In optimization school, we are explicitly told not to optimize toward Quality Score. We are told to optimize toward conversions, or conversion rate, taking into account cost-per-conversion. But really, we are optimizing toward profit.
“You can only have one KPI for each campaign,” I tell clients, giving them the line I learned in optimization school.* “The algorithm can really only optimize toward one thing at a time. You can optimize toward conversions or engagement time. One or the other. One might affect the other. But you can’t have both and clearly define success.”
It’s a helpful framework for prioritizing: What’s most important to you? Because if you’re going to optimize using a computer, the computer is not really that smart and can only have one true north: optimize towards this one value, taking the other values into consideration, but prioritizing the one. Computers make choices to solve problems, but people define what solutions are the best. People say, “Computer, go toward that best outcome and keep this other good outcome too, but not if it damages that best outcome too much.”
If you’re working with an algorithm, you can optimize for:
Long-term engagement or immediate conversion
Subscriptions or reach
Number of people who receive COVID-19 tests or profits
That doesn’t mean that you can’t have both at the same time. You can! Algorithms take into account many factors, such as the Quality Score metric for Google Ads and the paid Google search algorithm. Quality Score measures contextual relevance to the keyword you’re bidding on. No matter how high you’re bidding, if the content of your ads isn’t relevant, then you can’t win the auction and your ad won’t appear.
But no one optimizes toward Quality Score. In optimization school,* we are explicitly told not to optimize toward Quality Score. We are told to optimize toward conversions, or conversion rate, taking into account cost-per-conversion. But really, we are optimizing toward profit: what combination of factors will make us the most amount of money in the least amount of time with the least amount of cost?
You can have both a good Quality Score and high conversion rates, but an algorithm can’t optimize toward both. You can optimize toward one, then the other. But, to a computer, and to our optimization-driven mindset, only a single destination exists: the one best answer. An algorithm sees the black and white that we’ve directed it to see and directs all its efforts toward reaching that as fast as possible, given the rules provided.
That’s why algorithmic optimization has been so popular. If we give an algorithm a destination, it will find the most efficient way to get there.
We call it Defining Success.
*For most people, optimization school is on-the-job training from some people who have MBAs and other people who have learned from other people who have learned from MBAs or read Harvard Business Review. I did not learn about optimization in my social science graduate school program. I learned about it at my job.
Strategy and the crisis optimization plan
That’s how all algorithmic content recommendation works, and that’s what optimization culture is: You’re optimizing for one outcome over another. You focus and concentrate on one goal at a time. Once one goal is optimized, you move to another goal, not losing sight of maintaining performance of that first metric you optimized.
For paid supporters only. Upgrade to access.
Become a paid supporter to read this content. Supporters can opt in to every issue of the newsletter and always have full access a growing library of essays and instructional guides. Supporters also ensure The Content Technologist's continued publication.