Led design for Datadog’s Integrations and Marketplace platform, enabling thousands of users to seamlessly connect external services, import their data, and empowering partners to showcase and sell their offerings directly to customers.

Problem

Integrations are the onboarding journey and the entry point into Datadog. They are how a customer connects multiple data sources so they can use the rest of the platform, and, through Marketplace, how a third-party developer reaches Datadog's install base. The Integrations page carried a decade of technical debt, and that same foundation was the one Marketplace was initially going to be built on. The debt showed up as business cost, not just user friction.

The most used integration, AWS, was losing money on setup. With over 100,000 active AWS accounts, the legacy integration had outgrown the architecture it was built on. Setup was error-prone enough that one recurring account-creation error alone was costing Datadog $8M a year. Users could not see the full picture of what data was being collected per account, so they could not tell exactly what they were paying for. Managing accounts at that cardinality was painful, with no way to search, filter, or act in bulk. Rigid patterns made features hard to configure or scale.

Partners had no real surface to sell on. The original integrations experience was built on a modal design that capped how a partner could present a product. There was no rich media, no markdown control, no way to display the asset details, OAuth requirements, or release notes a user needs to trust a new integration before installing it. Integrations were hard to search or browse, so good ones stayed invisible. And partners had almost no insight into how their listings performed after install.

The two problems looked different from the outside, one an internal management pain and one a missing revenue surface. Underneath, they were the same problem: an architecture that could not carry what integrations had become. That framing is what made a reusable fix possible instead of two separate cleanups.

Approach and key decisions

The through-line of the work was to fix the foundation once and reuse it, rather than pay down the same debt per integration.

Rebuild on a modern SPA rather than patch the legacy stack. The debt was the constraint, so both surfaces were rebuilt from the ground up on a single-page-application framework instead of extending the old architecture.

Tradeoff: rewriting a system running 100,000+ live accounts is slow and high-risk. That is why the AWS redesign leaned on extensive user interviews, usability testing, and QA across Product, Engineering, and Customer Success, to de-risk a rewrite of something customers depended on daily.

Solve it once as a shared pattern. The management surface and the partner storefront were built to ride the same full-page architecture, so the pattern proven on one carried to the next.

Tradeoff: one pattern serving two different users, the customer managing accounts and the partner selling a product, risks fitting neither cleanly.

Attack the $8M error at its root. Rather than validate around the recurring setup error, the setup flow was rebuilt on CloudFormation, with AWS Organization Master Accounts to manage large fleets.

Tradeoff: a CloudFormation-driven flow adds an AWS-native dependency, and it was validated as the way customers actually work. Setting the AWS integration up correctly through CloudFormation is what enables proper installation and metric collection.

Make data collection legible. The AWS redesign surfaced what was being collected per account and why, with real error feedback, plus search and filtering built for high-cardinality environments and bulk account actions.

Tradeoff: more transparency means more surface to design and maintain, and more that can confuse a new user if it is shown badly.

Move trust into the Marketplace listing. OAuth requirements, installation assets, and release notes were surfaced directly in the listing, so a user could evaluate an integration before installing it, and partners got a tile system and rich marketing components to present with.

Tradeoff: the more a partner controls their own listing, the more listing quality varies. Giving partners markdown and media editing without developer intervention is a bet that partner autonomy is worth the loss of a locked, consistent template.

Protect quality with contribution guidelines, not gatekeeping. We set contribution guidelines that partners met to publish an app: a clear app description, installation steps, the dependencies it required, image and demo formats for the gallery, and a bar for ease of installation. The guidelines did the quality work at scale so the storefront stayed trustworthy as it grew.

Tradeoff: guidelines set a floor, not a ceiling. They keep bad listings out and consistent ones in, but they cannot make a listing great, and a partner who meets the bar minimally still ships something mediocre.

Impact

The redesigned AWS integration dramatically improved usability and system reliability. It transformed a critical product experience and set a new design and technical standard for every platform integration that followed.

• Faster onboarding. Setup time dropped significantly, saving customers hours of configuration.
• $8M in annual cost eliminated. Retiring the recurring account-creation error removed the cost it had been generating every year.
• Manageable at scale. A materially better management experience for thousands of high-cardinality accounts, with search, filtering, and bulk actions.
• Reusable patterns, validated. Full-page layouts and the SPA architecture proved out on AWS and are now being adopted across all Datadog integrations.

The Marketplace launched with strong adoption and overwhelming interest from developer partners eager to onboard. It established a scalable foundation for growth, positioning Datadog's platform as a viable ecosystem for third-party developers and unlocking a new revenue stream for the company.

• A foundation that compounds. Establishing and validating a reusable full-page design pattern and SPA architecture cut integration development time sharply. The GCP redesign shipped in 4 weeks instead of 6+ months.
• Seeded App Builder. Growing developer demand surfaced a clear opportunity: a simpler, low-code way to build Marketplace applications. That insight became the foundation for App Builder, a new platform built to unlock faster innovation and expand the Marketplace ecosystem.