and making decisions with full confidence.

WYSIATI, again.

What This Means for Your Business

If you're running paid advertising and measuring success by form fills alone, you're feeding your algorithm incomplete information and wondering why its recommendations aren't translating to revenue. Here's how to fix it.

Audit your conversion definitions. Open your Google Ads account and look at what's set as a primary conversion action. If it's a generic "form submission" or "page view," you've built your entire optimisation strategy on a proxy that barely correlates with money. Build a conversion stack. Define three tiers: Set up offline conversion imports. Feed your actual sales outcomes back to Google. Which leads became quotes? Which quotes became customers? What were they worth? This transforms your algorithm from a form-fill finder to a customer finder. Measure speed to lead. Time the gap between form submission and first human response. If it's over five minutes, that's your highest-return fix before you touch a single ad setting. Ask the defining question. If your algorithm could see only one number about your business, what number would give it the truest picture of success? That number is your primary conversion. Everything else is noise that should be labelled accordingly.

Your algorithm is brilliant. It will find exactly who you ask it to find. The only question is whether you're asking the right thing.

Further Reading


Dream Outcome is an Australian digital marketing agency helping SMEs grow through Google Ads, Facebook Ads, and Email Marketing.
Book a free call

Ready to grow profitably?

Get a free digital marketing plan tailored to your business. No lock-in, no long commitments.

Book my free call  →