The CAC Bathtub
Optimization is not a destination, it is an ongoing byproduct.
How to get my CAC/LTV model.
I have created a predictive CAC/LTV model that will respond well to many pricing strategies, business models, etc. I use a modular CAC/LTV equation to ensure that these values are as close to reality as possible. While this model was made to be predictive, you can also input some of your historic data to generate a more accurate perspective with regards to the health of your business.
I've spent hours ironing out the kinks in this model, but, as usual, if you find any errors or points of improvement, please feel free to email me. I'll update as feedback comes in.
CAC - the make or break of a business.
CAC (Customer Acquisition Cost) is one of the most important metrics that helps executives and investors determine the health of a company. LTV (Lifetime Value) is an estimate of the average gross revenue (revenue before marginal expenses) that a customer will generate before they cancel your services (churn). CAC and LTV are closely related - the relation between these two values is a large determinant of the health of your business. As long as your CAC is reasonably below the LTV, then you’re in good shape.
However, determining where your CAC is relative to your LTV is extremely difficult to calculate and predict, especially given the rise of many hidden variables and complex pricing strategies. In this post, I'll conduct a deep dive into some practical and experimental methods that I use to predict and analyze both of these key metrics.
Introducing CAC and the traditional means of calculation.
CAC, in the simplest way, is calculated by adding all of your sales and marketing expenses and dividing by the total number of new customers.
CAC = (Sales + Marketing Expenses) / New Customers
This is an easy calculation to conduct after you have the data, but predicting CAC with this is difficult because, in reality, you'll be putting together data from a bunch of different channels, sources, etc.
While there are industry trends that are somewhat demonstrative of how much each user will cost to acquire, the most important thing is not necessarily the acquisition cost, but the ratio of CAC and LTV (although I have heard some horror stories of startups burning through cash on crazy acquisition expenses).
The failure of traditional means of predicting and calculating CAC.
Once you have all of the data, it's relatively easy to calculate CAC. However, if you are a startup, have zero data, and only a small amount of runway with regards to funding, a small deviation or incorrect CAC calculation could make or break your venture.
Traditional means of predicting CAC do not parallel the innovation that we see today in products, pricing strategies, and business models. For example, Amazon has likely accepted that their CAC is in constant flux. With all of their acquisitions, multi-tiered pricing strategies, programmatic pricing algorithms using Bayesian inference, and diversified product line up, predicting and calculating CAC is quite the herculean task (if not impossible). While Amazon may be ok with taking the hit if their calculations are incorrect, 99% of early stage companies are not.
Subsequently, we need more flexible, adaptable, and modular means of predicting CAC. Math nerds like meh have a lot of catching up to do.
A more pragmatic approach to predicting CAC.
All of this research has arisen because many of my clients don't necessarily care - obviously within reason - how much each user/customer will cost at launch. Many founders are more interested and excited in seeing their product in action given the pressures of an already mounting sunk cost.
Instead, I often get asked, much more pragmatically, how much money it would take to acquire x amount of users. This is difficult to measure accurately, and is not a value that the traditional CAC equation can accurately measure. Subsequently, I have researched and analyzed CAC/LTV thoroughly to make accurate predictions for early stage companies.
The CAC/LTV ratio.
The CAC/LTV ratio should - at the most - be 1:3. I recommend trying to stay around 1:4 or 1:5 to ensure that you have enough cash on hand.
For example, a subscription model that charges $5 a month and the average customer will stay on for an estimated 18 months (therefore, $90 LTV), then the CAC for should be in the ballpark of $15-$20 (I think this is crazy high, but this is industry standard).
If you feel as though you have exhausted all marketing measures and still have not achieved a healthy CAC to LTV ratio, don't worry. There are so many ways to push CAC down and increase LTV simultaneously (see later).
The ideal CAC.
The ideal CAC, which is characterized by reverse-hockey stick graph, begins with an initially inflated acquisition cost followed by a rapid and consistent decrease period over period.
This exhibits an initially high acquisition cost, followed by a reverse-hockey stick shape that asymptotically approaches zero.
More often then not, this occurs from a front-loaded media budget that acts as the initial catalyst to a viral growth kicker (VGK) that snowballs out of control. Many companies have VGKs of 1.3-1.8 - so, every user who signs up invites 1.3-1.8 users. This creates exponential growth that obviously reduces CAC drastically. By the end of period 12, this company is acquiring users almost for free - and it is likely 80-90% organic growth.
More specifically, this kind of growth is caused by a mix of structural, programmatic, and viral components. Some of these often include:
1. > 1 Viral Growth Kicker. Check out my CAC/LTV model to see how small fluctuations of the VGK have drastic effects on growth rates. Normally, a 1% increase in the VGK will lead to a 3.8% increase in the user base.
2. Increasing Conversion Rates Overtime. Also, companies with this growth will likely use programmatic or advanced retargeting methods to increase their conversion rates over time.
3. Stable or Proportionally Increasing Marketing Expenses. Keeping costs low is another way to decrease CAC. To minimize CAC, try to keep your marketing expenses stable over all periods.
4. Product Enhancements. Product enhancements (additions to the MVP) after launch normally augment all of the above factors. For example, this enhancement could increase the amount of time users spend on the platform on average. Therefore, this will likely lead to an increased VGK.
While this graph is ideal, this is almost never the case (except, for example for tech giants such as Facebook, Instagram - not even Twitter or Snap). There are a TON of unexpected events and hidden variables that cause CAC to increase or plateau. This leads us to what I like to call the CAC bathtub.
The real CAC (the CAC bathtub).
All of this is to introduce the real CAC graph, which I like to call the CAC bathtub. If you have an econ background, this is similar to the average variable cost curve that depicts (dis)economies of scale.
Similar to the ideal CAC graph, this graph exhibits a decreasing cost from periods 1-4. However, at period 5, the cost plateaus for a few periods until it finally increases from period 7 to 8. Yikes.
This is the reality that many early stage companies (in fact, all companies) face. Here are the underlying components that produce this:
1. < 1 Viral Growth Kicker. Nothing viral here. This company might not be positioning their product correctly, and they probably are not taking advantage of simple gaming mechanics that seamlessly encourage sharing.
2. Decreasing Conversion Rates Overtime. As the company expands their traffic sources, conversion rates could fall pretty drastically. Ad fatigue (below) also has a role here.
3. Increased Marketing Expenses Overtime. This could occur as the company adds more people to their marketing/sales team, or could also be attributed to an increased ad spend relative to CPAs.
4. Ad Fatigue. This is where things get tricky. Ad fatigue means that your audience is saturated - they're sick and tired of seeing your ads and are likely not going to convert. Your CPAs will drastically increase here. There are MANY different ways to cure ad fatigue (coming soon to this blog).
Hidden variables behind CAC/LTV.
No model is correct, some are simply more helpful than others - is something that we often forget.
I have never had a client who has correctly predicted their CAC/LTV upon launch. It is critical to the success of a startup to conservatively predict CAC/LTV, unless you have endless amounts of cash.
There are many hidden variables, some of which I have listed below, that distort CAC/LTV models. The degree to which a company predicts, prepares, and adjusts to these distortions is a key determinant of its continuity as an organization.
Seasonality is a characteristic of a time series in which the data experiences regular and predictable changes that recur every calendar year. On the consumer side, the most obvious example of this occurs during the Holiday season - consumption is elevated, thus so is employment. We observe during this season, that consumers spend more relative to their income; therefore, they are more likely to use your services (given that it is a good product). On the business to business side, this is more difficult to calculate and predict because it so industry specific.
Seasonality can have serious effects on the traditional freemium model - you may see an increased share of users paying for your product over the course of different season, followed by a sharp drop off (e.g. Increased Audible subscriptions during the summer months - long days by the pool).
Ask. What time series/during what time of year is growth over-index? Under-index? How do these growth rates effect churn patterns?
Intermittent/Incremental MRR Variations
Intermittent MRR Variations occur most often in the business to business arena when there exists a 2 to 3 part pricing strategy. For example, let’s say a media company has three tiers of monthly retainer services, to each cohort in these tiers they offer small upsells that complement their services. Some months, their clients agree to the upsell, while the next month they might not - seasonality is another factor here (e.g. a company may agree to increased marketing services during the holidays to sell more, but not during the summer). These small, intermittent fluctuations that are difficult to predict and track have pretty preponderate effects on LTV.
Ask. What cohorts can I break my customers into based on how much revenue they generate? Are there patterns in MRR fluctuations for each cohort? What fluctuations are most often occurring and how do they affect CAC/LTV?
Variable Churn Patterns
Churn (which occurs when a customer stops your services) does not occur linearly - this means that you cannot necessarily calculate CAC/LTV using simple analysis. Churn often happens sporadically and often without coherence, subsequently, it is critical observe basic trends and prepare for the worst. There are four main churn patterns, I'll delve deeply into these in a coming post simply because it in itself is a mountain of a discussion.
- Annual Renewals.
To account for variable churn, I always calculate a conservative LTV estimate to account for all of these hidden variables. At the end of your LTV calculations, I recommend multiplying it by 0.75-0.8 to ensure that your business will be in good health even if things turn south.
Ask. What churn pattern does my customer base most often follow? Which churn patterns does each customer segment most often follow? What are the main factors affecting churn?
Hugely Different CAC/LTV by Customer Segment CAC
One of the biggest mistakes when it comes to calculating CAC occurs when we place the weight on each customer. Most of the time SaaS products have several different user groups/segments (if you don’t know if you do, you need to conduct some user research).
This is extremely difficult to calculate. Depending on your funds/expertise, Bayesian Inferential models could be the solution here. In short, Bayesian Inference systematically avoids the effects of outliers on the data set by narrowing the distribution. This allows you to see a CAC that is closer to the true CAC. I use Bayesian Inferential models for clients who have a diverse user base (meaning that the user base can be segmented into vastly different psychographic and demographic buckets). One user group, let’s say it accounts for 18% of the total number of users, may be significantly more expensive to acquire than another, but may also have a higher LTV. Subsequently, we want to calculate a CAC cost that is unbiased with respect to this group.
Ask. What cohorts exist in my user base? How does the CAC/LTV differ between groups? How can I systematically group users by cohort?
Short-run CAC optimization.
Expanding the runway.
Remember - CAC optimization is an ongoing byproduct of consistent and constant testing and analysis. You will never truly optimize CAC, it's an endless journey that is present during the entire life of an organization. You simply want to create as much runway as possible at the lowest attainable CAC level.
I use this method of graphical analysis to observe fluctuations in CAC. From here, I often know when I need to implement new methods to decrease CAC (keep it 'sticky-low') in the short run - this is what I like to all 'expanding the runway.' This means that we are trying to use preventative measures to deter our CAC from increasing in the short run - we're pushing the inevitability of an elevated CAC to a later date.
Check out the graphs below (math disclaimer). The first is a realistic CAC graph that shows increasing costs from Periods 7 to 8. The second graph is a linearized derivative of the first and shows the inflection point - this demonstrates the point at which where CAC begins to increase. I generated these graphs using the data from a predictive CAC/LTV model that I created.
See below updated graphs after I use preventative measures to push an increased CAC to a later period to maximize short run growth.
After implementing new strategies, I push increases in CAC from period 7 to period 12 - this simple push can mean life or death for an organization.
Methods to decrease CAC.
Putting all of this into practice, here are some methods to put downward pressure on CAC and expand your runway to push increasing costs further down the road. This is an ongoing, rinse-repeat situation.
Increasing the viral growth kicker.
A 1% increase in the VGK, on average, increases the user base by 3.8% - that's a damn good return. This is the lowest hanging fruit when it comes to growing a pretty massive user base relatively quickly.
1. Referral Program. Think Uber and Lyft here - they have huge user bases because they built inherent viral systems into their product. For every person that you refer, both of you get $5 off your next ride. Create a system that is conducive to viral growth, put into place proper incentivizes, and watch the users do the work.
2. Product Enhancements. Consider sending automated emails based on activity. For example, the app letgo sends (annoyingly) automated emails for every message that a user receives, fluctuation in price on related products, etc. While I personally find these emails to be a little over-the-top, I have told at least 12+ people about letgo in the last month. Effective, yes.
3. API Integrations. This one is simple. Make it as easy as possible for users to share your product. For example, see how Spotify finally worked with Instagram to link Stories to specific songs. This is why Spotify is killing Apple Music quarter after quarter - they understand that their user base will share songs if they make it easy to do so. I'd love to see these CTRs.
4. PR Stunts. PR stunts can be a great way to spur massive amounts of viral growth. The only downside of this strategy is that it normally only lasts a few days to a week, but it can catalyze a new wave of users. Most recently, Deadpool performed a brilliant stunt that created massive amounts of publicity. Before this stunt, I personally would not have seen the movie, but this ingenuity created a positive association with the film that attracted me to the theaters.
5. Combination Strategies. The best strategies are those that combine multiple strategies synergistically. Consider combining any or all of these tactics to see a significant VGK increase.
Optimizing the acquisition funnel: test-fail-optimize.
This is more difficult than it sounds. Often, this means conducting rapid fire multivariate testing (more info coming soon), and spending a dedicated amount of money and time on testing.
In short, to decrease CAC, it is vital that you test every channel and specific methods to get closer to optimization. Test each method, fail as fast as possible, close the funnels that failed, move on - this is the cookbook so to speak. In the process of funnel optimization, failing is more insightful than succeeding because you now know what doesn't work, rather than banking on one successful campaign that could have been a fluke.
Another strategy could also be expanding your traffic with new, larger audiences. Try making your campaigns a little less targeted and appeal to a larger audience. Take this lightly - expand your audience incrementally (like 5-8%) and see how it changes your metrics. If your conversion rate stays the same, continue to broaden your audiences until they drop. Traffic expansion is a tool to be used lightly, but it can be very effective at scaling your growth rates.
Increasing retention can be accomplished in several different ways. Conventionally, and most effectively, you should work on product enhancements to increase the amount of time each user spends on the platform or has the app downloaded on their phone. Unconventionally, and more easily, consider implementing several things that augment the user experience, see below.
1. Killer Social Media. Having an awesome social media is a great way to increase retention — try to use your social as a sales channel, as a social proofing method, and most importantly, as a way to get in touch. Check out how ASOS uses their social media to increase customer retention by being 100% accessible for order help on all channels.
2. Chat Bots. Chat bots are often thought of as an easy funnel stacking method or a means to augment the user experience, but they can also be used as a retention method. Like email retention methods, chat bots can often be used to broadcast information. They also have significantly higher CTRs and open rates than email.
Increase customer LTV.
This one is simple. To decrease the ratio of CAC:LTV and get your business in better health, you can simply increase the lifetime value of each customer. This can be accomplished by re-evaluating your pricing, retention, and business strategy. Perhaps you could increase the cost of the top tier of your pricing plans and test the sensitivity of each cohort related to changes in price.
Go on, say hello.
Woah, that was a lot of info.
If you have any questions on any of these topics, I would welcome the opportunity to sit down for coffee or hop on a quick call. Or, simply email me.