Leader Spotlight: Building customer trust through authenticity, with Craig Saldanha

---

Craig Saldanha is Chief Product Officer at Yelp. After getting his MBA from Carnegie Mellon University, he spent nearly a decade at Intel, where he led engineering teams to advance microprocessors. Craig then transitioned to product management at Amazon and spent his first few years developing Kindle. Most recently, before joining Yelp, Craig helped propel Amazon’s Subscribe with Amazon and Prime Video products.

In our conversation, Craig talks about how Yelp’s value proposition is customer trust and how deep, authentic customer reviews draw users to the platform. He discusses how his teams combine the tactical aspects of customer obsession with prioritizing customer feedback to foster trust with users. Craig also shares exciting AI-powered developments in Yelp’s roadmap to transform how it connects consumers with local businesses.

---

Key learnings in developing marketplace products

Balancing the needs of both sides of a marketplace can be very complex. Given your background in this area, could you share best practices, as well as challenges, that you’ve experienced?

I’ve been a PM in this area for a long time. I’ve taken away three key learnings: recognize that the needs and goals of different parts of the ecosystem will vary, know your flywheel and its levers, and have a monetization plan early on.

First, two-sided marketplaces have constituencies with goals that are not always aligned. A great example of this is from when I worked at Kindle. Authors and publishers want to maximize the value they can get from a book, which is typically through pricing. As a consumer, on the other hand, you want to read a book for as cheaply as possible. There’s a fundamental misalignment there. It’s important to recognize that.

The second is knowing your flywheel. To create a self-sustaining two-sided marketplace, you have to achieve product-market fit, which creates scale. If one side of the marketplace is lacking, it’s going to erode the other side of the marketplace. For example, when we acquire new consumers at Yelp, local businesses on the platform get more valuable leads. That draws more businesses to our platform, which, in turn, creates more selection for consumers. Those consumers write more reviews, which brings other consumers to the platform, etc.

Lastly is monetization. In the PM world, many of us focus on growth early on with the expectation that we’ll be able to add monetization later. Sometimes that’s true, but often, it’s not. Most fundamental decisions we make early on in the product life cycle dictate how effectively we can monetize. I always encourage PMs operating in a two-sided marketplace to have a robust plan for monetization, even if they choose not to enact it early on.

You mentioned that the number of reviews was a key factor in being able to scale. Since one side of the marketplace feeds the other, it can be difficult if there are challenges getting one side off the ground. Have you been in situations like that before?

In a previous role, I worked on creating a marketplace for subscription products — everything from HBO and Netflix to gym memberships. The basic idea at the heart of every subscription service is recurring billing. So, we thought that if we removed that friction point, consumers would feel more confident in exploring a wide array of selections, from fashion boxes like Stitch Fix to content like HBO.

In theory, it was a really good idea. A lot of businesses saw the value. We got some big names signed up based on the promise of a large consumer base, but that never scaled. This is because we made a fundamental error in understanding how consumers think about subscriptions.

Consumers think about a subscription as a friction point between them and the value they’re seeking, whether that’s content, a box in the mail, magazines, etc. Nobody wakes up and says, “I want to buy a subscription today.” Because we underestimated that, we could only scale the business side of the platform, so the flywheel never started. Despite our best efforts, we couldn’t achieve product-market fit because we fundamentally underestimated consumer behavior in an area that was key to making the whole thing work.

Rallying around customer trust

Could you talk about your approach to implementing and monitoring feedback loops to get insights from both sides of the marketplace?

At Yelp, our value proposition is customer trust. We have deep, authentic customer reviews that bring folks to the platform. We combine the tactical side of customer obsession by taking every single piece of feedback and using it to improve our product. Those two work nicely together to build customer trust.

On the tactical side, the PM team reviews every piece of customer feedback we receive. A specific PM is assigned to triage each piece of customer feedback, no matter the size, and we audit that regularly. We have a giant database of all of the customer feedback, the PMs assigned to it, and what they decided to do with it. Sometimes it goes into the backlog, sometimes we deem it urgent and implement a new feature, etc. I also audit that database and read the voice of the customer report every Saturday morning. It’s an opportunity for me to really understand the pain or delight that folks feel when they use our product.

The fundamental piece is that we create products with the guidance of core building principles. One of those tenets is customer obsession. We start every new feature with the value it delivers for the customer — whether the customer is a consumer or a business. We determine what the impact and long-term value is to them and to Yelp. We’ll even prioritize features that may not add to our top or bottom line but are purely delightful features for either side of the marketplace.

Do you have an example of a feature that came directly from customer feedback?

Yes — a very recent one. A few weeks ago, we launched our AI chatbot, Yelp Assistant. People have interesting feelings about this. There are two main personas that stuck out to us via feedback. One is early adopters — they embrace AI and want it in every part of their day. For those folks, it was most valuable for us to make the chatbots feel as human as possible.

The second persona group is those who are not adopters of AI and love Yelp because it’s made up of real humans sharing their experiences. They want AI to be completely optional. They said, “I want to choose whether to talk to this bot or choose a different path.” That heavily influenced our design.

That’s why now, Yelp Assistant is a choice. There’s a big button on the site that says “Find a pro with AI” and customers can click to engage with it. Or, they can find a pro on their own, without AI, through our detailed reviews and search and browse capabilities. It’s very clear that if you go down the AI path, you are going to have an AI-driven experience. Once down that path, though, we made sure the interaction still feels human. We did a lot of user research to ensure that this tool still feels like a natural extension of a person.

Driving alignment across worldwide teams

I was hoping you could talk about your work leading cross-functional teams. Are there certain methodologies that you rely on for aligning diverse teams and stakeholders toward a goal?

Because Yelp is a fully remote company, we have a diverse and geographically distributed team. We have teams all over the US, North America, and Europe. While that contributes to a phenomenal talent pool, we think very carefully about how we drive alignment because a lot of the decision-making ends up being distributed, and we favor fast execution.

There are five principles that we follow and that work well together:

1. Leading with context and not control — We make sure that everyone is well aligned on both the goals and the guardrails. Then, we give them a lot of flexibility to meet the goals with just the guardrails in place
2. Making our mental models and decision frameworks explicit — This sounds obvious, but a lot of folks have decision frameworks that are very different, and that seems to be the cause of a lot of unavoidable conflicts. At the start of every new feature or new product, we spend a lot of time debating what our mental models are, how consumers behave and how they’ll use that product, as well as what all of our assumptions are
3. Identifying a single decision-maker — This doesn’t have to be the highest-ranking person in the room. It’s the owner of that product or feature, but it’s very explicit. That way, whenever there’s conflict, we know there’s a decision-maker to resolve it
4. Having well-defined escalation paths — We embrace escalating. It’s not considered a faux pas if you disagree with someone and then speak to their manager. We have a well-established protocol even for sending a note saying, “We’re at an impasse. This is the decision. Here are the two points of view.” Then, we can have a quick resolution, which leads to the fifth principle of disagreeing and committing
5. Rewarding disagreeing and then committing — We view it as every person’s responsibility to the customer to speak up when they think something’s wrong and we really reward that behavior. We want people to disagree with every decision that we’re making if they think that we might be missing something. But, once the decision is made, we expect folks to commit to it, even if they disagree

What about prioritizing all of your product initiatives? Could you walk us through your decision-making process when you have multiple competing priorities?

We have three explicit mental models for prioritizing. First, we are rigid on our mission but flexible about how we get there. This goes back to knowing our goals and having a large portion of our resources dedicated to achieving them. The features and the products that help us get to those goals might change multiple times, but we know that a large portion of the team is focused on getting there.

The second mental model is building features that deliver value for customers on both sides of the marketplace — consumers and local businesses. We value the folks who come to Yelp and want to continue to reward them. We have a portion of our roadmap focused on this short-term delighting of customers, even if it doesn’t impact our top or bottom line. We’re disciplined about how we invest, so even with these short-term features, we’ll do an ROI assessment. We’ll look at how much effort it takes to deliver them, but we know that we’ll have a portion of our roadmap going toward building for customers that exist on our platform today.

Lastly, we build for the customers we haven’t met yet. That’s the long-term lens. These are the more speculative opportunities that we know might not come to fruition, but when we find the right opportunity, we’ll get to meet a whole new batch of customers.

You mentioned being remote and how that’s a driver of success for Yelp. Is that because of the talent pool that opened up for you?

It’s twofold. One is that folks are more productive than they’ve ever been. That’s a function of both the flexibility and the tenure. We get a lot of folks who tell us that they might not be in the workforce if they didn’t have the flexibility that Yelp affords, so we have open access to a lot of talent.

Second, I’ve been able to hire some incredible folks who would not have moved to one of our office locations if we were in person. Many employees and candidates tell us that they remain at Yelp both because they’re passionate about the mission and because they don’t have to move — they can work in a way that feels harmonious with the rest of their lives. It’s that combination of the talent pool and productivity.

Celebrating the learnings that come from failure

We talked about some things that can be hard, like change and disagreement. Along that same vein, failure can be difficult for folks. How do you promote a culture that really embraces failure to drive success?

It is hard. The culture has to be built over time. At Yelp, we value failure for the learnings it brings and reward folks for it. Nobody wants to fail, but when failure happens, the learnings lead to more success than we might have otherwise had. Every new hire that joins the product team is taught our five product tenets, and one of these tenets says, “We sunshine failed bets to celebrate the learnings they provide and the new opportunities they expose.”

In my monthly emails and chats with the team, I focus on a failed experiment and celebrate the detailed learnings that came from it. I want to share my thoughts about why it was a great or productive outcome. We truly value that at all levels of the organization. We don’t conflict the delivery of results with the impact that those results have.

Do you have an example of a failed bet that taught you and your team valuable learning?

One of my favorites has to do with grabbing a quick bite to eat after working late. Previously, Yelp would display the search results first, and you’d then tap on the restaurant you were interested in to find out if it’s closed. I thought we could sort search results by restaurants that are currently open and only display those instead.

Well, it didn’t work. Users did not like it at all, and we could see that in the data. When we dug in, we figured out that a lot of folks who were looking at restaurants at night were actually looking to make reservations for lunch or a weekend plan. It was frustrating for them that the choice they wanted was lower down the page just because it was closed at that exact moment.

This was a meaningful insight into how consumers have different needs and how we can satisfy those needs in different ways. It led to a redesign of how we display hours and other logistics, and how we give the consumer the information they want while making it easy to dismiss the information they don’t.

Elevating Yelp’s product with AI

Lastly, are there other AI initiatives on your roadmap that you’d like to share that Yelp users can look forward to?

I’ll talk about two more that I’m really excited about. One is AI-stitched videos. The strength of Yelp has always been the breadth and depth of its reviews. Consumers write awesome reviews and give us really cool photos and videos. Later this year, we’ll start testing AI-stitched videos at scale where we can feed in all the great reviews, photos, and videos together. We’ve trained an AI model to create a script, match photos and videos to it, and then create this overall, experiential view of what a restaurant is like. It’s super cool.

The second one that we recently announced is our Yelp Fusion AI API. We’re taking Yelp’s great user-generated content and making it available everywhere. For example, Expedia just announced that they’ve integrated with our API, and that lets folks who’ve booked a trip on Expedia ask, “What’s a good place in Charlotte, North Carolina to get brunch on a Sunday morning that’s open at 10 a.m. and is kid-friendly?” Expedia will pass that to us and we’ll parse that natural language query and essentially provide a list of great restaurants that match the query.

Yelp has always been a great way for folks to connect, but now we’re taking our content and exposing it across the web and in people’s homes, integrating it into their technologies, etc.