Leader Spotlight: Designing for marketplace network effects, with Alex Kim
Alex Kim is VP of Product at Saatchi Art, where he leads product, design, data, and engineering for one of the world’s largest online marketplaces for original art. Previously, he was Chief of Product & Operations at Cubic.ai, an AI-powered smart home startup that was later acquired. Alex also founded KIM’S, an online food delivery company that grew to serve thousands of customers monthly before being sold in 2015. Earlier in his career, he worked in product at Belkin, focusing on the launch of connected audio products.
In our conversation, Alex talks about the nuances of scaling a two-sided marketplace, including solving the inherent tension between supply and demand. He also discusses the human side of marketplaces and shares how Saatchi Art approaches complexities like network effects and tradeoffs between buyers and sellers.
The unique power and complexities of marketplaces
You’ve spent a significant part of your career in marketplace environments. At a high level, what makes marketplaces such a uniquely challenging and powerful business model compared to traditional ecommerce?
Marketplace businesses, if done well, drive a lot of value to customers, especially in a world with so much information overload. A good marketplace resolves friction and helps users navigate choices. Marketplaces typically thrive in fragmented industries where there is no dominant player and where there is friction or information asymmetry. For example, if I’m buying expensive artwork, how do I know I will not be scammed? How do I know whether it’s real? How can I be confident it will be delivered without damage? The transaction has risk. Marketplaces can reduce that risk and friction.
In a two-sided marketplace like Saatchi Art, you effectively have two businesses in one. You have demand and supply, and both need to be satisfied for the business to work. Another challenge is that marketplaces generally don’t own inventory. The core assets are liquidity, discovery, trust, and match quality, rather than unique inventory. The real asset is the network effect.
This leads to inherent challenges, especially the cold start problem. You have no demand without supply and no supply without demand. Typically, new marketplaces focus on demand first and “fake” supply — using contractors or internal resources — and then transition to real supply. If done well, all of this creates a moat around the business because these systems are difficult to replicate.
Measuring durable network effects
What leading indicators suggest you’re building durable loops and network effects rather than short-term liquidity spikes?
The idea that “when you have it, you will know” tends to be true. You really need to understand your business and market, but generally, traditional e-commerce metrics still apply — conversion, retention, repeat purchases, and the balance of paid versus organic traffic. There are also marketplace-specific metrics like GMV, supply liquidity, buyer-to-seller ratios, and how your top sellers are trending over time.
The strongest signal is when both sides grow together. The “holy grail” is cross-side growth — when buyers promote the marketplace, sellers promote it, and sometimes users participate on both sides. Eventbrite is a good example — in the same marketplace, you can post an event, and you can attend events. It becomes a closed-loop ecosystem where everyone can contribute to both sides. Our artists do that as well — they shop for art on our platform.
How do you make those tradeoffs without just shifting friction around the system?
Tactically, tradeoffs exist, but long-term, it shouldn’t be a zero-sum game. Both sides need to see value, otherwise the business won’t work. For example, we had issues with inaccurate inventory on our website. A listed artwork may have already sold elsewhere, like at a gallery or through friends or family. A buyer would purchase it on our site and then find out it wasn’t available, which is a terrible experience. You can imagine how much frustration the scenario causes a collector.
So, we introduced stricter listing management rules, which created more work for artists. Some pushed back, but others understood that without buyers, there is no marketplace. This is one example where we had to be more strict on the artist side to benefit collectors. On the flip side, we allow buyers to make offers, but we set a minimum threshold to protect artists from extremely low offers. That limits buyers, but ensures fairness.
A rough model we use is: if a change improves one side by X, how much does it hurt the other side by Y? You try to find the balance. If you’re forced to choose, you prioritize the side that is most constrained. Everyone will know what’s best for their own business. At Saatchi Art, the demand side is definitely more challenging.
There must be so many nuances in searching for artwork on a site. How does that break traditional thinking around discovery, matching, and conversion?
Each artwork is unique, which creates challenges for product discovery and recommendations. Traditional recommendation systems rely on popularity, but if something is popular, it’s already sold. Every new artwork also faces a cold start problem because it lacks historical popularity metrics a traditional SKU would accumulate over time. But even if a recommendation engine is able to identify a popular artwork, purchasing decisions are highly subjective, which presents an additional challenge.
We also see more browsing than searching. It’s difficult to describe an artwork in words, so users prefer to browse and discover rather than search for something specific. Constraints like size or price still matter, but beyond that, browsing is more valuable, as many of our customers say, “I’ll know it when I see it.”
Technicality and complexity in digital marketplaces
Artists aren’t typical sellers. Where do standard marketplace assumptions, such as level of technicality, fail with this audience?
Artists are very creative, but are not often as technical or business-savvy, and those skills are important for successfully selling online. More importantly, they’re not always motivated by sales. Recognition matters a lot in terms of how their work is perceived and seen. For example, all artists want to know how many views their artwork got. They want to know that someone saw what they created.
We see decisions that don’t always make business sense. An artist might cancel a sale because they want to keep the piece, or offer a discount because of an emotional connection with a buyer. We often see artists give discounts just because they want to make people happy. Their motivations are not necessarily financial — they go beyond that.
At the same time, artists who are more business-savvy and understand how to present and sell their work tend to be the most successful. We’re trying to educate them and provide them with tools to be more successful, but there’s only so much we can control in that sense.
You mentioned earlier that Saatchi Art does not hold inventory. How does that add to the unique complexities of the marketplace?
This manifests in a couple of ways. First is listing quality. All listings are user-generated, so listing quality is critical. Images, descriptions, accuracy — all of it affects buyer experience and can lead to returns and friction. On fulfillment, we rely on artists to package and ship correctly. We provide a return policy to remove financial risk for buyers, but there is still time and operational cost.
Logistics are particularly complex. We ship high-value, one-of-a-kind artwork globally from a residential address to another residential address, with packaging handled by non-professionals. You can imagine how many things can potentially go wrong here. Some countries have very strict regulations for art exports. There are many points of failure — packaging, pickup, customs, and delivery. We’ve had orders that included shipping large sculptures internationally, requiring custom crates, multiple transport methods, and special equipment for delivery. Some shipments take months.
This complexity is part of the value we provide. Without a marketplace facilitating it, many of these transactions wouldn’t happen. Our logistics and operations teams are outstanding. And this creates defensibility as well as value, because this complex operational capability is difficult to replicate.
Where AI comes into the arts
There’s a lot of talk today about AI and its impact on creative fields. Where are you cautious about using AI, especially when it comes to trust?
Accuracy and authenticity are the biggest concerns. There’s always a risk of losing the artist’s voice or producing something generic, so we have to be thoughtful about how AI is applied. AI-generated art itself created a lot of fear in the artist community when tools like Midjourney appeared. People thought artists would be replaced, but in many ways, it had the opposite effect — handmade artwork became more valuable and more appreciated.
AI-generated and human-created work can coexist — they serve different use cases and audiences. AI-generated images might work for certain contexts, but for something like a living space, people often want something authentic with a story behind it.
We do allow AI-generated artwork on the platform because the lines can be blurry, especially with photography and mixed media. It’s hard to define where AI starts and stops. Our policy is that artists must disclose what tools they used. As long as buyers know what they’re purchasing, they can decide for themselves. What’s not acceptable is misrepresentation — presenting something as handmade when it’s actually heavily AI-generated.
On the flip side, where are you seeing real value from AI in marketplaces, and particularly at Saatchi Art?
I try to operate from the principle that we should try using AI until it’s proven not to add value. It may sound a bit riskier, but I think there is a lot of potential. With that said, accuracy and authenticity are very important to us, so we are very careful and intentional about how we use it and whether it makes sense for our business and our users.
One example is description enhancement. Authenticity and artist voice are very important, so rather than giving a blank check to an AI tool, we require artists to write their description first. Then, with a strict prompt, we allow AI to enhance it while preserving the artist’s style and tone so it still sounds like them. An AI-written description is better than no description, and it also helps non-native English speakers create strong descriptions and be more competitive. They can still edit it and make it their own, but it levels the playing field.
We also use AI for styling — room visualizations to help buyers see artwork in a space. It can be inaccurate, so we take a conservative approach, focusing on higher quality outputs and making sure we don’t misrepresent scale. It gives us flexibility and cost savings compared to traditional photoshoots. The highest-impact use cases for us are personalization and search — creating user affinity scores and surfacing more relevant artwork.
We also use AI for fraud prevention, listing quality checks, and categorization. We have about 1,500 new artworks uploaded every day, all user-generated, and AI is instrumental in spotting quality issues, flagging spam, and enhancing attributes. AI unlocks capabilities that would be too expensive to do manually at scale.
What does LogRocket do?
LogRocket’s Galileo AI watches user sessions for you and surfaces the technical and usability issues holding back your web and mobile apps. Understand where your users are struggling by trying it for free at LogRocket.com.