Supermetric’s Martti Vuopio on building an accidental career as a Martech product manager

My academic background is in industrial engineering and management. After high school, I was undecided about my career path, considering medical, business, or technical schools. Industrial engineering seemed to combine engineering and business, and it was also the most challenging program to get into in technical school, which made it feel prestigious. Initially, I thought I would pursue management consulting, following the path of many of my friends who joined large firms like BCG or McKinsey. Consulting also similarly combined many different industries, again postponing having to decide on just one choice. However, during my studies, the startup culture began to grow in Helsinki, mirroring trends elsewhere. I decided to join and became an early employee at Smartly.io.

When I joined Smartly.io, we were about 50 employees. The company already had a robust business, growing and profitable, providing ad management and optimization software for Facebook ads. Facebook’s own tools at the time were somewhat lacking, so Smartly.io built a superior tool that attracted many advanced, demanding advertisers. I started as an account manager, which involved a mix of sales and customer success management. My role was often about debugging customer problems and finding workarounds using existing features, even if they weren’t explicitly designed for what the customer needed. For instance, for eBay in San Francisco, we developed a custom Python script to process their website feeds into Facebook catalogs because there was no native solution for handling 200 different feeds.

Although I initially aimed for sales and the business side, my background as a “computer nerd” who excelled in physics and technical subjects in school led me to become increasingly interested in solving these technical challenges and using code to address real-world problems. While working on customer issues, I grew more fascinated with the software product itself and how it could be developed. Like many, I believed I had better ideas for features and improvements. After a couple of years in that account manager slash customer solutions engineer position, I made the decision to join Smartly.io’s product team. I was returning to Helsinki after a year in San Francisco, and the product management team, then only two people, was recruiting. It sounded very cool, prestigious, and interesting. At that point, the company was still quite small, under 200 employees, and everyone knew each other, so there was no formal interview process for me. During my first two years, I had become an industry expert, even giving talks on Facebook ad measurement and optimization. They were looking for a product manager for the optimization team, and since I already knew the topic and subject matter, I suppose they didn’t need to interview me.

When I joined the product team, we already had two data scientists who had developed algorithms for things like pausing ads or reallocating budgets between Facebook campaigns. My role was to package these features, improve their market positioning, and prioritise what to build next. Everyone was asking for more optimization features, often tailored to their specific methods. My focus was primarily on prioritisation and market research. A significant part of my job was also bridging the communication gap between the data scientists and customer wishes, as they often spoke very different languages. I had to understand both worlds.

Data scientists relied heavily on statistical methods. For example, if Campaign A had 10 conversions and Campaign B had 20, data scientists would argue that this wasn’t enough evidence to make a definitive decision. Customers, however, would see a clear difference, perceiving Campaign B as “doubly better” and wanting the algorithm to immediately allocate all budget there. This meant a lot of explaining about statistical significance and optimal campaign optimization. In the early days of Facebook ads, there were many “hacks” that advertisers discovered to trick or bypass Facebook’s optimization algorithms to achieve higher ad delivery. Clients often wanted us to implement these hack features. We faced a tough decision: do we implement these short-lived hacks, or do we maintain a good relationship with Meta (Facebook) and align our goals with their product vision?

Ultimately, we decided that these hacks were usually short-lived, so we chose to focus on the long-term strategy. Time has shown this was the right decision, as Facebook and Meta ads have increasingly moved towards an “auto-pilot” model where advertisers provide assets and goals, and Meta handles everything else, reducing the need for micro-management. Long-term, channels aim to simplify things for advertisers, even if clever advertisers initially find ways around new features.

One of my most impactful projects was the budget scheduling feature. Before this, Smartly.io’s tools were heavily performance-driven, assuming advertisers always wanted to spend as much as possible if their goals were met, or optimally spend a daily budget for maximum conversions. However, in reality, agencies, for example, had specific use cases where a precise budget for a defined period was paramount, with performance being secondary. Their contracts often stipulated exact spending amounts, and they had a percentage markup on the spend, so hitting the exact budget was crucial.

It was almost surprising that Facebook lacked robust features for automating spending across an entire campaign portfolio. For instance, if you set a lifetime budget and later needed to adjust it due to another campaign performing better, you’d have to manually change multiple lifetime budgets. When a period, like a quarter or month, ended, campaigns would stop at midnight. Someone would have to be online, even on a Sunday morning, to re-enable them. Our budget scheduling feature automated all of this. You could define multiple future periods, and we handled everything: adjusting budgets, optimising spend between campaigns, keeping them running, and ensuring the correct amount of money was spent. This made our agency clients very happy, highlighting how Smartly.io was able to compete against the free native tools of Facebook by catering to the larger advertisers with their niche problems.

Unrelated to budget scheduling, I recall some genuinely difficult customers. One in particular was strategically important to Smartly.io, generating significant revenue, but their operational methods were completely contrary to what we believed smart advertisers should do. They were heavily invested in micro-managing. For example, they had rules running on Smartly’s trigger system that checked ad performance based on Cost Per Acquisition (CPA) every ten minutes. This CPA metric is highly volatile when checked at the ad level every ten minutes; a single conversion can drastically alter it. Our data scientists tried to explain that such data was not statistically significant and that they should wait for more data and let Facebook’s optimization algorithms work.

In the end, we kept this customer happy by implementing more features tailored to their specific style. The key was knowing when to bend and when to hold the line. It highlighted a common challenge in SaaS: sometimes, meeting the unique demands of a very large client, even if it deviates from your core product strategy, is necessary to retain them.

After seven and a half years at Smartly.io, I decided to move to another startup, Shook Digital, co-founded by a Smartly.io alumnus. About six or seven other Smartly.io alumni also worked there. Shook Digital operated in a similar industry, providing a platform for collaborating on, planning, briefing, and producing video ads, initially for TikTok. The main idea was that advertisers, at the time, struggled with creating TikTok video ads, as it differs significantly from polished Instagram ads; TikTok content needs to look user-generated and blend naturally into the feed. We focused on creator-led content, collaborating with creators, and automating as many parts of that process as possible.

Traditionally, creating creator-led ad content involves contacting numerous independent freelance creators, which quickly becomes tedious, managing communication, ensuring adherence to guidelines, and meeting technical specifications. Our platform centralised everything: briefing, scripting videos, instructing creators, and providing a central place for them to upload their videos. On top of this, we implemented numerous automated video generation, stitching, and editing features, along with automated integrations to ad platforms.

I joined Shook Digital in 2023, the year of the massive AI boom with the launch of ChatGPT and advancements in image and video generation models. It was fascinating to explore how we could integrate AI. For example, instead of hiring creators for video variations, we could shoot one video and use AI to lip-sync multiple AI-generated voiceovers. In the future, entire videos could be generated from scratch by AI. We used AI for scriptwriting and generated captions and subtitles. My experience was heavily focused on how to productise these AI features. The internet is full of point solutions for things like voiceovers or lip-syncing, so the challenge was integrating these into a process that large-scale advertisers could easily use.

I was the first product hire at Shook Digital, joining as Head of Product without any product managers to manage. We had a small development team, and the experience was incredibly refreshing compared to Smartly.io, which had grown to 800 people. Shook Digital was around 20 employees. The speed was astonishing. Whenever I had a product or feature idea, I’d discuss it with an engineer, and by the next morning, it was often already implemented. Working with a small, agile team and a platform that wasn’t yet bloated or complex was genuinely fun and allowed us to move very fast. At that stage, there’s indeed a constant pressure to move quickly; if you’re not fast, it simply won’t happen.

Unfortunately, the company eventually (temporarily) ran out of money. We weren’t moving fast enough to find product-market fit. The company started with an agency model, with a team dedicated to making videos and contacting creators. Later, they hired a development and product team to build the platform and automation, and the website was rebranded to appear 100% SaaS. However, it was challenging to break old habits. The internal team creating videos didn’t easily adopt the platform, and customers were accustomed to contacting a person for small changes and tweaks, rather than relying on AI and a more standardised video creation process. I was temporarily laid off for a while and started pondering my next move.

I applied to Supermetrics when I saw a job opening, and it was actually the first job I ever applied to and got. All my previous roles came through connections and networks. During the interviews at Supermetrics, they strongly emphasised customer centricity and staying close to the customer. Their leading principle in product management was to empathise with the customer, not just understanding what they do, but also how they feel, and how to genuinely make them happy. This deeply resonated with me because it mirrored the way of working I really enjoyed at Smartly.io.

At Supermetrics, this customer-centric approach convinced me it was a great place to work. Additionally, Anssi, who I used to work with at Smartly.io, became Supermetrics’ CEO a couple of years ago. Anssi was the COO at Smartly.io, and he is a very firm leader who makes expectations incredibly clear. He’s also an excellent culture champion, always reminding people about company values and encouraging them to take care of their personal lives and health. My perception of Supermetrics as a successful, youthful, and fast-moving tech SaaS company in Helsinki, complete with a great brand image and fun parties at their headquarters with a rooftop sauna, also played a part.

Traditionally, Supermetrics has been purely a pipeline tool, moving data from point A to point B without storing or even deeply analysing it. The challenge is that our users, typically data engineers and analysts, ideally set up a pipeline once and forget about it because it works perfectly. We are now expanding our market and offering to reach leadership and C-level executives, encouraging them to visit the platform and utilise the value-added features we are building.

We are developing dashboarding features for Supermetrics Hub, transforming it from just a data pipeline into a destination for data. This involves significant improvements to our storage capabilities so we can serve data more effectively. Supermetrics Storage will become a more opinionated storage solution for marketing data, not just storing numbers but intelligently handling online marketing data to enable fluent reporting across different ad channels and platforms. On top of this, my team is specifically building AI and machine learning capabilities to help you not only see and analyze your data, but to understand what to do about it.

There’s no definitive process for building AI products yet; it’s largely about discovery and exploration. In the past couple of months, I’ve focused on understanding agentic AI, exploring examples of AI analytics, examining what competitors are doing, and studying the latest developments from major AI companies. I’ve also been speaking with AI experts who have tried to build these systems, learning what makes them most likely to succeed. For instance, trying to do too much often leads to failure. The promises and stories of agentic AI on social media like LinkedIn are often far grander than what is currently implementable in real life, so finding that balance is crucial.

This presents a significant challenge. When people hear “AI agent,” they imagine something that can do anything and everything they ask. It’s difficult to narrow down the use cases we want to get right. If you go too narrow, users might ask why it isn’t just a simple UI with buttons instead of a chat interface. We’ve seen opinions suggesting that chat UIs will replace all other UIs, but we’ve also heard the opposite: users performing the same job multiple times a week prefer clicking buttons over chatting. Our product philosophy on this is still being built, as it’s early days.

We’re starting with what is perhaps the easiest or simplest use case: conversational analytics. This will allow users to ask questions about their data, generate dashboards and visualisations, and produce executive summaries of performance and campaigns. We’ve heard from agencies, for example, that this is a core need, as reporting to their advertisers forms a large part of their work.

After that, we plan to build a co-pilot style assistant. The Supermetrics platform has many capabilities, but users need to know they exist and how to use them. This agent will help in setting up data pipelines and performing various configurations. The agent should enable customers to start from a business goal, rather than needing to understand data analytics specifics or the exact column names in each ad platform, or building dashboards themselves.

Beyond that, we anticipate moving into the planning side: predictions, planning future campaigns, and modelling the price elasticity of advertising campaigns. This would allow users to see, for instance, how many conversions they might gain if they double their ad spend. We believe we are well-positioned for this because forecasting the future requires extensive data from multiple sources, and since we already centralise all that data, we can apply advanced modelling on top of it.