Adobe Advertising has launched a custom algorithms product, letting advertisers tune the logic that decides how their programmatic budget is spent. Control over the algorithm is powerful for buyers who know their commercial goal. For those who do not, it is a faster way to optimise toward the wrong number.
An algorithm does not know your margin. It knows the target you gave it. Give it the wrong target and it will hit it beautifully.
Adobe Advertising has launched a custom algorithms product, giving programmatic buyers the ability to shape the logic that decides how their media budget gets spent. Instead of handing everything to a black-box optimiser, advertisers can tune the algorithm toward their own definition of success. It is part of Adobe's long effort to build a serious programmatic business inside its marketing cloud.
Control over the algorithm sounds like an obvious win. It can be. It can also be a way to get to the wrong destination faster, because an algorithm optimises exactly what you tell it to, whether or not that thing makes you money.
Why it matters
Most bidding problems are not algorithm problems. They are goal problems. A campaign optimised to cost per click will happily buy cheap clicks that never convert. A campaign optimised to conversions with no margin data will chase sales that lose money. Handing a buyer more control over the algorithm does not fix that. It gives them more power to encode a bad goal.
This is where the commercial thinking has to come before the technical setting. Custom algorithms are a real advantage for a buyer who knows their cost tolerance, their margin and what a good customer is actually worth. For a buyer who does not, the extra control just widens the gap between a campaign that looks efficient and a campaign that is profitable.
The number that matters before you touch a custom algorithm, the real commercial goal it should optimise toward
What to do about it
Define the commercial goal before you tune anything. Profit, margin, customer value. Not a proxy metric that feels close.
Feed the algorithm your real numbers. Cost tolerance and margin, not just conversions, or it optimises toward volume that costs you money.
Do not mistake control for improvement. A custom algorithm pointed at the wrong target performs worse than a default pointed at the right one.
Test against a simple benchmark. If your bespoke setup cannot beat a straightforward one, the complexity is not earning its place.
Keep a human reading the outcome in commercial terms. The algorithm reports on its goal. You have to check the goal was right.
Custom algorithms are a powerful tool for buyers who have done the commercial thinking. For everyone else, more control over the bidding is just a more sophisticated way to optimise toward a number that was never going to grow the business.