The 999 Algorithm Tax Case Study

How It Started

Anyone who has been doing this for a while knows the feeling. You have something real to offer, you have done the work, and you decide it is finally time to put some money behind it. So you go to Meta Ads Manager, you set up a campaign, you let the wizard walk you through it, and somewhere in that process a toggle called Advantage+ is sitting there, quietly turned on by default, and nobody tells you what it actually does to your money.

I knew what I was doing, or at least I thought I did. Fifteen years in performance engineering. I have rebuilt WordPress sites from 28 PageSpeed to 94. I have restructured Google Ads accounts that were burning thousands every month on garbage placements. I know what a Quality Score means and why it costs you money when it is low. But Facebook's algorithm is a different animal, and I made the mistake of trusting it.

The campaign was for vSourceCode. The offer was clear. The creative was sharp. The targeting made sense on paper. I set a budget, hit publish, and watched the numbers come in over the next few days.

What the Dashboard Showed After $999

159 reactions. Impressions in the thousands. Engagement numbers that looked healthy at first glance. And then I checked the one metric that actually matters for a service business: landing page views. Zero. Not low. Not disappointing. Zero. Nearly a thousand dollars and not a single person had clicked through to see what I was actually selling.

This is not a rounding error. This is not a targeting miss where you reach the wrong people and get low engagement. The engagement was there. 159 reactions means people saw the ad and interacted with it. They just never went anywhere. The algorithm collected reactions, the dashboard looked busy, and the money was gone. Cost per landing page view was technically infinite. You cannot do worse than infinite.

What Advantage+ Actually Does to Your Money

Advantage+ is Meta's AI optimisation layer. When it is switched on, you hand control of your audience, your placements, your bidding strategy, and your creative variations to an algorithm whose job is to optimise for the objective you selected. Sounds reasonable. The problem is what it optimises for in practice.

When you tell the algorithm to optimise for engagement, it finds engagement. It is actually very good at this. It will find every person on Facebook and Instagram who double-taps posts for sport, who reacts to everything in their feed, who has been categorised by Meta's data as a high-engagement user. These people react to your ad. They do not read it. They do not click it. They definitely do not buy anything. But the algorithm gets a gold star because engagement went up.

This is not a bug. It is working exactly as designed. The algorithm's definition of success and your definition of success are two completely different things, and nobody at Meta is in a hurry to explain that distinction when you are handing over your budget.

The Algorithm Tax is the money you pay for results the algorithm wanted, not the results you needed. It is the gap between what the dashboard celebrates and what your business actually required. In this case that gap was $999 and every rupee of it was gone before a single potential client saw my landing page.

The Numbers That Tell the Real Story

Algorithm Way (Advantage+ ON): $999 spent. 159 reactions. 0 landing page views. Pure waste.

Engineer's Way (Advantage+ OFF): $282 spent. 811 real reach. Actual results delivered.

Less than a third of the budget. Real reach to real people. The kind of numbers a business can actually work with. The algorithm found 159 people who would react to anything. The engineer found 811 people who might actually care. That difference is not luck. It is calibration.

What I Changed and Why It Worked

Advantage+ off. That was the first and most important change. The moment you turn that off you get control back. You define the audience. You define the placements. You define the bid strategy. The algorithm stops making decisions for you and starts executing the decisions you make for it. That is a fundamentally different relationship.

But switching off Advantage+ is not enough on its own. You also have to actually know what you are doing with the controls you get back. This is where most people get stuck. They turn off the AI, stare at a hundred options, panic slightly, and either turn it back on or start guessing. Guessing is just a slower version of the same problem.

What I did was go back to engineering principles. Audience first. Who is actually going to buy a performance audit? What do they look like in Meta's data? What are their interests, their job titles, their behaviour patterns? Not who the algorithm thinks will react to a post, but who has the business problem I solve. These are different people. Finding the right ones costs less and converts more.

Placement second. Advantage+ will put your ad anywhere Meta can collect an engagement. Feed, Reels, Stories, Audience Network, Messenger. Some of these placements are genuinely useful. Some are filled with accidental taps from people watching videos. The engineer's approach is to pick placements based on where your specific audience actually pays attention, not where the algorithm can most easily harvest a reaction.

Creative third. Not last. Third. Most people put creative first because it is the visible part. The image, the copy, the hook. These matter. But they matter less than showing the right ad to the right person in the right place. A perfect ad shown to the wrong audience still gets zero landing page views. I know this personally.

Why This Matters Beyond Facebook

This story is about Facebook but the same pattern exists everywhere AI has been given budget authority. Google's Performance Max campaigns do the same thing. Smart Bidding does the same thing. Any system where an algorithm controls spending without human calibration will optimise for what it can measure easily, not what your business actually needs.

The reactions in the Facebook campaign were measurable. Landing page views from genuinely interested people were harder to get, required more precision, and took more thinking. The algorithm took the easy path. That is what algorithms do. They are not lazy. They are efficient at the wrong objective.

This is the core of the Algorithm Tax. Every time you hand budget to an AI system without engineering the constraints first, you pay a tax. The algorithm does its job. It just does a different job than the one you needed.

The Lesson I Would Give Anyone Starting Out

Do not let the algorithm drive until you know where you are going. The AI tools in Meta and Google are genuinely powerful when they are given correct signals, real conversion data, and constrained enough that they cannot optimise for the wrong thing. But they need an engineer at the wheel first. They are not a replacement for knowing your audience. They are an amplifier. Amplifying the wrong direction at $1,000 a month is expensive.

Run your first campaigns manually. Learn what your real audience looks like. Build real conversion data before you hand anything to an algorithm. Once you have that foundation, the AI tools can actually help. Without it they just spend your money on reactions from people who will never call you.

And yes, the account was later suspended which is a whole other chapter for another day. The point stands regardless. The $999 was gone before that happened. The algorithm did that part entirely on its own.

What the Technical Tax Audit Covers

When I audit a business's paid advertising setup I am looking for exactly this kind of misalignment. Not just the ad account settings, but the entire chain from ad to landing page to conversion. A slow landing page means even a perfectly calibrated campaign loses half its clicks before anyone sees the offer. Broken conversion tracking means the algorithm is optimising on incomplete data. Overlapping audiences mean you are competing against yourself in your own auction.

The $397 Technical Tax Audit goes through all of it. What the algorithm is actually spending your budget on. What your landing page speed is costing you on every click. Where your conversion tracking is lying to your AI. And what the combined monthly cost of all of it is in real money.

Most clients find the audit pays for itself in the first month just from turning off what should not have been on in the first place.