What I Learned Designing Ads

After the emotional rant, I also want to share what I’ve learned in the past year working on search ads. I am not trying to defend ads ¹, but rather document what was surprising to me and counterbalance my gut feelings with rational arguments. Hopefully it also helps shed some light on the topic with different perspectives. After all, reality has a surprising amount of details.

Not everyone hates ads

It sounds obvious, but it’s easy to forget our users are not the same as us when designing a consumer product because we are also our users! It’s a double-edge sword: we can leverage our own intuition and experience but at the same time risk relying too much on them. It’s humbling to be reminded that people from all walks of life with vastly different goals and preferences use the product everyday.

One thing for certain is that most people aren’t as morally disgusted by ads as I do. While the public discourse often looks unfavorably at ads, in reality, people don’t care about whether it’s sponsored or organic if the result helps them find what they need², especially when they know what they are looking for and ready to buy. Of course, helpfulness is the key here. Helpful ads are usually relevant and timely. They show people what they might be interested in at the right time.

It’s not an easy task, but search ads are actually in a good position to be helpful. When people are searching about something, there’s an intent behind each search query. Some queries are more commercial, like a brand or a product; others are less so, like looking up a fact or the link to another website. When I search for “new balance 574 grey”, I am more likely to be looking to buy a pair of shoes than when I search for “new balance founded date”³. Thus, showing promoted brands and products for queries with high commercial intent has a better chance of helping people find what they want, which—by no coincidence—also benefits the advertisers.

One key ingredient of Google Search Ads’ success in the 2000s was using Click-Through Rate (CTR) as a proxy for relevance in its Ad Rank algorithm. Ads are not only ranked by how much an advertiser is willing to pay in the auction (bid), but also the likelihood of click. As a result, a lower‑bid, higher‑quality ad can beat a higher‑bid, low‑quality one. This⁴ motivates advertisers to improve ad quality and match user intent. At the same time, ranking by expected value (relevance × bid) also maximizes revenue per query and long-term trust for Google. More relevant ads mean more satisfied users, more searches, and more bidders—compounding value with incentives aligned.

It’s interesting to me that the ability to deliver quality ads was a key part of Google’s business model innovation. Search ads wouldn’t have been successful if all it did was taking users out of the equation and spamming search results with bad ads.

It’s not all about making money

It’s true that ads are revenue-driven, but there’s also nuance in how revenue is measured.

In general, we look at both short-term and long-term metrics. While short-term metrics are readily observable in experiments, long-term metrics are statistical prediction that takes user learning into account. User learning means users’ behavioral change over time in response to a particular change. If a button was changed from blue to green, the model would predict if users will learn to interact with the ad more (ad sightedness) or less (ad blindness) because of the green button.

The implication is twofold. First, it helps prevent short sighted decisions that only optimize for immediate return. Making an ad flash and dance might boost short-term engagement but also make it more intrusive and likely to be skipped, increasing ad blindness and decreasing long-term metrics. On the other hand, it also allows for changes that sacrifice short term gain but benefit the business in the long run, such as prioritizing organic results over ads when they are more helpful.

In a decade-old paper from Google researchers that details this methodology, they provide a case study that proved more ads is not always better. While experiments showed that adding more ads lifted short‑term revenue, long‑term models predicted those gains would erode as users learned to ignore lower‑quality placements. Instead, showing fewer, higher‑quality ads improved user experience and, over time, delivered neutral to positive business impact because higher satisfaction increased sustained interaction with the remaining, higher‑quality ads.

Great, with user value baked into the metric, all we need to do now is maximizing that magical number, right? Not quite. Under that umbrella, there is a lot of leeway to optimize more for one or another, which is why designers still need to advocate for users.

The user-centered lens for ads

So what do designers do to advocate for users exactly?

As I work on the shopping vertical, the first and most straightforward one is solving for users’ shopping needs. That often means organizing product or merchant information in a way that’s easy to parse by adjusting information hierarchy and density, minimizing visual clutter, and maintaining coherence with the rest of the page and broader ecosystem. Metric-wise, we also look at quality clicks (clicks where users don’t quickly click back—typically a signal that a user is interested in the website) as one of the user value signals in experiments.

To understand ads’ usefulness and users’ sentiment on them, we partner with researchers on qualitative and quantitative studies. Depending on what the hypothesis is, these studies can dig into the “why” behind users’ behavior or their overall perception of how “commercialized” the page feels. The research findings will help inform visual treatment decisions or framework level decisions (such as ad load and placement). In these studies, designers and writers adjust different visual and messaging levers in the format to understand if and how they affect people’s understanding and feeling towards the ad.

Finally, designers need to design for trust and transparency. We are responsible for preserving platform’s trustworthiness by ensuring we don’t deceive users for a quick click. This is a delicate balancing act, especially when designing ad formats that are embedded within organic results. There is a tension between visual coherence—making the ad feel like a native, high-quality part of the page—and necessary differentiation. The experience should feel seamless but not at the cost of fair disclosure.

It takes a village—ads and organic

In the video “Why Google Search is Falling Apart”, YouTuber Mrwhosetheboss hypothesizes that the reason why organic search becomes worse is that Google deliberately sabotages organic result to promote ads.

To my knowledge, this is very unlikely. As engineer Logan pointed out from a technical perspective, it’s more accurate to think of ads and organic as two separate feeds with distinct machine learning systems that complement each other. What’s more, on ads side, we don’t just look at ads metrics but also how certain changes affect people’s interaction with the whole page.

If you’ve ever worked at a big company, you know that a poor experience can be structural and not intentional because how a giant organization is composed of differently incentivized sub-organizations that operate rather independently. There is collaboration but they will never be perfectly in sync. Thinking of Google as one monolithic evil corp is giving it too much credit.

It truly takes a village—creating a good ad experience is not on the ads org alone. Organic team also needs to be transparent and collaborative, including ads earlier in the process. If ads are downstream to changes in the organic experience and have to play catch up reactively, what you end up with is an existing ad rashly bolted on organic results due to monetization pressure. On the contrary, an earlier collaboration will give the time and space for ads to navigate the process and adjust the experience to be more cohesive.

Summary

This essay is longer than I expected, so here’s a summary of all the learnings:

1. Average users don’t inherently hate ads. Ads don’t stink if they are helpful.
2. Search ads was a successful business because of the incentive alignment among users, advertisers, and Google. If we were to transition to a privacy-first world, we can’t just think about maximizing users’ interest but also how the incentives of different parties align in this new world.
3. Long-term metrics that take user learning effect into consideration allow for trading short-term loss for long-term gain. Showing less or demoting ads, decisions that benefit users can also benefit the business in the long haul.
4. Ads designers advocate for users by optimizing for information clarity and click quality, working with researchers to understand user preferences and perception, and preserving trust with a balance between coherence and transparency.
5. Disjointed experience is often a byproduct of how large organizations work. Involving each other earlier in the process is critical for platform and feature teams to build a cohesive experience together.

For me personally, the most important lesson might be learning to distinguish what’s subjective from what’s objective, recognizing both my own feelings and the reality as it is, and developing a nuanced view that sees and holds the tension between them. “Be brutally honest with myself” and “be radical and radically open” have always been my tenets. I hope I was successful in upholding them with these essays.