How to win by doing things that don't scale

Any why you shouldn't optimize for efficiency in every single thing you do

👋 Hi, it’s Torsten. Every week I share actionable advice to help you grow your career and business, based on operating experience at companies like Uber, Meta and Rippling.

P.S. If you enjoyed the post, consider sharing it with a friend or colleague 📩. If you didn’t like it, feel free to send me hate mail.

---

In 2013, Paul Graham, co-founder of the startup accelerator Y Combinator, wrote one of his most famous essays called “Do things that don’t scale”.

The basic idea is that while software companies are valuable because they are scaleable, you sometimes counterintuitively need to do things that explicitly don’t scale to make your company take off. One of the examples he gives is of Airbnb going door-to-door to recruit hosts for the platform.

In his words:

“Actually startups take off because the founders make them take off. There may be a handful that just grew by themselves, but usually it takes some sort of push to get them going. A good metaphor would be the cranks that car engines had before they got electric starters. Once the engine was going, it would keep going, but there was a separate and laborious process to get it going.”

His advice is specifically focused on getting the startup going and building initial traction. And while that’s the context in which “doing things that don’t scale” has become famous, I’d argue that it’s a winning strategy even well beyond the early days of a startup.

In this post, I will make the case that doing non-scalable things in almost any area of the business, and at any company stage, can create a competitive advantage (and, conversely, that pursuing efficiency at all costs sometimes destroys the very thing that makes a company great).

Specifically, I will cover how doing things that don’t scale can help you:

• Move faster

• Win local markets

• Delight customers

• Maintain excellence

Let’s dive in.

---

Doing things that don’t scale to move faster

Let me reiterate the claim I made in the intro:

It’s a common misconception that “doing things that don’t scale” is something only small startups can or should do.

In reality, non-scalable tactics are at the heart of many hyper-growth stories.

Take Uber as an example. Ridesharing, and later food delivery, have always been highly competitive markets, so the main focus in the first 5 - 10 years of Uber’s history was on growth and gaining market share.

And since the company was growing faster than the Product team could build the necessary features, a lot of this was enabled by labor-intensive, scrappy solutions built by Operations teams.

For example, when I worked at UberEats, we were struggling to make certain restaurants reliable because they were in remote areas with fewer couriers; we knew we had to price dispatches individually to ensure they get accepted, but that functionality didn’t exist.

If we had waited for the Product team to build this feature, it would have meant months of terrible customer experiences, leading to user and restaurant churn.

So we did what any good Ops team would do: Create a hacky (non-scalable) process to plug the gap.

Get the duct tape out and let’s MacGyver this problem.

1. We pulled lists of unreliable restaurants via SQL

2. We used Uber’s geofence tool to draw small circles around these restaurants

3. We calculated by how much we needed to increase the dispatch offers

4. We manually added special incentives just for those tiny geofences

In the beginning, the whole thing was manual. After doing this once or twice, one Ops manager wrote a Python script to automate drawing the geofences based on a CSV with coordinates, and we also started using a Python script to configure the incentives in batches.

Of course, even with these partial automations, it wasn’t a sustainable solution.

But this approach had two benefits:

1. We were able to solve the problem within a few days, enabling further order growth

2. Product hated the manual bandaid solution, so they worked hard to quickly replace it with a productized version

We kept repeating this cycle: 1) Run into a new problem, 2) develop an Ops-driven prototype solution, 3) extract learnings, 4) automate and productize based on these learnings.

Uber, of course, isn’t the only company doing non-scalable things to fuel hyper-growth.

• 💼 Deel was one of the fastest companies to hit $100M in ARR. They rapidly expanded to dozens of markets with their Employer of Record (EOR) offering by first starting with Ops-heavy prototypes and then productizing over time (e.g. building out native payroll capabilities in the product).

• 🏠 During the first years of rapid growth, Airbnb still handled plenty of core processes manually rather than in the product. For example, changes to existing bookings were handled via email and implemented by support reps. While not sustainable in the long-term, it allowed them to continue their hyper-growth and focus product resources on more critical features.

So how do these inefficient processes help companies win?

The simple answer is “speed”. Product and Engineering resources are limited, and waiting to build a fully-functional automated solution slows down growth and expansion efforts.

Besides moving fast, it also helps teams deeply understand the problem and how to solve all variants and edge cases well.

To be clear: You will have to automate eventually when the manual solution becomes a bottleneck; but doing this too early will put you at a disadvantage compared to competitors that are willing to do whatever it takes to grow.

“If you don’t solve problems manually first, you could lose crucial insights into what’s causing a problem and why. Moreover, it may not be the best use of your time early on. We knew that our cancellation flow was flawed, for example. But we also knew winning Europe was more important.”

- Jonathan Golden, then Director of Product at Airbnb (via Medium)

---

---

Doing things that don’t scale to win local markets

Grab vs. Uber

Although Uber was willing to lean on labor-intensive efforts from Operations teams to accelerate growth, there was one area where the company did prioritize scalability: The product experience Uber rolled out in each country was initially pretty much identical (and US-centric).

When Uber expanded to Asian markets, it initially only offered rides by car (even though other modes of transportation were very common locally), and it required payment by credit card even though many people didn’t have one.

This created an opportunity for a competitor to offer a much more localized experience.

Enter: Grab.

Grab customized the experience in each market in multiple areas based on local habits and requirements:

• 🏍️ Modes of transportation: Grab offered GrabBike in Indonesia, GrabTukTuk in Thailand, GrabRemorque in Cambodia, GrabTrike in the Philippines etc. depending on what vehicles were dominant in a given market

• 💵 Payments: Grab allowed cash payments from the get-go to make the service more accessible (it took Uber two years in some regions)

• 🤖 Localized algorithms: Grab tweaked its core algorithms like dispatch to account for local requirements (e.g. only odd or even license plates being allowed in certain areas in Jakarta)

• 🤝 Onboarding: Many drivers were unfamiliar with smartphone apps, so Grab offered in-depth trainings in person

Customizing the product for each country is not really a scalable approach. But if it means the difference between winning or losing a market, the effort can be justified.

The key is to be selective about the localization tweaks and standardize everything else:

“We assess whether we should localise a product or service based on some of these considerations: Does the impact we can create justify the effort required? Is localisation needed in order for us to adhere to local regulations? Is the issue that we are solving unique? When we make such exceptions, we make them count.”

- Samir Kumar, Head of Mobility at Grab (via Grab website)

With this hyper localized approach, Grab quickly gained market share from Uber and eventually acquired Uber’s Southeast Asia business in 2018.

Netflix

Another example of a company that successfully turned localization into a competitive advantage is Netflix.

There are two pillars to Netflix’s localized content approach:

1. Producing original content in local markets

2. Making content accessible to viewers in other markets

Both of these processes by themselves are very much not scalable, but together, they created an efficient growth engine that allowed Netflix to win even in fragmented regions like Europe.

Let’s start with content: While it’s a key differentiator, original content production is expensive and time consuming. Netflix is creating original content in at least 28 markets and is estimated to spend more than half of it’s $15B content budget on international movies and shows.

So how do you justify maintaining large local teams and spending tens of millions of dollars on content in markets as small as Denmark (~ 1M subscribers)?

That’s where Netflix’s dubbing and subtitling strategy comes into play.

Netflix dubs content in 34 languages and creates subtitles in even more. And to get high-quality translations and dubs that reflect cultural nuances, Netflix does this manually with professional linguists and voice actors.

These high-quality captions don’t write themselves

While the process is incredibly labor- and time-intensive (e.g. dubbing the Korean show “All of Us Are Dead” in English took more than three months and required 55 voice actors), it makes the content accessible to millions of additional viewers.

For example, the German Netflix original show Dark had 90% of its viewers outside of Germany, making it much easier to generate a return on the production investment.

Is it worth it? It’s hard to tie content directly to subscriber growth, but in terms of the resulting user engagement, the data speaks for itself — the majority of the movies & shows that stayed the longest in Netflix’s global Top 10 are foreign-language productions:

Image from Bloomberg; data from July 2021 to February 2022

---

Doing things that don’t scale to delight customers

Another common misconception is that you only do things that don’t scale because you have to, with the goal of automating them ASAP.

In the examples covered above, the companies were doing non-scalable things out of necessity; e.g. Uber only leaned on Ops solutions temporarily and ultimately productized most things.

But there are cases where the non-scaleable things you’re doing are actually what makes you special, and you’ll want to keep doing them for as long as you can.

Customer service

“Please listen carefully as our menu options have changed…”

Is it, tho?

Customer service is often seen as an unfortunate requirement of doing business; it’s expensive, so many large corporations do everything they can to cut it back to the bare minimum.

But some companies have turned white glove customer service into a sustained competitive moat. Let’s look at two examples:

Example 1: Zappos (E-Commerce)

Zappos’ brand has always been about exceptional customer support.

Where other companies try to make it as difficult as possible to talk to a real person, Zappos put their support phone number front and center on most web pages.

The calls themselves are also drastically different from what most customers are used to. Call centers typically optimize for maximum efficiency by following standardized scripts and trying to keep calls as short as possible.

Zappos doesn’t do any of that. In fact, one employee in Zappos’ customer service team famously had a 10 hour long phone call with a customer. Here’s how he talks about Zappos’ approach to customer satisfaction:

By the time the call ends, the customer should be perfectly satisfied. I should have done something they weren’t expecting - It's that WOW factor Zappos looks for. Also, making sure that I listen to exactly what the customer is calling for and proactively provide solutions. Customer satisfaction means that when the customer hangs up the phone they can't believe what Zappos has done for them and they are beyond satisfied. When this happens, I’m subtly inviting customers to come back again and again.

- Steven Weinstein, Zappos support employee (via ICMI)

This kind of customer interaction is not really scaleable, but it creates memorable moments and boosts loyalty. About 75% of Zappos’ sales come from repeat customers, much higher than the average for e-commerce stores.

Example 2: Looker (SaaS)

Anyone who has ever introduced a new BI tool in their company knows how difficult it can be to get everything set up and achieve broad adoption across teams.

And it often feels like you’re on your own; the customer support teams of most large vendors are unresponsive and typically don’t provide hands-on help.

Looker took a different approach with its “Department of Customer Love”:

“I had no qualms about sinking a disproportionate amount of resources here — if you make customers successful, they’ll use your product forever. […] It’s a retention strategy, not a cost center. […]

From the very beginning, if a customer was building the data model and had a question, we’d jump in and help them finish the job. That’s why we knew we had to make the leap from providing pure support to building a true community.”

- Lloyd Tabb, Looker co-founder (via First Round Review)

Unfortunately, after Google acquired Looker, customer support was outsourced to offshore contractors.

Looker’s differentiator: Gone.

Executive support

Some founders and CEOs take customer focus to the next level and answer complaints and support inquiries directly (or at least forward them to the respective leaders, like in the case of Jeff Bezos’ infamous “?” emails).

Take Rippling, the last company I worked at, as an example. Parker, the co-founder and CEO, regularly engages directly with customers via Twitter, Email etc. to solve their problems:

• Is this scalable? No

• Do interactions like this leave a lasting impression and create free word of mouth advertising? Yes

---

Doing things that don’t scale to maintain excellence

As a company grows, it becomes harder and harder to maintain a high bar for everything you do.

One way to combat this is to set an example from the top. Over the years, I have seen founders and execs do two things repeatedly that had a big impact on how teams operated and the quality of work they shipped:

1. Dogfooding

If you don’t use your own product, how can you empathize with your users?

When I worked at UberEats, all ICs, managers and execs would regularly do deliveries themselves. This made sure we all felt the pain points that our couriers had first hand and gave us an opportunity to talk to our restaurant partners in a real-life setting.

Rippling is taking it one step further: Parker, the CEO, acts as the company’s main payroll & HRIS admin. This is not a PR stunt; he actually runs many of the company’s key admin processes in the tool himself (like rejecting my expense reports 😔).

If a workflow is inefficient or confusing, or the UI is unresponsive, he shares the pain with other users and has a personal interest in getting it fixed.

“I’ve got this backlog of administrative work to do often, but each time I do it, there are frequently Slack pings that are going out to individual product and engineering teams about, ‘Hey, this didn’t work quite right, or this was slower than it needed to be, or the experience wasn’t clear’. That sort of drives a lot of iteration on the product side.”

Parker Conrad, co-founder & CEO Rippling (via TechCrunch Found podcast)

But the most extreme example of executive dogfooding that I ever witnessed happened at Meta.

In order to better understand the quality of games on Facebook, the VP in charge of gaming initiatives (who happened to be the co-founder of a game studio in a prior life) rolled up his sleeves and played hundreds of Facebook games over the course of a week of vacation and documented his experience.

This got us insights we wouldn’t have been able to get from analyzing data, and it gave a push to the rest of the team to spend more time using the product.

2. Going deep

With the recent “founder mode” debate, the idea of founders getting involved in the nitty-gritty details has gotten a lot of attention — but it’s not a new concept.

Mark Zuckerberg has been doing this for a long time. He is known at Meta for picking topics he deeply cares about and going deep into the weeds, debating nuances of how features should work or weighing in on design choices like colors and button shapes.

In an interview on Lenny’s Podcast, Meta CTO Andrew Bosworth described it like this:

“When Mark has determined that the thing you’re working on is the most important thing, there is no detail too small for him to notice. Like, he will be in a review, and in the same review will be like: “Strategically, I think we’re off course” and also “This one pixel is definitely wrong. You have to fix that.””

- Andrew Bosworth, Meta CTO (via Lenny’s Podcast)

It’s clearly not scalable: You can’t be involved in the details of everything that’s going on in the company.

But if you believe that at any point in time there are one or two things that are the most critical to the company’s success, it’s only logical to spend a disproportionate amount of time on these select topics (and delegate the rest).

---

Closing thoughts

“Doing things that don’t scale” isn’t just for early-stage startups.

Counter-intuitively, it can help you move faster even if you have hundreds of employees or tens of millions in revenue. And it’s not always a crutch you want to get rid of as soon as possible: For some companies, non-scaleable tactics have turned into a lasting competitive moat.

With increased adoption of AI, hands-on, high-touch human activities will become a more and more valuable differentiator. Don’t sacrifice what’s making your company unique just because you think you should streamline and automate every single thing.

---