Chief Product Officer (2022 to 2026), promoted from Vice President, Product Strategy
I joined Yottaa in 2022 to help with M&A and a longer-term product view, and I stayed to help reinvent the company. Yottaa had spent a decade as an e-commerce performance pioneer: first as a CDN built specifically for online stores, then as the company that made the browser smarter by intelligently sequencing the hundreds of third-party tags that slow modern storefronts down. A private growth-equity firm had recently invested, with a mandate to move the business off its capital-intensive networking roots and focus on its high-growth software. The opportunity was real. So were the headwinds. My early, private description of our position was a small canoe heading down a fast river crowded with other canoes, except ours was made of paper. I decided the interesting work was to rebuild the boat while it was still moving.
Reading the market before touching the roadmap
Before I proposed changing anything, I wrote down what I actually believed about the market. About six months in, that reflection became a strategy. Several forces were eroding our two oldest value propositions at once. Faster consumer bandwidth and 5G were shrinking the raw page-load problem we had been founded to solve. Every commerce platform was now bundling “good enough” CDN and security. And the biggest quiet threat was architectural: many of our core customers, a large share of our revenue, ran on Salesforce Commerce Cloud and were rebuilding their storefronts as single-page applications on frameworks like Next.js and Vercel. As they did, they shifted from loading vendors through our JavaScript tag to calling vendor APIs directly, for example taking a JSON response from a personalization vendor and rendering it themselves, which quietly removed the surface our optimization depended on. Server-side tag management pointed in the same direction, even though it never reached the wide adoption I predicted. The honest read was that the ground under our flagship product was moving.
That pointed to a longer-term direction I would spend the next two years pursuing: stop competing on yesterday’s terms, and reframe Yottaa around digital experience monitoring and real user monitoring, winning on business-relevant insight rather than feature parity with the observability giants. Anyone could show a developer a waterfall chart. Almost no one was connecting web performance directly to conversion and revenue for the business stakeholders who held the budget. I wanted Yottaa to be the trusted advisor that wrapped a retailer’s commerce platform with performance, protection, and insight, aimed at the middle-market teams who could not stitch best-in-class point solutions together themselves. But a strategy on paper does not sell anything. The first job was to bring order to a product and a price book that had drifted, and that is where I started.
Bringing order: the marketecture
The first thing customers and our own salespeople tripped over was the product itself, or rather its names. We sold a sprawl of vanity-branded SKUs that took half a sales call just to explain. I rebuilt the “marketecture,” collapsing that sprawl into a single clean umbrella, Yottaa Digital Experience Optimization, with simple editions a buyer could understand and a free entry point that opened the funnel. Underneath the packaging I pushed the architecture toward a single JavaScript tag with backend entitlements, so we could unlock and cross-sell capabilities without sending an engineer to change code on a customer’s site. I also did the unglamorous work of standardizing our SKUs, order forms, and contract terms, the foundation any product-led-growth motion needs.
I did not win every internal debate, and that is part of the story. I argued we were past the point of competing in raw CDN and security against Cloudflare, Amazon, and Akamai. The decision at the time was to keep selling those edge services, so I committed to it and drove it like it was my own idea. Debate, decide, commit.
The hard truth about price
Our customers kept telling me something uncomfortable: we had become one of the most expensive line items on their budgets, second only to the commerce platform itself, at exactly the moment post-pandemic traffic was collapsing and our performance impact was shrinking. So I did the thing product leaders are not supposed to do. I argued for a price cut.
I built the case two ways. Cohort analysis and clustering on our own customer data surfaced the pattern that explained the churn: roughly three out of four of our SaaS customers were paying for far more traffic than they actually used, a pricing mechanic that punished them for a downturn they did not cause. Separately, I stood up a deliberately third-party win/loss program, because customers tell a neutral party things they will never tell a vendor’s own sales team, and it confirmed the same story from the buyer’s side. On that evidence I ran two resets, in 2023 and again in 2024, moving us toward usage-aligned pricing, proactive right-sizing of at-risk accounts, and a clean free-to-paid ladder simple enough to publish on a pricing page.
It worked where it mattered. As the new pricing cycled through renewals, company-wide net revenue retention climbed back into the low 80s by the end of 2025 and gross retention into the mid 70s, on a run-rate the business carried into 2026 of roughly 85 percent gross and 90 percent net, expectations rather than full-year actuals. The reset protected the core: across the mid-market and enterprise accounts that made up the large majority of our revenue, we lost a single logo to churn across all of 2025, while the deliberate losses stayed in the small-business segment we had decided we could not serve at a profit.
Building the engine around the product
A good product with no supply chain behind it still loses. We had a large sales force and almost no enablement, so I took over product marketing during a transition and rebuilt the foundation: clear, accurate documentation of what every capability actually did, then sales-ready material on top of it. I brought in a firm I had worked with at Adobe to codify a sales playbook, the “Yottaa way of selling,” complete with an ideal-customer profile, buyer personas, and a whiteboard story reps could actually use.
I leaned hard into partnerships, because our solution often belonged inside someone else’s contract. We became an SAP silver partner and launched Yottaa on the SAP Store to reach the commerce base that platform served. Across the customer base, brands like Discount Tire, Maui Jim, Samsonite, and Signet Jewelers relied on Yottaa to keep their storefronts fast and secure. The partner research also crystallized a strategic call that would shape the next two years: since every commerce platform now bundled a CDN, continuing to sell our own network was a losing position. We needed to let it go.
The bigger partnership bet was Salesforce. We built and launched an offering on their AppExchange and worked to align with the Salesforce Commerce Cloud B2C go-to-market team. The timing worked against us. Salesforce was pushing its Commerce Cloud base toward a newer composable storefront and bundling more performance and CDN capability into the platform itself, and at the same time brands on the legacy platform were increasingly migrating to Shopify, where the pitch was that performance would simply be handled. Many of those joint customers churned from Yottaa. We won some back when they discovered their new storefronts still carried dozens of third-party tags we could tame, but it was a hard slog, and the dedicated Salesforce sales and customer-success team we had engaged was reorganized out from under us. It was a clear lesson in how completely a platform partner’s own strategy can move your business, for better and for worse.
Becoming CPO and making the turn
By late 2023 the case I had been making, that we had a product-market-fit problem, had carried, and I was promoted to Chief Product Officer to act on it. Part of what I had been arguing was that we were trying to do too much: we ran two very different businesses, an edge network bought by infrastructure teams and a web-analytics and optimization product bought by digital and marketing teams, and serving both well was stretching a small company thin. The turn I pushed for, the tack as I preferred to call it, was to end-of-life the Yottaa network, move customers onto best-in-class CDN and security through Fastly while keeping the relationship, and refocus engineering on the software, so those same customers could stay with us for web-performance analytics and for Application Sequencing where it still added value. It also created optionality: a cleaner software company, with the managed-services business able to stand on its own later.
We executed the cutover fast. Our customer and engineering teams moved about three quarters of the accounts running on the Yottaa network onto a Fastly-backed managed service in under six months, and we kept the large majority of that recurring revenue while taking the cost of running our own data centers off the books. Migration-driven churn was close to zero: the accounts that did not move were leaving for unrelated reasons, and the one account we had worried about internally ended up moving and staying.
Reinventing the platform
In 2024 we turned to rebuilding the product itself, and “we” is doing some work in that sentence. A lot of talented people had spent years running that network, and I underestimated how big a shift retiring it would be. It was not until the migration was finished that we fully saw it: the network’s retirement changed the shape of the engineering organization, from deep expertise in networks, CDNs, and web application firewalls toward browser, full-stack, and cloud-data depth. Some people made that leap, new skills had to come in, and the change was harder than the architecture diagram made it look. We migrated to the cloud, wound down our data centers, and pointed R&D at a new Web Performance Cloud built to give cross-functional e-commerce teams visibility, analytics, and the ability to act. I owned the product vision for that reinvention.
The piece I am proudest of is the foundation we built underneath it. We re-architected the data platform as a modern lakehouse, streaming real-user-monitoring data through Kafka into a Databricks and S3 architecture. It cut our infrastructure cost per page view by an estimated 60 to 70 percent, a deliberately conservative figure, and made the platform AI-ready by design rather than as an afterthought. On that foundation we shipped Hybrid RUM, the release that correlated performance across the front end, third-party scripts, the edge, and the origin, and that finally spoke to engineers and business leaders in the same product. That was the strategy from my first month, made real: tie milliseconds to revenue.
None of that foundation was mine alone. It was built by an engineering team that consistently punched above its size, led by a CTO I have built with across earlier chapters of my career, on an integration that later anchored a major commerce acquisition and on the work to rebuild Gap’s content and commerce stack. Talent that joins you from company to company is one of the truest signals in this work, and getting to build alongside people I trust, who then raised the bar for everyone around them, is the part I would do again in a heartbeat.
One caveat about that year. With the platform mid-rebuild and the network being retired, we deliberately idled the new-logo engine and concentrated spend on the rebuild. So 2025 was about proving we could hold the customers who mattered, which we did. Proving the new model could win new ones was the job we were lining up for 2026, with Hybrid RUM as the opening move.
Betting on AI, in the product and in how we worked
I treated AI as both a product direction and an operating method. Internally, after we had run the company lean, my product-marketing team built a Yottaa GPT assistant that cut the research-and-drafting stage of a solution brief from days to minutes and the end-to-end cycle from more than two weeks to a few days. The lesson generalized into a principle I still use: get the ground-truth data right first, and generating everything downstream with an LLM becomes trivial. This very history document is me applying that principle to my own career.
And I do not just direct this work, I build with it, well past chatting with a model in a browser. To get an unsentimental read on where our monitoring really stood against the field, I wrote an automated system, BeaconEval, that forensically grades any RUM vendor’s browser agent against 131 technical criteria, every finding cited to the vendor’s own publicly served browser code and cross-checked across multiple models. I ran Yottaa through it, baselined the audit against our source, defined what “elite” would look like, and turned every gap into a prioritized engineering backlog the team could start closing. It runs on its own now, and it is the kind of competitive and product intelligence that used to take a team and a quarter.
In the product, I pushed for an agentic future before it was fashionable, and we built the foundation for it deliberately. The Web Performance Cloud put AI to work directly on the data: anomaly detection that flags performance regressions automatically, and YoBot, an agentic assistant that answers performance questions in plain English and points to the fix. A third piece, an attribution service to quantify exactly how each third-party vendor moves conversion, is the harder problem and was still in development as I left. It is the connective tissue for the fully closed-loop vision.
Underneath YoBot is a real agentic architecture rather than a chatbot wrapper. A Model Context Protocol server exposes live telemetry, retrieval runs over Yottaa’s own performance data, an orchestration layer decides which tools to call, and a foundation model does the reasoning, with a serverless action layer in build when I left to apply optimizations and feed results back. The fully autonomous loop, where telemetry drives validated changes on its own, was where the platform was headed rather than where it stood the day I left. But the thesis is the part I care about, and I think it is still early: the dashboard era is ending, and the next interface is a customer’s own AI agents querying your data directly. A protocol like MCP is necessary but not sufficient. The harder and more valuable problem is becoming the thing those agents actually depend on, the real-user sensor that tells a coding agent a change hurt someone, and the verifier that proves a fix worked before it ships. Owning both ends of that loop is the position I kept pushing Yottaa toward.
Sharpening the story the market could feel
I also used product and a tuck-in acquisition to make the value undeniable. We acquired SpeedSense, bringing in deep web-performance consulting talent and the Sensai technology, which accelerated our insights product and gave us synthetic data to solve the cold-start problem for new customers. We launched Web Performance Services, a fully managed offering combining Fastly, HUMAN Security, and Yottaa. And we replaced our old static annual benchmark reports with a live, interactive Web Performance Index drawing on more than 500 million shopper sessions across hundreds of brands and storefronts, a living benchmark for an industry raised on yearly PDFs. And the proof showed up where it counts, in customers: across the more than 1,500 storefronts Yottaa optimized, published results ran from site-speed gains north of 30 percent to double-digit conversion lifts, including a 15 percent lift across one retailer’s seven Shopify storefronts. My framing for it became my framing for the whole category: performance is not a pass/fail badge, it is a race.
What I take from it
Yottaa was turnaround work inside one of the harder markets a software company can face, and I am clear-eyed about that. What I am proud of is the discipline we brought to it: reading where the market was actually going and acting on it, making the unpopular but correct calls on price and on retiring a legacy business, rationalizing a portfolio around customer value rather than engineering inertia, and building an AI-native foundation that was genuinely ahead of the field. I also developed a thesis I still believe in, that digital analytics and observability are converging and that the durable winners will be the platforms that are open to AI agents by design.
And when it came time to find Yottaa a new owner, I was hands-on in the sale itself. Working alongside our banker, our advisory firm, and the CEO and CFO, I did a lot of the work people do not expect from a product leader: building the target list and the rationale behind each name, drafting the teasers, owning the pen on the core buyer materials, and pitching in call after call, staying hands-on through my final day to hand the work off in good order.
My tenure ended in May 2026 with that work still in motion, and we carried it a long way: the platform rebuilt on modern economics, the core customers protected, the strategy documented for the team carrying it forward. The market for standalone performance tooling had been consolidating under our feet the whole time, and we ran the reset knowing it. However Yottaa’s next chapter reads, I am proud of the discipline we brought and the platform we built, and I am rooting for the people writing it.
Where I think this is going
If one thread runs from my first job to this one, it is the gap between when software is built and when you learn whether it actually worked for a real person. I was trained on the V-model at Accenture, where we tried to catch every defect before shipping. The industry spent two decades closing that gap through agile, test-driven development, and experimentation, and yet real-user performance data still arrives late, in a dashboard, after the code has moved on. What changed for me was the Model Context Protocol. I did not read it as an AI feature; I read it as middleware, the same decoupling pattern I worked with decades ago, finally cheap enough that real-user signals can become a queryable input while code is still being written. That is the loop I pushed Yottaa to close, and it is why I believe the next frontier is Agent Experience: as AI writes more of the front end, performance becomes the baseline signal that tells an agent whether a change helped or hurt, and the platform that can both detect the regression and verify the fix becomes part of how software gets built. I wrote about that thinking in this essay, and I left Yottaa with a product spec for the pivot.
This chapter is part of My Work, my career told one company at a time.