Lessons from scaling to 32'000'000 MAU in 2 years - Smallpdf.com

Hyper growth and hyper broken

A preview of my product-led growth cohort which I’m running this year on Maven.com. Gain foundational knowledge on PLG and how to implement it in 2 weeks (The next one 1st of May)

When things go crazy, they go really crazy

I had my fair share of hyper-growth startups with SmallPDF.com (B2C Document Management SaaS platform) being the most recent one. The numbers…

In a matter of 2 years:

• I was employee N° 20 and left when there were 150 of us in 3 locations

• From 2 to over 30mio ARR

• We grew from 14’000’000 monthly active users to 32’000’000 monthly active users and became one of the most visited websites in the world (top 120)

• We had 2 product teams when I joined and about 15 teams in total when I left. (Product, Growth, Business, P&C, support, etc.)

• 1 merger and acquisition to fight a core dependency of the business

• Product scaled into many different sub-silos:

  • core product

  • growth

  • marketing

  • business

  • technical product teams (after an M&A)

3 Learnings

Sh** breaks constantly - Keep it running

Everything broke constantly. Communication lines, processes, and a lot of intermittent drama. If you grow this fast you cannot fight all fires. Not even all of the important ones, you need to fight with intent what cannot break in a product-led growth company:

• R&D is annoying but necessary. Don’t rely on your product sense. It’s worse than you think.

  • Access to data → Our planning was always demanding and we forced ourselves to become proficient in data

  • Access to customers → Weekly interviews per product team

Why do I say that? I know from about 200 experiments that only ~35% of those succeeded. Those are objectively quantifiable experiments that did not meet expectations. None of these experiments was started just randomly.

At some point, a product manager thought that this is a good idea and likely to succeed. We spent time writing it all down before we ran so we had an objective check whether we were right.

Most of the time we weren’t. And that insight is key. Your predictions suck and if you ever let loose of your measurement you will fall back into this fallacy.

It happens to me and it will happen to you

A great book on the topic is “Superforecasting” by Philip Tetlock which deals with our ability to forecast anything.

Knowing your ICP matter

About one year into this adventure we did a thorough ICP (Ideal customer profile) research trying to figure out who is using the platform.

The value of these profiles cannot be overstated and they sound basic if look from far away, let’s look at 2 of them:

• Content Creators: People with complex needs who use the platform seldom but intensely when they do with a wide breadth of tools. Typically these are people who need to prepare presentations/documents and pull together content from sometimes dozens of different sources. Email, presentations, phone calls, slack, etc.

• Operators: People that use very few tools but with a high repetition frequency. For instance, accountants have to do something to statements before the end of the month and so forth

How did this affect operational work? It affects everything. Once you know how you can identify them your perspective on what to do becomes different.

Operators use specific sub-tools (Smallpdf had over 20 at the time) whereas content creators use different ones. Knowing this allows you to optimize their experience:

• ‘Operators’ pay a lot of attention to how fast their flow is. You optimize them for speed and efficiency. (How to process multiple documents etc.)

• ‘Content creators’ pay a lot of attention to how flexible their flow is. They don’t process a lot of documents but the few they do need to be served well.

Countless hours were spent analyzing our massive data stack and knowing how to subsegment the data by these ICP’s sharpening the picture a lot.

So let’s just say you’re believing me that this was worth it. I go as far as to say that it is worth figuring out for every company.

The way we did it was an intense effort through qualitative and quantitative surveys of our user base and then matching this up with data analysis to see whether there are different patterns.

You often don’t know what you’re looking for and segmentations only make sense if these users start to behave differently. What works for a data behemoth like Smallpdf might not work for a sales-led heavy company because they can’t leverage the same amount of data.

Be that as it may, ICPs are important for every company. Whether you are sales-led or product-led.

It allows you to communicate a different value proposition to each segment through marketing, product, and sales. That’s efficiency and the difference between worthwhile startups and one that runs out of money.

Strategy is everything

One of my favorite things is to create strategies. It’s not the process itself because that one is arduous with bringing everyone together and discussing endlessly.

I love the process of simplifying it and bringing it into a seemingly simple form.

When I created the domain strategies for Smallpdf’s core business with all of our product and growth teams we created 2 subdomains (of 4 in total):

• Document creation: Anything that the product might need to have an optimal creation experience went into that.

• Document collaboration: Anything that users might need around a document to collaborate which was essentially adjacent use cases and nothing we dealt with so far.

If you had to rename the two then it came down to: “Core and Adjacent” This played extremely well with already knowing our ICPs and overall better knowledge of our segments.

Strategy creation Kraken @ Smallpdf

While B2C was the core market of Smallpdf it turned out that a ton of people are using the product while being in a micro business (1-20 people). Those are extremely hard to identify and just as Grammarly B2B started to intermix with B2C. But how is this relevant?

It gives you a solid data basis to pivot your business towards team use cases and broader collaboration features. It’s not difficult to imagine why people might want to collaborate around documents but… do you know? And how many?

The difficulty is not finding things that make sense. It is to identify what makes the most sense. At the point of strategy creation, we ended up with 100s of ideas. Defining and simplifying it into a one-pager feels like magic.

It’s what matters in the end. All the other problems we had over the years seem insignificant. I feel like that is one of those things that we got right without a doubt.

My process is always roughly the same, my free mini-strategy course goes into more detail:

What else?

There’s much more since I bounced between leadership and being an operator:

Other learnings I talk about in my growth cohort:

• Contract Negotiations with Sales-Led Companies as a PLG company

• Growth and Product Separation

• Partnerships with Industry Giants like Miro

• And more…