From Brand Queries to Revenue: An Interview with DoorDash Machine Learning Engineer Prashanth Srinivasan

Retail media is moving closer to the moment when shoppers declare what they want. Global retail media ad spend will reach $196.7 billion in 2026, but the next layer of competition is not just more ad inventory. It is whether commerce platforms can handle high intent search moments, including brand related queries, while keeping ads relevant to the user’s search experience.

Prashanth Srinivasan, Senior Machine Learning Engineer at DoorDash, works on ads quality, ranking, and retrieval systems that drive revenue growth and user personalization. His expertise spans machine learning, ads monetization, data engineering, and distributed systems. A speaker at Agentic AI Summit Silicon Valley, he brings a decade of experience across machine learning and backend systems, including prior work at Google on large-scale infrastructure for YouTube Ads and Assistant NLP.

Prashanth, thanks for joining us. Why do brand related search queries matter for ads platforms?

Brand related searches often carry strong commercial intent. When a user searches around a brand, the system is no longer dealing with a vague browsing session. It is dealing with a query that may reflect purchase interest or a user trying to navigate toward a familiar product or merchant.

At DoorDash, the project focused on enabling ads on user search queries that had previously been blocked because they were related to brands. I was the DRI for that work, which made the project a direct test of how ads relevance, search systems, and monetization could come together on a previously unavailable class of queries.

What made the project technically difficult?

The difficulty came from how many systems had to move together. This was not a narrow change inside a single model or service. Enabling ads on branded search queries required end to end investigations and fixes across the whole ad serving stack.

That included search platforms and search systems, ranking and relevance models, ad targeting stacks, budget and pacing, and retrieval. Each layer had a different responsibility, but the user experience depended on all of them behaving correctly once a previously blocked class of queries became available for ads.

What was your role in the branded search query work?

I was the DRI for the project, which meant I was responsible for driving the investigation and execution needed to make the launch work. The role required looking across multiple components rather than treating the problem as a local fix.

The project involved working across organizations and teams, because branded search query enablement touched systems owned by different parts of the ads and search stack. My responsibility was to help connect the technical path from query handling to retrieval, ranking, relevance, targeting, budget and pacing, and final ad serving so the experience could operate as one working system.

Why did user testing and scale matter so much here?

User testing mattered because brand related queries can change how the search experience behaves once a previously blocked query class becomes eligible for ads. Those testing rounds helped the team evaluate how the newly eligible branded queries behaved across the ads experience before the launch path was expanded.

The project required multiple rounds of user testing, along with collaboration across organizations and teams. It also required investigation into many components of the ad stack and support for increased scale. Opening a previously blocked query class is not just a policy or eligibility change. It can change traffic patterns, candidate flow, budget behavior, and the volume that downstream systems need to handle.

How does this connect to the broader growth of retail media search?

Retail media search is becoming a more competitive part of the advertising market because it sits close to purchase intent. EMARKETER reported that almost $5 billion of incremental ad spending will flow into retail media search in 2026, which explains why platforms are putting more attention on query handling, relevance, and measurement.

That context matters for DoorDash Ads. Branded search query enablement was tied to a millions of dollars in incremental revenue impact and was critical to the Ads organization hitting revenue goals. For a business where ads can be an important driver of profitability, opening new high intent query surfaces can have business value only if the system still protects relevance and user experience.

How does relevance engineering support this kind of work?
Relevance engineering keeps monetization connected to what the user is trying to find. In the branded search query project, that meant treating new ad eligibility as only one part of the work. The project also required ranking and relevance models, targeting, budget and pacing, retrieval, user testing, and scale readiness to move together.

That is why branded search query enablement could not be handled as a simple switch. DoorDash had to evaluate how previously blocked brand related queries behaved across the ads experience and make sure the serving path could support them across the ad stack. The business impact came from opening a high intent query surface while keeping the system grounded in relevance.

What should other machine learning teams learn from this project?

The main lesson is that query monetization depends on coordination. A branded search query may look like one line of input, but the decision to serve an ad passes through many layers. Search systems interpret the query. Retrieval decides what can be considered. Ranking and relevance decide what deserves attention. Targeting, budget, and pacing determine whether the ad can compete responsibly.

In the DoorDash project, the work required end to end investigation across those layers, multiple rounds of user testing, and support for scale increases. That is the kind of machine learning and systems work that often matters most in ads. The value is not only in a model decision, but in making the full path work for real users, real merchants, and revenue goals.

Where is search ads infrastructure heading next?

Search remains one of the largest revenue pools in digital advertising, with IAB and PwC reporting that search revenues, including AI search, reached $114.2 billion in 2025. As that market keeps growing, the quality of query handling, relevance decisions, budget systems, and retrieval paths will matter more for platforms that want advertising to remain useful inside search.

For DoorDash Ads, the direction is clear from the branded search query work. New commercial surfaces require more than eligibility changes. For branded search query enablement, that means the same disciplines Prashanth applied at DoorDash: end to end investigation across search systems, ranking and relevance models, targeting, budget and pacing, retrieval, user testing, and scale readiness.