9 lessons from growing paid search for SaaS by 10x (in 3 years)

My first attempts at growing a SaaS for Amazon sellers

Before I started as the first marketer in a young startup with 12 employees, I had only tinkered around with paid search. The company offered a suite of tools for Amazon sellers, with features from PPC optimization, to product research and inventory management. They had just achieved product market fit and my task was to use that momentum to scale by finding and optimizing the right acquisition channels.

Paid search has one big advantage at an early stage: The immediate results lead to a tighter feedback loop and thereby faster learning.

So I started testing and quickly found something that worked, or so I thought. After seeing a spike in conversions, I was excited: Paid search is the perfect channel. I thought scaling was just a matter of doing more of the same from that point on — Adding more keywords and ads, bidding higher to get more impressions, and expanding the target audience.

It didn’t take very long to realize that I was wrong and lost sight of the big picture when optimizing all the little details. When looking at spreadsheets all day, it’s easy to forget that real people, with specific intentions are clicking on those ads. And it’s very easy to fool yourself about the real ROI of your campaigns when there's a lag between signup and actual revenue.

The growth of new trials (red) looked awesome at first

When the increase in free trial signups didn’t result in the expected ARR growth, I couldn't make heads or tails of it. My first response was to look for new tactics, maybe some clever hack will do the trick. It didn’t, or course. We figured it out eventually and grew from $600K ARR to $5M ARR, and later even $10M ARR, but it was a long learning process.

The 9 lessons I wish someone had taught me from the start.

They fit into three categories:

Understanding customers better

• Understand the customer journey to identify leading metrics

• Understand which segments are the most valuable

• Structure accounts based on those journeys and segments

Using data correctly

• Don't get stuck in analysis paralysis by trying to be scientific

• Be selective with what you test

• Don't confuse yourself with attribution model

Anticipating changes early

• Be prepared for sharply increasing competition

• Monetization often limits your paid search ceiling

• Anticipate the channel ceiling

Let's go through them one by one.

Get a deeper understanding of who your customers are

1. Understand the customer journey to identify leading metrics.

To understand the bigger picture, zoom out and look at the whole customer journey instead of just the small fragment between clicking on an ad and signing up for a trial.

What are prospects doing before that, and what actions predict that they will convert to paying customers after the trial? What actions predict that they will stay customers for a long time and have a high LTV?

That’s why it’s essential to track the right leading metrics and optimize for those. In our case for example a free trial signup wasn’t worth much, if users didn’t also integrate their API with the tool. This was the strongest predictor if they were going to become a customer or not. It should have been obvious because users couldn’t get much value out of the tool without doing that. But our main KPI was “new trials”, so we just looked at that everyday and didn’t notice anything else.

Taking the step to activate their Amazon seller central API predicted a new lead’s value

2. Understand which segments are really contributing to the bottom line

Just like specific actions in the customer journey can predict the value of a new user, so can other variables. That’s why it’s important to slice and dice your customer data in many different ways and hunt for patterns to see which new trials are most valuable based on CVR and LTV.

In our case, this was region (e.g. people from the US vs India) and use case (e.g. Amazon ppc optimization, vs Amazon product research). When we looked at the intersection of these segments, we realized that some trial signups are almost 100 times more valuable than others in the long run. That’s because smaller individual differences in ARPA, trial-to-paid CVR and churn compound to an enormous overall difference.

Value of a trial = Trial to customer CVR x (ARPA / Churn)

Segment A: 5% x ($50 / 8%) = $31.50

Segment B: 25% x ($300 / 3%) = $2500

So the ratio of the least valuable to the most valuable segment is 1:79 (!)

In the intersection of this Venn diagram, we found the leads that were 79x more valuable than the least valuable ones

3. Structure accounts based on journey and segments

With a clear picture of our ideal leads and which steps we needed them to take, we had a solid foundation. Instead of trying to “optimize this ad title”, or “adjust the CPC for that ad group” a little bit, we changed the fundamental account structure.

Having a separate account just for the US, and specific campaigns based on use cases made it much easier to see the true ROI that was obscured by looking at “averages” before. Only after having a solid foundation like that does it make sense to think about things like optimizing your impression share and ads.

Using data correctly: Stay skeptical but don't overthink

4. Don’t get stuck in analysis paralysis, by trying to be scientific

Jeff Bezos once said that as a rule of thumb, you should act when you’re 70% sure because otherwise, you’ll be too slow. This is way below the scientific standard of >95% to call it a “significant” result. While the scientist in me cringes at the overconfidence many people have in test results that are barely better than guessing, Jeff has a point there.

At an early-stage startup, you will always lack resources, data, and especially time — because speed is your biggest advantage. So don’t waste that advantage by getting lost in data and trying to do a Ph.D. on paid search.

On the other hand, don’t lull yourself into a false sense of security by pretending that your AB test with 70% likelihood of beating the control is a clear indicator your hypothesis is confirmed. Test, evaluate, use intuition to decide if you have to, but be clear that the decision has been made based on intuition, not data. Then test again.

5. Be selective with what you test

Not having a lot of data makes testing expensive. Not necessarily by ad budget, but because you can only run a limited amount of them to achieve a relevant sample size.

Way too often have I seen people resort to “Let’s just run an AB-test” to resolve a dispute.

This can quickly lead to squandering the little traffic you have on tests that either don’t move the needle even if they’re successful, or don’t have a chance of getting anywhere near significance anyways.

6. Don’t confuse yourself with attribution models

So I just told you that you need to take the whole customer journey into account, and now I’m telling you to ignore attribution models? Sure, check if things look very differently when you compare last-click to first-click for example. But if it doesn’t make an obvious difference, don’t obsess over it.

At this point, Google’s algorithm is advanced enough to figure things out so long as you feed it the right data (Signal to noise). This also means not giving it less relevant data points to not dilute the message. Just be aware that there are a lot of external factors like brand, product, competition, and word of mouth Google has only incomplete data on.

The brand awareness from a display campaign permanently increased the conversion rate of paid search ads… but never showed up in any attribution model

Anticipate changes early

7. Be prepared for sharply increasing competition (especially in an emerging market)

One of the great things about working at a new startup in an emerging industry is that there isn’t much competition… yet. On the one hand you are limited in the volume of leads you get, since not that many people know that there’s a solution to their problem yet. But if people do search for it, they will come straight to you, because only few people will be competing for those precious “above the organic results” ad slots.

It’s tempting to assume things will stay like this, and base your forecasts on today’s CAC. However, as more and more competitors push into the market, this level of optimism might turn out to be a bit naive. In our case, the CPCs increased by a factor of 10 over the years, and for a long time, we assumed that we must be doing something wrong. Maybe it’s our quality score? Nope. It’s just that the early mover advantage was over.

This is how the average CPCs for some of our most important terms developed over time (in EUR)

8. Monetization often limits the paid search ceiling

There is a way to deal with the rising CAC though, and it has nothing to do with optimizing campaigns. When trying to decrease CAC by optimizing campaigns while competition is on the rise, it’s like shoveling out water from a sinking ship with a bucket. Instead, plug the holes: Churn and monetization. While churn rate is an obvious issue for most SaaS companies that's usually owned by the product team, increasing your prices is a growth lever that often gets little attention.

This is either due to the fear of losing customers and a tanking conversion rate, or because it’s just not clear whose responsibility it is. If it’s being tested and iterated upon however, this could be your strongest lever.

9. All good things end: Anticipate the channel ceiling

There are only so many people searching for a solution. You might think you can grow by just adding terms that are more “informational”, but we already found out that this isn’t likely to move the needle much when it comes to ARR and will just drain your budget.

So when you see diminishing returns from your paid search campaigns, there might be nothing wrong with them at all. You might just have found the ceiling to that channel. It’s time to take a step back and re-evaluate the entire growth model. Then you can expand your scope, test and learn about new channels, product-led growth, and so on.

So what now?

If you're working on paid search at an early stage startup, you can use the nine points as a checklist to evaluate where you are in the overall life cycle of paid search as a channel.

Save the article so that if you find yourself getting lost in a sea of contradictory data, obsessing over details, and optimising for siloed KPIs like CTR and trial signups, you can come back to it.

Want me to elaborate on any of the points? Comment below.