Methodology
Every click has a value
Advertising is a brand’s strongest growth lever precisely because we can calculate value per ad. On Amazon, we pay per click, and we can calculate Value per click like so:
The hard part is boiling this value down to the most granular level and building a system to continuously recalculate value to calibrate spend.
Grounded in your true margins
A click is worth what it returns after every cost that stands between the sale and the profit. Revenue is not profit. A return on ad spend target treats every dollar of sales as equally valuable, regardless of what the product cost to source, what Amazon collects in fees, and what comes back as returns. Contribution margin is the number that makes a click worth bidding on or not — the cost of goods, the platform fees, the return rate, subtracted from the price the customer pays. A bid grounded in revenue overpays for products whose margin is thin and underpays for products whose margin is strong. A bid grounded in contribution margin prices the click against what the sale actually leaves behind.
Conversion rates, statistically modeled
A conversion rate is not a fixed number. It is a range — a distribution of plausible rates centered on the evidence a term and placement have accumulated. When a search term has thousands of clicks, the range narrows; the rate is known with confidence. When a term is new or lightly trafficked, the range stays wide, and the bid carries a risk discount proportional to that uncertainty, so an unproven term is never priced as though its rate were already settled. A term with little history of its own borrows strength from similar terms — structurally related searches whose conversion patterns inform the rate before the new term has earned its own. The model holds a distribution for every term, at every placement, and prices each click against the rate that distribution supports.
For thousands of search terms
A single bid applied across many search terms is calibrated to the average of that pool. The terms above the average are worth more than the bid pays — the ad underbuys them, leaving profitable impressions on the table. The terms below the average are worth less than the bid pays — the ad overpays for them, spending against clicks that do not return their cost. The average is a smoothing that hides the very dispersion it is meant to price. Each search term carries its own conversion rate, its own order value, its own margin against cost. A bid matched to the term prices the click against what that specific term is worth, not against what a group of terms is worth on average.
At each placement
The dispersion deepens within a single search term. The same keyword converts at a different rate in the top-of-search slot than it does in the rest of the search results, and differently again on a product detail page. A bid is one number; the placements it buys are three distinct markets, each with its own conversion rate and its own click cost. A bid set to the term-level average pays the same for a click that is likely to convert and a click that is far less likely to. The placement worth its own conversion rate and margin deserves a price matched to that placement, not a price averaged across all of them. One bid cannot price what three placements are separately worth.
A system that sings
This is how the methodology becomes your ads running themselves. One campaign per product, per keyword cluster — exact match on every search term in that cluster. Each search gets its own bid, calibrated to that search's exact value and normalized to each placement's performance, repriced every hour as the evidence moves. The result: no wasted spend, maximal profit capture, and growth you can measure.
Trust your data
Get started free