On Growth in Startups at Scale

How does the role of Growth function evolve with scale of startup?

Photo by Anastasia Zhenina on Unsplash

Quick Recap

This article is 4th in the ‘On Growth’ series. A quick recap of the series so far:

• In the first article, we covered the objective and outline of the series.
• In the second article, we covered how to quantify Product-Market Fit (PMF).
• In the third article, we covered how to understand different aspects of PMF to establish the customer segment(s) and the perceived value of product before designing a Go-to-market(GTM) strategy to acquire more customers, while also setting foundation for a culture of experimentation.

In this article, we will cover:

• How does the role of Growth function evolve at scale?

Let’s begin.

For the love of jargon

All specialists love complicating things. Everyone seems to enjoy the confused looks a freshly coined terminology gets. People working in Growth are no different: just when you thought you had understood product-market fit (just a fancy way of saying a business is profitable/viable) and could move on, we have now invented the term product-market-channel fit.

Nevertheless, the term is important and is central to understanding the role of Growth function in a startup that’s scaling — keep tuning the running growth engine and keep building new ones— but rather than defining what product-market-channel fit is, let me take a few steps back and build to this point from the ground up.

Zeroth law of business

We all know this formula from middle school mathematics:

Profit = Revenue - Cost

And since businesses need to make profit, revenue should be greater than expenses. Simple enough.

Now, as we covered in the second article of the series, since new-age businesses can track unique customers across transactions, we need to aim for profitability per customer, before increasing number of customers.

So, the formula can be written as:

Profit = (Revenue - Cost) per customer * Number of customers

But, why should we track profit for internet businesses at customer level and why not at discrete time intervals (monthly) or at transaction level (per order) as we have always done?

This is because a major part of lifetime cost (LTC) for a customer is borne by an internet business in acquiring them i.e. customer acquisition cost (CAC). (To the extent that people often say LTV minus CAC, instead of LTV minus LTC) However, CAC is typically front-loaded (Facebook charges your credit card everyday or sends you an invoice every month), while the lifetime revenue/value (LTV) is realized over a period of time. Tracking profitability at customer or cohort level, therefore, paints a more accurate picture of feasibility of the business.

Profit = (Lifetime Value - Lifetime Cost) per customer * Number of customers

What PMF means is that per customer profitability has been achieved, that is first term in the formula above is positive, and number of customers can now safely be increased. There is no point, of course, in increasing the number of customers if net value per customer is negative (refer to zeroth law of business).

Another important point the term product-market fit illustrates is that the fit is only for a particular market (customer segment, rather), and not necessarily for all customer segments. That is to say, the equation above is for a homogeneous set of customers. If the business was selling the product to two or more customer segments, the equation will then become:

Profit =

[(Net value per customer of segment 1) * (Number of customers acquired in segment 1)] +

[(Net value per customer of segment 2)* (Number of customers acquired in segment 2)] + …

To summarize, a product can simultaneously have product-market fit in one customer segment (positive net value per customer), and lack product-market fit in another customer segment (negative net value per customer); and it’d make sense only to acquire customers from the former segment in such a case.

Enter product-market-channel fit

Since customer acquisition cost (CAC) is a big component of lifetime cost per customer (LTC), and since the cost depends on the distribution channel being used to acquire the users, the choice of channel dictates whether or not cost is lower than revenue.

The ‘channel’ part of product-market-channel fit, therefore, underscores the importance of picking the right channel for your product-and-market combination. That is to say, painting with a broad brush, you won’t try to acquire users for your gaming app with newspaper ads, and you won’t try to acquire users for your B2B service with Snapchat ads.

While having net positive value per customer indicates you have achieved product-market fit, if you pick wrong channel to increase the number of customers, your cost per user will get bigger than value per user; and you will break your PMF.

So, product-market-channel fit is all about picking channels which do not break, if not improve, your product-market fit while you try to increase the number of customers. Putting it in form of equation:

Profit =

[(Net value per customer of segment 1 through channel X) * (Number of customers acquired from segment 1 through channel X)] +

[(Net value per customer of segment 1 from channel Y)* (Number of customers acquired from segment 1 through channel Y)] + …

Framework of picking growth experiments to tune product-market-channel fit

Apart from indicating the importance of picking the right channel, the term product-market-channel fit is also indicative of the order in which distribution experiments will be conducted.

First, channel side experiments within same customer segment:

• Experiments within same distribution channel e.g. video ads instead of image ads, or vernacular ads instead of ads in English, or conversion ads instead of install ads, for Facebook ads. Or, changing incentive structure of your user referral program. Such intra-channel experiments will be done most frequently, and will be least likely to break product-market-channel fit.
• New distribution channel for the same customer segment e.g. Newspaper ads instead of Google ads. Experiments in this bucket have lower probability of success than the former, as product-market-channel fit means PMF has been proved for that channel, and not necessarily for all other channels even for that customer segment. This might seem pretty obvious but people often ignore power law and jump to trying out too many distribution channels at once (our business is growing, so let’s buy some TV spots). This to not to say that you should be pessimistic and shouldn’t try out new channels; it is to say every new channel is a new experiment to establish product-market-channel fit and shouldn’t be taken for granted based on fit of another channel.

Then, new market experiments:

• Similar but new markets (customer segments): e.g. moving from professionals in metros to professionals in smaller towns, or moving from homemakers to students. Again, since product-market fit was established for one particular customer segment, and so, every new customer segment is a new experiment and viability needs to be proved before scaling.

Similarly, order of growth experiments on the product side:

• First optimizing user funnel within same product: Improving activation rate through split test experiments in on-boarding; improving early retention through content experiments in marketing automation; etc.
• Then new products: Building an altogether new product to capture more value that can be captured from that customer segment (e.g. Swiggy building grocery delivery service, in addition to food delivery service). Or, building a similar product to enable other distribution channels for same or new customer segment (e.g. Flipkart building a lite app). Generally a strategic decision, rather than a tactical experiment.

Enter marginal CAC or product-market-channel-scale fit

A quick trip from the past to the present:

• You create a product;
• get some early adopters;
• iterate to get good retention among them;
• declare PMF has been achieved;
• formalize growth function;
• study the customer segment and the perceived value of product;
• design go-to-market strategy to position product’s value to the segment;
• pick one or two distribution channels to experiment scaling with (often Facebook and Google Ads, in consumer tech).

All good till here.

Perhaps, after some struggle, one acquisition channel clicks for the customer segment, and LTV continues to be more than CAC. Hurrah, we have product-market-channel fit!

Now, keep scaling this channel while trying out new channels till we go public. Right?

Until, one day, you realize the channel is getting so costly that you don’t have product-market-channel fit anymore. This is because distribution channels often have a particular scale where they hit their sweet spot of cost: lower scale than that and they don’t have enough data points to optimize the campaign for; higher scale than that and they don’t have enough relevant people in the customer segment who are at the cusp of trying your product at that moment.

I call this product-market-channel-scale fit: the scale (of expenditure, or of number of required customers) at which the respective channel will deliver product-market fit at that point in time.

This concept can also be understood as marginal CAC. Often people think of CAC as a constant number existing in a vacuum (as evidenced by questions such as, ‘what’s the CAC for e-commerce apps?’, or rather, ‘why is your CAC not so-and-so number?’); however, CAC is a function of 1. customer segment, 2. channel, and 3. scale (in that order).

The solutions:

1. Short term: Don’t scale a channel when marginal CAC is positive (every additional customer is costing more than the previous customer).
2. Mid term: Run experiments (as per the framework above: intra-channel experiments > intra-product experiments > diversify channels > diversify customer segments > diversify products) to bring down CAC or increase net revenue per user.
3. Long term: Build brand awareness and interest to increase size of population that will action upon the acquisition campaigns in future.

We have now covered key questions that Growth function in a startup faces as it scales (around Series B and Series C for funded startups, but the core questions continue to remain similar, even beyond). Got a question or a comment? Please do share.

In the next article in this series, we will cover the organizational aspects of building and running a Growth function through different stages of a startup. See you there!