Demystifying the Difference between CAC and CPA

With help from one of the greatest minds in growth, Brian Balfour (Founder and CEO @ Reforge and ex-Vp of growth @ Hubspot) we will demystify the differences between CAC and CPA to understand where, when and how to use them.

As a guiding principle to your growth strategies, it is important to note these key differences in order to make the best decisions for your growth. The common misconception with these two metrics is that you can use them interchangeably, and that is wrong. 

Firstly:

CAC = customer acquisition cost

CPA = cost per acquisition

Note that customer acquisition is not CPA. And vice versa. 

CAC specifically measures the cost of acquiring a customer, where the definition of a customer is a paying user. Conversely, CPA is the cost of acquiring a non-customer, or more aptly, a non-paying user. This can be anything depending on your OMTM: cost per registration, cost per signup, cost per lead, cost per activation. 

Even though the two metrics are intrinsically different, CPA is used to measure the cost of leading indicators for the overall CAC.    

Models for CAC and CPA

B2C model

Typical B2C models don't require a sales cycle since the conversion time is much shorter. Someone decides in the moment whether to convert into a user (download/register etc). From then, the product would then spend time converting a free user into a paying customer. 

• CPA - cost per registration, cost per activation, cost per signup, cost per lead

• CAC - cost per paying user, cost per advertiser (as Facebook example)

B2B model (and Freemium models) 

Conversely, a B2B model (and Freemium model) requires a sales cycle with lead time to convert a user into a customer. Additionally, the CPA of  a user would then be a leading indicator for the overall CAC of your customers. 

• CPA - cost per signup, cost per lead, cost per registration, cost per activation

• CAC - cost per paying user in Basic plan, Pro plan or Enterprise plan (or equivalent)

Your model, your customer

At this stage you need to understand and define your business model and customer. Understand who/what a customer is to you and at what stage a user is converted into a paying user. The definition of your customer must stay consistent to dispel any confusion. For CPA, choose your OMTM to be the guiding light for all acquisition initiatives. 

CAC and CPA calculations

Here is the minimum viable CPA equation. Which works for any use-case and OMTM. 

CPA - cost per armoire...?

Below is the minimum viable CAC equation. It works well for typical B2C models with very short sales cycles and decision making. 

CAC - cool asa cucumber...?

Sales expenses take into account sales salaries, hardware, software, licenses and phone bills, to name a few. Marketing expenses also mean salaries, hardware, software licenses etc. Any expense or cost related to item that helps the sales and marketing teams convert users. 

What the basic CAC calculation doesn't account for are three key instances that will change how we look and structure said equation. How long is the sales cycle between lead and customer (for freemium: how long before user converts to customer)? How do you distinguish a new customer from a returning customer? And what are the costs for supporting a lead (or free user) before converting into a customer?

Sales cycle

Let's say you're a B2B payments platform like GoCardless. A lead comes in, and is initiates contact with the sales team. From this point it takes 60 days for them to convert to the Plus package. By not taking into account the sales cycle when calculating the CAC, you will skew your data and set you up for making the wrong decisions based on inaccuracies. 

For instance, imagine you run a large marketing campaign that launches in October. Your total spend (marketing and sales) for the month hits £50,000 and the users acquired that month is 500. You then divide the amount spent in October by the new users in October giving you a CAC of £100. 

Before this fictitious campaign began, a KPI of yours was to acquire new users at <£80, so in your eyes this marketing campaign actually failed.

However, if you took into account the sales cycle of 60 days, you would have noticed that in December the new users spiked and hit 900. Meaning the £50,000 campaign amount should be divided by 900, giving a true CAC of ~£55. Since £55 is less than £80, the campaign was a success. This would have been missed if the sales cycle was not taken into account.

Taking things further

We can make the CAC even more accurate through using average sales cycle lengths, but we would need to adjust the calculation to suit. Let us assume that the average sales cycle length is 60 days and sales expenses remain constant over the two month period. 

note: n = month

CAC - cost acquisition customer...?

Customer acquisition cost is the marketing expenses of two months ago plus half of the sales expenses of last month plus half the sales expenses of the current month. All of which is divided by the amount of new customers acquired during this current month. 

Marketing expenses in October were £25,000.

Sales expenses in November were £22,000.

Sales expenses in December were £22,000.

New users in December were 900.

Therefore, CAC = (£25,000 + £11,000 + £11,000) / 900 = ~£52

Key takeaways here are to really understand what your average sales cycle is, and to make sure you have a fully loaded CAC.

Fully Loaded CAC

Expenses should include:

• Salaries

• Overhead (rent, equipment, coffee)

• Money spent on tools (CRM etc)

Expenses should include (depending on your business model):

• Customer success (since your customer success teams improve retention which in turn positively affects customer acquisition)

• Product, engineering, design (as long as these teams build, create and implement product that influences customer acquisition)

• Shipping for free trials (if running free trials is integral to incentivising a non-paying customer to become a paying customer)

To make things easy on us, any team, product or tool that is used to acquire new users should be used to create a fully loaded CAC. E.g. if you have a free trial for a product, any costs associated to hosting that free trial should be used as part of the expense. Similarly, if you have B2B businesses with large customer success teams, such as GoCardless, then those should also be added to the CAC expenses since they definitely enhance the user experience and affect customer acquisition and retention.

Conclusion

• Define what a customer is to your business.

• Be aware of the differences between CAC and CPA.

• CPA is a leading indicator of CAC, but they are not interchangeable.

• Be aware of your sales cycle, if you have one, and be thoughtful when calculating your CAC.

• Any team, product or tool associated with the acquisition or retention of a customer must be included in the expenses to give you a fully loaded, accurate CAC. 

How To Run A Growth Experiment

Using the G.R.O.W.S method

Editor's note: I promised another prioritisation essay in last week's piece. This is it. Plus more. Much more. Similar to last weeks piece, best results from this approach happen when you have plugged as many holes as you can in that leaky bucket product of yours :)

In order to get the most out of your growth strategies, having a clear cut process in running growth experiments is paramount. As long as you approach this with some scientific methods - allowing any test to be analysed from a quantitative perspective - then you are on a good path. 

A simple scientific experiment outlined below can be found in any secondary school textbook.

Minimum viable experiment

We will use the same concept, just in a more scalable way.

The key here is to have an unbiased decision making process when focusing your time, effort and money into finding your core growth channels. Putting your trust into a scientific growth experiment allows you to follow the data rather than your gut (don't get me wrong, your gut is very intuitive, but it won't be able to consistently tell you whether your multivariate-tests-of-the-future will work). It will ultimately help you make the most informed decisions you can in un-earthing the growth for your business. 

There are a few ways (or anagrams) you can follow for your experiments, however, one that has worked for me is the G.R.O.W.S process, coined by the Growth Tribe out of Amsterdam. 

G.R.O.W.S

The G.R.O.W.S method

The G.R.O.W.S process follows this order:

1. G - Gather Ideas

2. R - Rank Ideas

3. O - Outline Experiments

4. W - Work

5. S - Study Data

Let's dive in to each stage.

G - GATHER IDEAS

Trello board for backlogged ideas across the Pirate Metrics

Creating a company-wide spreadsheet for ideas that popup anytime is great way to start. How I organised our growth backlog is by using the conversion funnel (or pirate metrics) as a guide, going from awareness, acquisition, activation, retention, revenue and referral.

You can take things further by setting up formal brainstorming sessions with the whole company or multiple teams depending on size of co. To get the most out of these sessions, it's best to learn how to brainstorm like a Googler.

Simply put, the growth team ask every individual to brainstorm on their own around a specific stage of the conversion funnel, e.g. user acquisition. Everyone then comes together to flush out the best, most creative and innovative ideas to start ranking. Each person must:

1. Know the user

2. Think 10x

3. Prototype said ideas

R - RANK IDEAS

Next step is to rank or prioritise your ideas. If you take a look at my previous essay on how to prioritise your growth, I map out a prioritisation framework (Bullseye). You can use that or, since we are focusing on experimentation, you can create any simple ranking system to help rank best ideas. Such as:

Who ate all the P.I.E.S

What I like about this system is the semi-scientific approach. Good tip is to make any decision a quant one. The image above shows a mixed variety, meaning ideas from different stages of the conversion funnel. It's actually best is to focus on one stage every time. 

O - OUTLINE EXPERIMENTS

Now that you have one highest ranked idea, you can start experimenting. Key here is to design a test that will verify whether the specific idea/channel/approach will be a success or failure. Best way to go about this is to build out an experiment sheet.

Minimum viable version:

Minimum Viable Version

• Top ranked idea: Create a popup page after created a lineup to incentivise users to play multiple paid games with one-click

• Research: Currently we have paying users playing 2 paid games per gameday. This popup page will make it much easier to play more

• Hypothesis: By using the popup page to incentivise more paid games, we will increase paid games per user per gameday by 50% and therefore increase average revenue per paying user (ARPPU) by 20%

Maximum viable version:

Top ranked idea: Create a popup page after created a lineup to incentivise users to play multiple paid games with one-click

Maximum Viable Version

Top ranked idea: Create a popup page after created a lineup to incentivise users to play multiple paid games with one-click

• We believe that there will be an increase in paid games because of a popup screen that incentivises users to play more with one-click

• To verify this we will send out a beta version with said integration to our top 5% users - without telling them the update - and analyse their behaviour when prompted

• We will measure number of paid games played (#) and ARPPU (£) to see if there is an increase in either or both due to the integration

• It is a success if we improve the ARPPU by 20%

• Results Quantitative: # paid games played up by 75% and ARPPU up by 33%

• Results Qualitative: users say "very useful", "quick and easy", "organic placement of popup", "much more fun, quick and useful"

• Next steps are to rollout integration and update game 

W - WORK WORK WORK

Rihanna says it best

This the area that separates the men and women from the boys and girls. You can do a number of things to maximise the effectiveness of your work and experiment, however for me, weekly sprints worked best. They really allow you to laser focus on one goal (the experiment) and truly smash it out the park. No excuses. 

S - STUDY DATA

Which pill...?

This is where the insights kick in and you make your growth decisions. This is where the test is verified as either a success or failure.

If you get the analysis wrong, you make the wrong decisions. Therefore you could jeopardise your whole growth model. Which could mean the end of your business, depending on its stage of life. No pressure.

Make sure you understand, record and analyse both quantitative and qualitative data for every experiment. Understanding how the users reacted and felt, matter as much as the data behind their behaviour.

Try to find any differences in user behaviour from before, and ask yourself why this has (or has not) occurred. Does it fit with your original hypothesis? Is this healthy, scalable behaviour?

From all of the data you have acquired and consumed you can now conclude success or failure. And depending on answer, either use the data for further background research in future projects, or use to rollout the experiment company-wide. 

Key tools I used were Fabric.io (for real-time acquisition metrics), Amplitude (for retention, segmenting, behavioural cohorting, data visualisation and Metabase (for all data retrieval and querying). 

RINSE AND REPEAT

Rinse and repeating growth experiments

Even if your first seven experiments fail, you can use the data acquired as more background research for the next seven. This iterating experimentation process theoretically means the more experiments you run the higher value they become (so by experiment 1,273 you'll be pooping gold). 

Remember, rinse and repeat until you find that one core channel that really grows your business. The one that hits the sweet spot. When you do, ride that channel as long as you can, dig deep, iterate and squeeze as much growth as you can out of it until it's fully saturated. 

Then rinse. Then repeat. 

You can find my other musings here, here and here.