From Cost Per Click To Cost Per Outcome: How AI Flips Optimization

Sergii Denysenko, CEO at MGID.
Today's advertisers care about one thing: outcomes. As budgets tighten, media fragments and brand loyalty wanes, advertising channels must prove their ability to uplift tangible business metrics. However, optimizing toward outcomes is notoriously challenging, with a huge number of variables affecting whether spending on an ad achieves its desired result. Optimizing for outcomes requires the ability to predict the future—an impossibility, until now.
Cause And Effect? Try Effect And Cause
AI-powered bidding strategies reverse typical programmatic inputs and outputs. Instead of adjusting costs and hoping for conversions, machine learning models begin with the desired conversion and work backward to identify the optimal bidding strategy.
This is made possible through the predictive capabilities of machine learning models. These models ingest and analyze signals from hundreds of sources—from ad placement and local time of day to device type, audience behaviors and contextual relevance—to identify complex patterns of cause and effect that can be used to gauge the likelihood of a conversion.
Whereas traditional cost-per-click (CPC) models treat every impression or click as equal until proven otherwise, AI-powered cost-per-acquisition (CPA) models dynamically weigh the value of each opportunity in real time. If the model determines that a particular user, at a particular moment, is significantly more likely to convert, it will automatically adjust the bid upward, even if that click costs more.
You might worry that such upward revisions would rapidly exhaust budgets, but such models balance out instances of increased spend by, conversely, devaluing or stripping out low-probability impressions. This way, budgets are preserved while efficiency is significantly boosted.
This real-time recalibration allows for automated, high-frequency optimization at a scale that no human team could feasibly manage. Not that humans are—or ever should be—removed entirely from the process. You as the advertiser still remain in control of all the cost and key performance indicator levers that are important to you, while engineers work behind the scenes to monitor the algorithm's performance and correct for edge cases.
After all, no model is 100% accurate in 100% of situations. The end goal of such technological developments is not to remove human oversight and strategy from the process, but to significantly expand the reach and effectiveness of human decisions. AI without people steering its course is like an orchestra without a conductor.
For all the revolutionary technology chugging away in the back end, the user experience is largely unchanged. You're still billed on a CPC basis and can set the same thresholds you're used to, while the underlying engine makes decisions based on conversion potential. It's as much a mindset evolution as a technical one: Trust the model, feed it accurate data and it will reward you with better outcomes and less manual tinkering.
Outcome Optimization Ripples Across The Open Web
Should it achieve its full potential, AI-driven optimization will fundamentally alter the bloated, top-heavy digital advertising ecosystem of recent years.
For one, AI levels the playing field. Advanced algorithmic optimization techniques were the domain of big tech giants such as Google and Meta, which had a massive advantage due to the reams of data exclusive to their walled gardens, derived from platforms meticulously designed to extract as many user signals as possible.
Now, we all can tap into similarly sophisticated predictive technologies across the connected world, diversifying our channel distribution and reach without inflating our budgets. Smaller brands and advertisers without sprawling analytics departments can pursue outcome-based advertising with precision and scalability on whichever channels they choose, rather than default to the path of least resistance offered by big tech.
Over time, this will redefine the concept of value in media buying. Under CPC regimes, cheap inventory could still be an attractive gamble, even if it doesn't convert. With CPA models in control, low-quality placements will wither from lack of spend, clearing a path for high-quality publishers whose audiences are more engaged and conversion-prone. Outcome optimization serves as a filter, reallocating spend away from made-for-advertising (MFA) sites and toward genuinely valuable environments.
There are challenges and limitations, of course. AI models require both sufficient data and time to learn. You must set realistic CPA targets—despite appearances, AI is not magic—and give campaigns room to breathe. Impatient tweaking or overly aggressive benchmarks can 'confuse' the model, leading to misfires or underdelivery. But these are bumps in the road that will eventually be smoothed out. As more campaigns feed these algorithms, the systems will grow more resilient and their benefits will become more widely accessible.
AI's capacity to evolve is perhaps what's most exciting about it. Digital advertising often feels like laying the tracks in front of a speeding train that veers wildly through market fluctuations. With AI's ability to adapt incredibly quickly, advertisers can step back and gain a wider perspective. At a time of rapid and tumultuous change in digital media (ironically, much of it caused by AI itself), it's more important than ever to be able to look to the horizon rather than be bogged down by busywork.
Forbes Agency Council is an invitation-only community for executives in successful public relations, media strategy, creative and advertising agencies. Do I qualify?

Try Our AI Features

Explore what Daily8 AI can do for you:

Learn with AIFact CheckingText to SpeechAI Summary

Related Articles

Bloomberg

28 minutes ago

• Bloomberg

Irish Regulator Warns Against Caving to Pressure on Bank Rules

Ireland's top banking supervisor warned of a need to stick with high standards of financial regulation particularly now when there is increasing pressure to ease them amid trade wars, geopolitical tensions and fears for global growth. 'If we're trying to chase growth with a deregulation engine I don't think that's going to end well for anybody,' Mary-Elizabeth McMunn, Deputy Governor for Financial Regulation at the Central Bank of Ireland, said at Bloomberg's Future of Finance event in Dublin. 'It also can't be that you simplify things just to such a degree that you really fail to capture complex risks.'

Forbes

33 minutes ago

• Forbes

Integrating AI Into Your Small Business

Kathleen Hurley is the founder of Sage Inc., a tech company that offers SMB businesses infrastructure solutions and next-gen technology. Understanding that AI will be part of the equation for a successful business going forward, you may wonder where to begin implementing AI into your small business. It seems like an enormous task, particularly for organizations that are often lean and operating with staff models that are close to their performance requirements. Getting started with AI doesn't need to be a huge undertaking for a smaller firm, but there are considerations before incorporating an AI platform into your day-to-day operations. For instance, ensuring that security is set up correctly so that the AI only sees what it should see is crucial. This is often referred to as the walled garden, reflecting that the AI is given room to roam within its own environment but walled off from things it shouldn't be interacting with. The walls are put up around secured information or data that hasn't been reviewed, allowing the AI to ingest or train on information that has been vetted and approved. Is this important for a small business where data security is often not the top priority? Yes. The data security of a smaller organization is still essential, particularly when AI can make data visible to users if it's not adequately secured. Imagine HR data improperly displayed as the result of an inquiry because it's left inside the walls when you didn't mean for it to be. This is the type of query result you want to control simply by removing the data that could result in an answer you don't intend. When considering your security, you can also think about your overall AI strategy within the organization. Designing your AI strategic plan is a great way to start your journey, even if you aren't entirely sure what your strategy is yet. The First Steps Of An AI Strategy The reason many companies fail to put a strategic plan into place is that they are unsure of the tactics, or platforms, they plan to implement. But that shouldn't stop them! They are still aware of their feelings about bias and ethics, their concerns about spending and platform approvals, the process that data transmission approval should go through and the different approaches that users can take when they want to test out a new platform. Requirements for agents and platforms, such as security and data privacy requirements, can be stated if they are not called out in other vendor platform requirements and policies. Some approaches involve AI innovation teams who can be responsible for the whole engagement in the organization, or there can be a single point of contact for the AI strategy. This is the base of solid, simple rules that the company needs in order to proceed carefully and safely in the AI environment, respecting the company's needs for security and data privacy. Finally, consider testing different applications in various groups within the company. AI note-taking is a great way to start, as it is simple to implement and generally has a very intuitive interface that all users can understand and benefit from. Anyone who attends many meetings can appreciate not having to take notes! ChatGPT and other Gen AI tools are likely in use in many organizations—regardless of size—and implementing a policy will help surface their use cases and enable a firm to understand how they are being utilized. Gen AI tools are great ways for users who are skilled at prompt engineering to take advantage of AI. For those who are not natively comfortable with writing prompts, offering training on prompt engineering is a good way to initiate the process of integrating Gen AI into a small business environment. Harnessing internal resources, such as having a power user on staff or bringing in an outside resource, are both great ways to get started with prompt engineering. Integrated tool packages, such as Copilot, are also excellent solutions for small businesses, as they significantly streamline the implementation process and can help smaller organizations improve their return on investment. It's essential to implement the strategic plan before releasing the package, as well as security measures since even Copilot has the same security pitfalls as other solutions. What Works For You There is no one-size-fits-all solution for small businesses when it comes to AI, and there is no perfect one tool that will do everything perfectly and solve all problems at once. However, there are steps you can take to implement AI that will make you more structurally sound and ready to embrace the new technology, no matter the size of your organization. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

The Verge

37 minutes ago

• The Verge

Apple keeps pulling its own ads

Apple has taken down a new ad just one day after posting it, making it the fourth one removed in just over a year, as spotted earlier by MacRumors. The nearly eight-minute-long ad, titled 'The Parent Presentation,' featured comedian Martin Herlihy giving students advice on how to convince their parents to buy them a Mac. Apple posted the ad on Friday, but it disappeared from YouTube and the company's webpage for college students on Saturday. The iPhone maker also released an accompanying 81-slide presentation template that's supposed to give parents '45 undeniable reasons why a Mac is essential to college,' which still remains available for download on its site. Last May, Apple apologized for its 'Crush!' commercial, which showed a hydraulic press flattening a piano, record player, paint, and other creative tools, only to lift and show its new iPad Pro at the end. It was meant to demonstrate how many creative tasks can be completed with the device, but it sparked widespread backlash instead. Apple pulled the commercial from TV before removing it from YouTube. Months later, Apple pulled a 10-minute ad, called 'Out of Office OOO,' which showed a group of coworkers using Apple products on a business trip in Thailand, after receiving criticism from Thai citizens and lawmakers for portraying the country in a stereotypical and outdated way. Then, in March of this year, Apple took down an iPhone 16 ad with Last of Us star Bella Ramsey. Apple used the ad to show off an AI-upgraded Siri with features that aren't available yet, like recalling the name of someone they met months ago. Unlike the other ads pulled by Apple over the past year, there's no clear reason why 'The Parent Presentation' was taken down — other than some users on social media calling it 'cringe,' or raising questions about who the commercial's target audience was. The Verge reached out to Apple with a request for comment but didn't immediately hear back.