How online advertising actually works (the version no one explains).

When you load a page on the internet, the ad slots on that page are auctioned off in roughly the time it takes for the page to render. Every advertiser who is targeting an audience that includes you gets evaluated. The platform picks a winner, the ad appears, and money changes hands. This happens billions of times per day across the open ad networks.

The auction is the foundation of nearly every online ad you have ever seen, including search ads, social feed ads, display banners, video ads, and most ads inside apps. The mechanics differ slightly by platform, but the basic shape is the same: an auction runs, a winner is picked, and the winner pays.

Most online ad auctions use second-price logic. You set a maximum bid (the most you are willing to pay), but the platform charges you just slightly more than the second-highest bidder needed to win. If you bid $5 and the next bidder was at $3.20, you might pay $3.21, not $5.

This is the same mechanism eBay uses and most ad platforms inherited it for the same reason: it encourages bidders to bid their true willingness to pay rather than play games. Google has shifted some of its auctions toward first-price recently, but the principle of "you are not always paying your max" still holds across most of the ecosystem.

The practical implication: bidding higher does not necessarily mean paying more, up to a point. It increases your chance of winning. The actual price you pay is set by competitive pressure, not by you.

Every major platform also evaluates the relevance and quality of your ad and its destination page, then bakes that score into the auction. On Google, this is called Quality Score. Meta calls it Relevance Score. The names differ; the function is the same. Google's own help documentation on Quality Score spells out the three factors that feed it: expected click-through rate, ad relevance, and landing page experience.

A higher relevance score lets you win auctions at lower bids than your competitors. A lower one means you have to bid more to win the same impression. This is why you can have two campaigns running side by side targeting the same audience and one costs three times more per click than the other. The cheaper one has better creative, tighter targeting, and a landing page that matches the ad.

The implication for owners: the single highest-leverage adjustment in most ad accounts is not the bid, the targeting, or the budget. It is the creative quality and landing page match. Improvements there propagate into the auction directly and lower what you pay per result.

Online ad targeting works because the platforms build profiles of users from a combination of sources: data they collect directly (you logged into Facebook, you watched a YouTube video), cookies that track your behavior across the web, mobile device IDs that follow you between apps, IP addresses that suggest your location, and contextual signals about what page you are on right now.

The infrastructure was substantially weaker after Apple's iOS 14 update in 2021, which let users opt out of cross-app tracking by default. Roughly 75 percent of iPhone users did. The result was a major degradation in deterministic targeting and conversion attribution, particularly for Meta. The platforms have largely adapted by leaning harder on modeled signals (statistical inference about behavior) and contextual targeting (where you are reading rather than who you are).

The next wave of change is the deprecation of third-party cookies, which Google has delayed several times but appears to be implementing in stages. Forrester's ongoing coverage of the cookie deprecation tracks the state of post-cookie targeting and the various replacement technologies the industry is testing.

Attribution is the question of which ad gets credit for which sale. In a perfect world you would know exactly which touchpoint caused the buyer to convert. In the real world, a customer might see a Meta ad, click a Google Search ad three days later, get a remarketing display ad the next week, and finally convert from a direct visit a month after that. Who gets credit?

The platforms each have their own attribution windows and models, which means Meta will report the sale, Google will report the sale, and your Shopify dashboard will say neither of them did. If you add up the reported conversions from every platform, you will often see two or three times the actual sales number.

The honest fix is to anchor on first-party data (what your own analytics and CRM tell you) and use platform-reported numbers as directional only. Mature operators run periodic incrementality tests: turn off a channel for two weeks and see what happens to total sales. The answer is often surprising. Some channels are worth far more than the dashboards credit them with. Others are worth nothing at all and the dashboards have been lying for a year.

The cost of a click on every major platform has risen 10 to 20 percent per year for most of the last decade. Three forces drive this:

• Auction inflation. More advertisers competing for the same finite inventory pushes prices up. This is a market doing what markets do.
• Audience saturation. The high-intent audiences are well-mapped now. The marginal new advertiser is paying to reach the same people the existing advertisers have been reaching for years.
• Tracking degradation. Post-iOS-14, the platforms are less efficient at matching ads to converters, which means more impressions are wasted, which the platforms recoup by raising effective CPMs.

The implication: the same ad spend buys you less performance every year unless something else changes. The owners who beat the trend are the ones investing in better creative, better landing pages, and (most importantly) higher customer lifetime value, which lets them sustain higher CPAs without breaking the math.

The honest single biggest predictor of whether a business can win on paid advertising is customer lifetime value. If your average customer is worth $200 once and never returns, your maximum sustainable CPA is roughly $60 to $80 in most categories. If your average customer is worth $1,800 over three years, your maximum sustainable CPA is $600 to $900. The same ad in the same auction is a winner for one business and a money-loser for the other.

The unfortunate truth is that most owners don't actually know their LTV. They know their average order value, which is a much weaker number. McKinsey's work on customer lifetime value remains one of the better published frames for thinking about this properly, particularly for businesses with subscription or repeat-purchase economics.

If you are reading this and thinking "I don't actually know my LTV," that is the homework. Calculate it before you spend another dollar on ads. The number will either give you permission to scale or stop you from continuing a losing game.

The inside-baseball picture changes how you should think about your account. Bid less obsessively, optimize creative more. Stop treating platform dashboards as truth and build first-party measurement. Calculate LTV before you decide what CPA you can afford. Test new platforms small and stay disciplined about killing them when they don't work.

And most importantly: assume costs will keep rising. The only durable response is to be better at the creative, the landing page, and the customer experience than the next advertiser in the auction. Every other lever is temporary.