What’s your TRUE customer lifetime value (LTV)? – DCF provides the answer





The old formula that everyone uses for customer lifetime value (LTV)) –average gross profit per customer divided by churn – ceases to work properly when you have very long customer lifetimes and negative churn. LTV can become infinite, which clearly doesn’t reflect reality. This post offers a new way to calculate LTV based on discounted cash flow analysis that takes into account the risks associated with revenue that is far off in the future, and the time value of money. The resulting LTV can help companies better understand and manage their future revenue streams and it much more accurately reflects what an investor would pay for that future flow of cash.

Most of the content in this article is aimed at SaaS companies that are past the point where they have a repeatable, scalable and profitable sales process, and are well into their expansion phase. If you’re an early stage SaaS startup, still trying to get product/market fit, or experimenting with different ways to make your marketing and sales predictably repeatable and scalable, it is useful to play around with CAC and LTV to get a feel for where you are. But it’s important to note that these formulae will only yield meaningful results when your sales and marketing process and costs are predictable and scalable. Instead of spending too much time obsessing over CAC and LTV, rather focus your energies on solving the problems of improving product/market fit, and making your customer acquisition repeatable, scalable and profitable.

My apologies that there are some complex looking formulae in this article. We have provided a summary below of the key concepts, and a link* to jump straight to the spreadsheet to model your own LTV. For those interested in understanding the theory behind this model, we provide our usual detailed explanation below.

My thanks to my partner Stan Reiss, who co-authored this piece with me, providing all the expert math help.

*Spreadsheet updated 2/23/16 – errors corrected in Cell B78 on Real World Model tab.


For subscription-based businesses who have “negative churn” (they expand revenue from retained customers at a greater rate than lost revenue from churn), you need new formulae to calculate LTV that includes both this expansion rate and churn.

In addition, because you are modeling revenues that will occur far in the future, you will need to apply a discount rate to account for the risks and time value of money. We recommend using a 10% discount rate, but this may differ depending on your own cost of capital.

This approach provides a lower, but more accurate view of your LTV. We have previously recommended your LTV to CAC ratio be greater than 3. However, we will be working with SaaS companies to understand the impact of this lower LTV calculation and will provide an updated recommendation in a future post.

Payback period does not change, and we still recommend a 12 month payback period unless you have easy access to a lot of capital. In that case, your payback period may be 18 months or longer.

Data you’ll need to model this new LTV, using the spreadsheet tool below:

  1. Average starting contract revenue per account
  2. Gross margin (this should include customer support and account management time to retain and upsell)
  3. Churn rate
  4. Growth rate for retained customers

**These #s can be monthly or annual #s, rates.

Real World Model

A more detailed second model in the spreadsheet linked below allows you to input data to reflect how churn and growth rates change over the life of a cohort of customers. You can input this data for whatever time period you have (first year, 18 months, 2 years, etc.) But, don’t obsess over the accuracy of this data and just use the best data you have. The model is much less sensitive to inputs in the far future; it’s more important to ensure your first year or two of assumptions are accurate.

Free Spreadsheet Attached

For those readers interested in modeling their own LTV, an Excel spreadsheet is attached. Click this link to download*.

*Spreadsheet updated 2/23/16 – errors corrected in Cell B78 on Real World Model tab.


All of my regular SaaS/subscription economy readers know that one of the core concepts that I stress is the idea of measuring Unit Economics at the customer level to help entrepreneurs figure out if they have a good business, and if not, what they need to focus on to improve it. Unit economics is a simple concept where we look to make sure that the profit we are getting from a customer over their lifetime (LTV) is greater than the cost to acquire a customer (CAC).  This is needed because in the early days of a fast growing SaaS/subscription business, the EBITDA losses can be huge, and investors and board members need a new kind of tool, that is not provided in the traditional accounting reports, to help them recognize the difference between the good businesses that will eventually produce great profits, and the bad businesses that will never turn a profit. (For more details on this topic, see SaaS Metrics 2.0.)

Those posts, and several others, focused on the importance of retaining customers, and trying to get to “negative churn”. It is this negative churn that now requires us to develop more complex measurements to understand future cash flows.

Traditional Churn and LTV

To study churn, we need to look at a particular  group of customers that signed up in a particular time period. We refer to these groups as cohorts. So you might have a Jan 2014 cohort which is comprised of all the customers that signed up in Jan 2014. We will then want to track how many customers we retain, and how the revenue for each cohort evolves over time.  Here is a graph that shows what happens to the number of customers in a particular cohort over several years with three different monthly Customer Churn Rates.

As you may be able to tell, the graph is an exponential decay, and for exponential situations, we can use a simple formula to get at the average Customer Lifetime:

And from there, we can simply multiply that average lifetime by the average amount of Gross Profit that we make from each customer every year to get to LTV:

Filling this out with the math for average customer lifetime, and gross profit per account, we get this formula:

                (ARPA is the Average Revenue per Account)

So far, we have only looked at a simple version of churn, which is the rate at which our customers churn. However, things are more complicated when we look at how revenue behaves when you have churn.

Why Dollar Retention Rate (DRR) is different to Customer Retention Rate (CRR)

Imagine a situation where we start the year with two customers, one small and one large:

Now let’s look at what happens if we lose Customer 1:

Our Customer Churn was 50%, but our Dollar Churn was only 17%. So we see that these two numbers behave differently, and need to be tracked separately to understand the full picture of what is going on in the business. This points out an obvious fact: it is more important to retain your larger customers than it is to retain your smaller customers.

Now let’s look at what happens if we were able to expand the revenue we are getting from Customer 2 from $5k to $7k:

Something very interesting has happened here: the expansion revenue of $2k from remaining customers was greater than the $1k that we lost from the churned customers, and we have Negative Dollar Churn, even though the Customer Churn remained at 50%.

This is an extremely powerful concept, which I have written about before here: Why Churn is SO critical to success in SaaS. It does raise an interesting question, which is how do we get our current customers to expand, and that requires that we have either variable pricing, and/or additional product modules to sell. I have covered this topic in a prior post on Multi-Axis pricing.


So far we have talked about Customer Churn and Dollar Churn. There are two other terms which you might see used, which are the counterparts to churn:

  • CRR (Customer Retention Rate) which is equal to 1 – Customer Churn Rate
  • DRR (Dollar Retention Rate) which is equal to 1 – Dollar Churn Rate

So a business that has a negative churn rate, will have a Dollar Retention Rate of greater than 100%. These are some of the very best SaaS business out there. Examples of DRR greater than 100% include: Zendesk: 123%, NewRelic 115%, Box: 130%, Rally Software 127%. (For more public company retention rates you might like to check out this article by Pacific Crest.)

Negative Churn invalidates the LTV formula

In the LTV formula above, I kept things simple by using the term “Churn Rate”. But I should have been more precise and used the term Dollar Churn Rate to avoid confusion with Customer Churn Rate. But now we have a problem, if we insert a negative value in the LTV formula, we don’t get the right answer. And you can tell quite easily without a formula, that if our cohort kept growing at 16% every year (which is what is implied by a negative 16% churn rate), then we would have an infinite future revenue stream that would grow and grow from this cohort.

Our first attempt to model this phenomenon, showed that there is a point where the customer churn starts to bring down the revenue. Here’s a graph showing what would happen if you had a cohort of 100 customers that initially started paying you $100 a month, but each remaining customer increased their payment by $5 every month. The monthly Customer Churn Rate is 3%:


In case of interest, the formula used to compute this graph is: clip_image014.png

a = initial ARPA per month x GM %
m = a fixed $ amount of monthly growth in ARPA per account (not compounding)
c = Customer Churn Rate % (in months)

The formula would work equally well for yearly values.
(Note: there is a huge simplification in this formula that we will address later: it assumes a fixed linear expansion in revenue over time, and in most cases this won’t be how expansion revenue happens.)

Using DCF to account for Risk and Time Value of Money

Given that so much of the customer value in this situation occurs way in the future, where there are risks of the market changing, new competition, technology platform changes, etc. it makes sense that we should factor in that risk in some way when computing LTV. Similarly, we all know that if we were offered a dollar in ten years time, we’d value it far lower than a dollar today. Applying a discount to future cash flows is the accepted way to model this phenomenon.

Here is what happens if we apply a 10% annual discount to the cash flows of a SaaS business with 10% annual dollar churn:

The vertical Axis represents the percentage of the initial cohort contract value, and how it drops over time with a 10% annual churn rate. Then the orange and green lines show the impact of adding discounting of future cash flows to the blue baseline.

This shows what we were expecting to achieve: a lower value associated with cash flows that are farther away in time, reflecting future risks and the time value of money.

This far more accurately reflects what an investor would be willing to pay today, for that flow of cash going into the future.

True LTV: DCF applied to a Negative Churn Scenario

The graph below shows what happens to Cash Flows in a Negative Churn scenario:

  • Churn rate of negative 10%, as a result of the following two variables:
    • Customer Churn is 10% annually
    • Remaining customers increase their spend by 22% of the original contract amount every year (note this is linear, not compounding and exponential)


This requires us to come up with a new formula for LTV:

LTV Formula

Where G is the growth rate for the customers that have not churned, and K is defined by the following formula:

LTV - 2

Where Churn is the Customer Churn rate.
Note: the formula above is included in the provided spreadsheet, on the Tab called LTV Calculator.

Using this formula, we can see that the LTV, for a starting contract with $1 of gross margin, changes as follows:

LTV Table

LTV Graph

The left hand bar shows LTV when there is no discount applied. All the CFOs that I talk to know that this number is too high. Then the four bars to the right show what happens to LTV when we apply discounts to future cash flows.

What is the right Discount Rate?

All financial theory is consistent about one thing: every time managers spend money they use capital, so they should be thinking about what that capital costs the company.  There can be many sources of capital, and the weighted average of those sources is called WACC (Weighted Average Cost of Capital).  For most companies it’s just a weighted average of debt and equity, but some could have weird preferred structures etc. so it could be more than just two components.

The discount rate that is used in a DCF should be your company’s Weighted Average Cost of Capital (WACC).

If we were looking at the true cost of capital, we’d suggest the following discount rates to use:

  • 10% for public companies
  • 15% for private companies that are scaling predictably (say above $10m in ARR, and growing greater than 40% year on year)
  • 20-25% for private companies that have not yet reached scale and predictable growth

The earlier the stage of the startup and the greater the risk in the business, the higher the discount.

However, there is an argument to be made that startup SaaS companies shouldn’t be using a different discount rate to public SaaS companies, as their goal is to show that they have the needed unit economics to eventually become a public company. And investors, board members and management will want to compare their LTV numbers to other larger companies. So given that thinking, we’d be inclined to suggest that the whole industry standardize on using one way of calculating LTV using discounted cash flow, and use one discount rate, which would probably be 10%. That way you could compare private company LTV numbers with public company LTV numbers. We are charting new territory with this analysis, so it will be interesting to hear readers’ feedback on this question.

We have provided the detailed thinking, formulae and calculations behind those numbers on the following page: “How to calculate the discount rate for use in a DCF” Here we go into how to calculate WACC. A key component of WACC  is the cost of equity, which is computed using CAPM, the Capital Asset Pricing Model to arrive at a discount rate. This is explained in more detail in that article for those who have the curiosity to dig deeper.

Impact of DCF on the LTV:CAC Ratio

My regular readers will know that I have long recommended that SaaS startups should have an LTV : CAC ratio of greater than 3.  Given that we will now be looking at lower LTV numbers because of using DCF, this recommendation will need to be changed to a lower number. I don’t yet have a firm point of view on what this new ratio should be, and will be spending time with a bunch of companies that decide to use the DCF version of the LTV formula to watch what this does to their LTV:CAC ratio before making a final recommendation.

However, my other recommendation, that SaaS companies should measure their Months to recover CAC remains a great metric to understand whether your company has good unit economics. For companies looking to optimize their cash flow, I have recommended that they try to recover CAC in 12 months or less. But if your startup is able to cheaply raise a lot of capital, you may be able to relax this, and recover CAC in 18 months, or even a bit longer.

Introducing CORE – Cost of Retention and Expansion

Most readers will realize that it requires Account Managers time and effort to retain customers, and additional sales effort to get expansion revenue. In most organizations this cost is part of the Sales expense, which may be fine for accounting purposes, but is not going to reflect correctly in our CAC and LTV calculations, as it will show up as a misleading number in CAC.  To give you a sense for this, imagine a mature SaaS company with a large installed base producing about $100 million in recurring revenue. To keep that installed base happy, and to maximize retention, that company will likely have a large number of Account Managers. If the cost of these Account Managers shows up in CAC for newly added customers, it will make the CAC number much higher than it really is.

The solution is to move the costs of these Account Managers, plus any extra sales people that are needed to generate expansion revenue in that customer base, into the LTV side of the equation, as these are really a true Cost of Retention and Expansion (CORE) and should be treated similarly to COGS when computing the Gross Margin for that recurring revenue. So in the formulae above*, when computing Gross Margin %, don’t forget to include both COGS and CORE.

How to calculate CORE and Gross Margin % for an average customer

You will likely know how many customers a single Account Manager can handle. And if you have specialist Expansion Sales Reps, again, you will likely know how many customers they can handle. As an example, let’s say that our organization has both Account Managers and Expansion Sales Reps. In that situation CORE for a single customer is calculated as follows:

CORE Formula

And Gross Margin % now becomes:


Where COS is the Cost to Serve a single customer (hosting costs + support costs).

Understanding the effectiveness of an Expansion Revenue Salesforce

If you have dedicated sales people that only work on generating expansion revenue, and they are in addition to the Account Managers, it will be interesting to see what the CAC and LTV is for expansion revenue. My recommendation would be to create a second parallel tracking system to track these two elements. That will allow you to figure out the return on this investment, and compare it to other potential investments that you could make elsewhere in the business.

The Real World is more Complicated!

All of the above formulae make some key assumptions that typically don’t hold true in the real world:

Assumption 1: Your customers will churn in an exponential decay month after month.

In the real world, churn curves have quite different shapes from one company to another, and frequently improve over time. For instance, many companies using annual contracts will have little churn until the end of the contract periods.

Assumption 2: Your remaining customers will expand their revenue in a linear fashion

In the real world, this is highly unlikely to be the case, and will also vary a lot from company to company. It’s possible that you might see a linear increase if they have some kind of pricing axis such as user seats that increases predictably over time. Other factors such as upsell of new modules will be very non-linear, and will occur at odd times. Some companies might see a lot of expansion in the early days, and then it could taper off.

Assumption 3: Your Gross Margin % will remain constant over time

In the real world, most SaaS companies are able to improve their gross margins substantially in the years before they hit true scale (around $150m in ARR). If it is clear that your gross margins will be increasing predictably, you may find it valuable to model this as a changing variable.

Revenue versus Billings

Another important factor to note is that the formula above focuses on the timing of revenue, which could be very different to how cash is collected. Many startups are able to bill a year in advance, which means that the cash flows would actually be quite different to the revenue flows. Applying DCF to the revenue instead of the actual cash flows would penalize those companies somewhat. Trying to build a formula to take this into consideration would add too much complexity, and we don’t think that the difference warrants that extra complexity. But if you were looking to be entirely accurate about this, you could modify the provided spreadsheet to use cash flows (i.e. billings) instead of revenue.

Modeling the real world

The spreadsheet that accompanies this article provides a way to model your real world experiences on how two of these variables will change. Just go to the “Real World Model” tab of the spreadsheet*, and enter in your values for the churn and expansion as you observe (or believe) they change over time from your cohort analysis. It’s unlikely that you will know more than the one to three years of data, so enter only what you do know, and then use the Residual Value formula to predict the other years.

How should SaaS Startups react to this article

Don’t obsess over getting to the last level of accuracy here. We are using a formula to predict the future, and the future, by it’s very definition is not predictable. So recognize that the value in this analysis is to get enough accuracy to make useful business decisions, such as what factors to look at to improve profitability, which customer segments are most profitable, and which have problems that need addressing, etc.

Make sure you have the data you need to understand Churn, Expansion and LTV

There is one thing that all SaaS companies should do immediately after reading this article, if they haven’t done so already: build a database to track revenue by individual customer by month that can be used with an analytics tool to analyze cohorts. You will need to be able to look at all customers that started in a particular month (a cohort), and track customer count and revenue for that cohort over time. Also make sure to tag each customer with attributes such as “Enterprise”, “SMB” so that you can look at how churn and expansion varies over time for the different types of customers. This analysis will become extremely valuable at showing you which are your most profitable customer segments, and where you need to do work to fix issues.

This data will allow you to get Cohort graphs such as the one shown below, that will allow you to understand the shape of how your customer count and revenue decay over time. To create this graph, we are simply overlaying multiple cohorts data on top of each other, and treating their starting month as Month 0. The shape of these curves can be used with the spreadsheet model provided to calculate your real Customer Lifetime Value (LTV).



I have had many questions over the years from SaaS entrepreneurs and CFOs about how to calculate LTV when they have negative churn. They clearly realized the that old formula of Average Gross Profit per Account divided by Churn did not work in their situation. This article and the attached spreadsheet hopefully addresses that question. My apologies that there are some complex looking formulae in the article. If you are a reader who doesn’t like the look of those, simply download the spreadsheet*, and enter your own data into LTV Calculator.

I am really interested in hearing readers’ feedback on the ideas expressed in this article. Please leave me comments below.

*Spreadsheet updated 2/23/16 – errors corrected in Cell B78 on Real World Model tab

My thanks to my partner Stan Reiss, who co-authored this piece with me, providing all the expert math help. Thanks also to Joe Ferrero, Sai Ho and Alan Black of Zendesk for contributing their insights.

About the Author

David Skok

Stay updated

Related Posts

466shares Customers hate being sold to. Customers hate being sold to. They don’t mind...

2.4Kshares In this article I interview Aaron Ross, co-author of a new book,...

681shares An Interview with Andrew Quinn of HubSpot Having a high performing team...

  • Danny Maloney

    Great analysis, David. I appreciate that you went beyond the limits of typical back of the envelope startup math. This is how sophisticated investors (should) look at expected returns and founders need to understand the impact of cost of capital in evaluating their model.

    As you imply, I think you’ll get some pushback on using the same discount rate for earlier stage companies- if for no other reason than it enables early stage investors to argue for lower company valuations.

    From a founder perspective, I think it’s valuable to calculate DCF using both a higher and lower discount rate. When using the lower rate, I’m thinking long-term about our company and – as you suggest – testing if we can achieve the economics necessary to go public one day. When using the higher discount rate, I’m thinking more near-term and being realistic about the higher market rate WACC for early stage companies.

    Perhaps the most accurate approach in modeling expected cash flows would be a blended one, decreasing the WACC input to the model as certain revenue milestones are hit? It complicates the logic of the model, but likely reflects reality more closely.

  • Peter Nesbitt

    I think its very interesting to think about discount rate being a decisive factor in the plausibility of a businesses model. i.e. at 15% discount rate the LTV is higher than our CAC but at 10% it would be less than CAC, therefore a business. As a PublicCo or if rolled into a LargerCo through M&A, then the transaction is immediately accretive by blending the two companies WACCs. A common analysis in M&A is looking at pre/post merger implied share price using a DCF/WACC. This could mean that for earlier stage companies, they should actually spend more than CAC depending on the probability of getting to a size that would make it an M&A target; essentially a discount arbitrage.

  • CustomerSuccessAssociation

    This has huge implications for Customer Success teams, as it’s the argument for making them much more than just churnfighters. Somebody has to be authentically responsible for generating that second order income (upsell, expansion), and who better than the people who have their finger on the day to day pulse of the customer?

  • The LTV thing has been doing my head in for a while. I don’t know how many months to use as “lifetime” because almost all of our customers from years ago are continuing to use indeni and on average are expanding at a good rate.

    So instead, my focus is on this:
    The marginal impact of another account executive (AE / sales rep).

    How long does it take for a new AE to be cash-flow positive (that is, the billings actually collected minus the total costs of the employee, but not fully loaded)?

    That helps me understand when I can add more sales reps and at what rate.

    Since CAC looks at Sales AND Marketing, the other side is the marketing. On the marketing side, all I care about is the ROI with specific activities.

  • Hi Michael, great to hear from you. I agree in principle. But I know of many situations where SaaS companies have turned to specialized sales forces to chase the expansion revenue, where there’s more selling work to be done after making the customer successful. Have you also seen this?
    Best, David

  • Hi Peter, you’ve uncovered the same issue I did while writing this article: what’s the right discount rate. Where I came out in that is that venture backed startups may be in a special class, as they’re not investing to get an immediate return on the capital they’re burning. Instead, they’re investing to build market share and create a much bigger company that can exit through IPO or M&A and provide a return on capital in that way.
    This only holds if the startup falls in the category that I describe. One indicator that a company is in this category would be a growth rate of at least 40%.
    Interesting discussion topic!

    Best, David

  • Hi Yoni, I’m a big fan of that way of thinking. You’re looking at the second form of unit economics: where the unit is the sales hire. You’ll see me reference that in the SaaS Economics blog post, and other recent presentations. It’s a really useful way of making decisions as the way many companies build their business plans is to start by adding a salesperson. That salesperson then needs leads, which drives marketing expenses and possibly Sales Development Reps. Then assuming the sales person is successful, it will drive support costs, etc. so, like you, I like to look at that entire package of costs and work out whether the business can be profitable, and what the cash flows look like.
    I also believe it’s very valuable to look at unit economics where the unit is the individual customer. This helps you to figure out which of your different customer segments are the most profitable. And leads to other value insights around retention, churn , expansion revenue, which are partially covered in this article.
    I hope that’s helpful.

    Best, David

  • Hi David – Terrific analysis and spreadsheet. Rhymes with the post I did on “Your LTV Math is Wrong” – http://bostonvcblog.typepad.com/vc/2015/10/your-ltv-math-is-wrong.html. To your question about cost of capital, I suggest start up founders use 30%. 10% is too low and results inover investment in sales and marketing before truly achieving product-market fit. IBM’s cost of capital is 10%. The startups that we deal with that are most focused on the LTV vs. CAC dilemma have never been FCF positive, never mind consistently FCF for decades, and need to raise capital from VCs at a much higher cost in the form of equity dilution. The other approach is to use probabilities against the cash flows to capture the riskiness of those cash flows and calculate an expected value of cash flows. But the assumptions are too generous if you assume both 100% probability that the cash flows remain constant (startups operate in a very dynamic and competitive a market) while assuming only 10% weighted average cost of capital (raising capital is much more expensive to startups). Anyway, that’s my two cents!

  • Hi Danny, thanks for the feedback. Very helpful. What I am seeing in the comments leads me towards your first suggestion: use the higher discount for a near term realistic assessment, and the lower rate to look at the long term potential of the company, and to compare it to other companies.
    I appreciate your input!

  • Hi Jeff, I had not seen you post before. Looks like you and I strongly agree on the broad principles. Thanks for the input on what Discount Rate to use. I think you are right about using a high rate for early stage startups where there is very little history and repeatability. But this should trend down as they evolve further and show repeatability. What are your thoughts on how you standardize the metric so you can make comparisons between companies? The best that I can come up with for that is to do a second calculation, using a standardized discount rate, for that purpose.

  • As you know, entrepreneurs are perennial optimists. I recommend using 30% as the standard until the company is public. When you are dependent on the private market to raise capital, you need to be more conservative about your cost of capital and its impact.

  • Nada

    You might want to fix your spreadsheet. You don’t appear to be multiplying Arpa/GrossMargin times your formula. So modification of the C13 cell results in no change in LTV.

  • Thanks Nada. I will take a look immediately.

    Best, David

  • Great post David. I’ve been scratching my head on LTV for a while. We are a lower volume / higher priced SaaS operation compared to most, so I always end up with nonsensical numbers when using the traditional “1/churn” approach.

  • Thanks for commenting. It is very helpful to know that this is resonating. Best of luck with your business!

  • Many thanks. This has now been fixed.

  • Nada

    Thanks. We actually incorporated your formula into our CRM metrics tracking system. I was wondering why in the world you number was thousands of times smaller 🙂 Your implementation in excel also helped me find an error (javascript doesn’t recognize ^2 as “squared”), which also improved our numbers. Big thanks. We enjoy your site.

  • Regarding the question of whether the CAC:LTV expectation should change:

    I believe the CAC component should not change definition and not have DCF applied, because CAC is an up-front cost and thus should not be amortized over time. Put another way, DCF adjusts for future risk, but there is no “future risk” for CAC, because it’s already been spent in the past. In a sense, “CAC” is the only part of “CAC:LTV” that is definite and risk-free!

    I believe the rule of CAC:LTV >= 3 should still apply. If you were to adjust the target ratio downwards, because you adjusted LTVs for DCF, you actually negate the effect of the DCF. The whole point of DCF is to make LTV more accurate. So if you adjust your target ratio, you undo that correction.

  • Nada

    We were with you as well. We’ve calculated in a discount for a while now, but even with that we were getting crazy ratios (200:1), because our dollar churn was near zero. With this updated formula, using user (our pricing model) as the churn number, but accounting for our inra-account growth, we get the desired impact, and far more reasonable looking ratios (< 100).

    I had to account for negative churn in other ways that I didn't like. The way David provides makes us feel much more comfortable with the numbers.

    Interestingly, though, I think there's one last issue we have to resolve. When we create a cohort to see something like year-by-year LTVs and ratios, we find that the older years (2013 in particular) we've actually already earned more than what LTV has predicted. This is because we had several much smaller customers bring down our average.

    I'm thinking about adding some time-value calculation in, that adds booked revenue to the base value. Figure out how to apply some DCF to that number (because it was over time), and then estimate the future years relative to the current longevity of the cohort.

    I'm also thinking that's for later next year, as what we have now is great… and there's so much more on which we can focus now 🙂

  • Typically as you get more data and confidence in the future, your discount rate can be reduced. So a brand new startup with no 2-year-old cohorts should use a large rate, but you might use a lower rate now that you’re more predictable.

    However! I would recommend that you NOT do that, because the past is not a good predictor of the future. You might notice a lot of articles on the Internet about how sometimes in years 3 and 4 cancellations suddenly increase. Your market might shift. Global economy might matter. So, remaining conservative is a good way of remaining strong even in the face of unpredictable adversity.

  • Nada

    I guess where I was going was that the formula makes assumptions on likelihoods in given historical years, that if you overachieved on, you could fill in with actuals (they were real, afterall)… but still use the future DCF at the same rate (10-25%) to accelerate the death of the lifetime

  • Excellent post. This is something I’ve struggled with and I didn’t know how to handle. I ran the numbers and while I only have three years of clean data this appears credible. Thank you.

  • Mike P8i

    Great job putting this together! From a practical perspective, I’ve seen some operators put a time cap (3 to 5 years) on LTV calculation to instill CAC spending discipline. Enterprise software might have a longer horizon (10 years?) given the general stickiness once implemented but the cap is put in place since longer horizons potentially encourage massive burn/spending. From a cash flow / working capital perspective, seems like payback period (i.e. 6 to 12 months) is also a bigger focus vs. LTV/CAC ratio since CAC is a tangible upfront cost while LTV can be a fluid/volatile number.

    Another thing to consider when doing LTV calculation is the market tends to be consistent with how it values recurring revenues. It’s a good sanity check to see if your LTV calculation meshes with what an investor is willing to pay for $1 of recurring revenue in the public market (that has similar margin and churn characteristics). It doesn’t have to be exact but I doubt you’re not going to see companies getting their recurring revenue stream priced at 20x revenue. Keep that in mind when doing the calculation.

  • Rodrigo Lobera

    Everything looks great but, what if you have variable churn over time? You are assuming that churn is fixed (I.e. 3% month to month) but, what if that value changes overtime? How do we calculate lifetime with variable churn?

  • In general you should use cohort analysis to understand how churn changes over the lifetime of the customer. Then you can “sum under the curve” to get a good picture of true LTV. I find it useful to get that total, then divide by *initial, first month* MRR, to get a “number of months” of LTV. This is of course a false metric in a sense, but it’s useful to compare with other things.

  • Hi Rodrigo, that’s exactly why I provided you with the spreadsheet. It lets you put in any slope for revenue churn over time. Also you can vary other element s like margin over time.
    Best, David

  • Rodrigo Lobera

    Thank you for your help, very useful!

  • Rodrigo Lobera

    Oh! I will take a deeper look at your spreadsheet and see if it works for me. Thanks for your reply, very much appreciated.

  • Fengqi Shi

    Hi David, great post! I have some specific questions according to the spreadsheet Real World Model Tab.
    1. In row 63, why do you use 0.8+0.8 from month to month instead of do a monthly compounding, )1.08^1, 1.08^2 etc. I see you get the 0.8% discount rate for time period that way so I assume the math should be hold true for row 63 and cell M63 should be exactly 10%.
    2.Cell b78,=ARPA*Gross_Margin*1/(1-K)+(Revenue_Growth_Per_Remaining_Customer*K)/(1-K)^2 should I add an extra parentheses before 1/(1-k) and after the (1-K)^2 for the LTV calculation?

    Another general question for gross margin? What do you recommend for a SaaS company to put into COGS? Do you recommend to include sales commission?

    Thanks you so much!

  • Fengqi Shi

    Hi David,

    I guess one more question. I might miss it somewhere in your blog but I want to know your reasoning behind using a linear fixed growth rate along with compound discount rate and customer churn rate. Thanks a lot

  • This was based simply on looking at a few companies growth curves, and guessing that this was the best curve to fit for them. However the whole point of providing the spreadsheet is to allow people to put in any kind of growth curve shape that they might see in their business, as we knew the real world wouldn’t fit a straight line. I hope this helps.

  • Rodrigo Lobera

    Hi again David, I was looking at your excel and I saw that, in the Real World Model sheet, you never use the Row 64 neither the cell B75. Is there something I’m missing?

    Thanks in advanced for your kind reply,

  • Hi Rodrigo, many thanks for your observant eye. You have helped me find a couple of errors in the spreadsheet, both in the formula in cell B78 (Discounted Residual Value). That formula now does use the value in B75. (I also added a missing set of brackets in that formula.). I will email the updated spreadsheet to you, and make sure the spreadsheet attached to the blog post is updated as well.
    With regards to row 64 (Customer Count), there is no need to use this row in the formulae, as those are based on what you have observed in your remaining revenue line. It is shown just out of interest.
    Many thanks for your invaluable help. And my apologies for the error. Best, David

  • Rodrigo Lobera

    David, thanks to you. Anything I can do to help, count on me.

  • Hi Fengqi, many thanks for your diligent look at the spreadsheet. These are indeed errors, and have been corrected. I will send you an updated version of the spreadsheet now so yo have it before the changes are uploaded. My thanks for your help, and apologies for the errors.
    Re- what should go into COGS, this should best be addressed to your accountants to get the complete answer. But in general the two main components are the costs of serving (e.g. AWS costs, etc.) plus the costs of your on boarding and support teams. Sales commissions do not go into COGS but are placed in the Sales expense line of the P&L.
    I hope that helps. Best, David

  • Robert Z.


    Thank you for providing this article. I have a few question pertaining to the LTV calculation that incorporates growth into the equation.

    First off, why do you take the square of churn and divide that by the monthly growth multiplied by the retention rate? Doesn’t that discount the lifetime value of a cohort too much?

    Also, is the “m” variable within the equation derived from taking an absolute growth rate divided by the total months in a given period or by taking the CAGR over the number of months in that period? (i.e. if we are looking at a cohort that began in 2006 and want to evaluate it all the way up to 2015, would we take the average selling price / ARPU from 2015 and divide it by the 2006 average selling price / ARPU and then divide that by the total number of months (108 in this example) or would we take the CAGR of the 2015 and 2006 ARPUs over 108 monthly periods?)

    Thank you very much in advance for the help and the response.

  • Mike C

    Hi David – Great blog, thank you! I’m trying to understand your LTV equation a bit better and so am triangulating against a simple amortization table with very simple assumptions. For some reason, your math gives a much lower result than I would expect (5.8 vs 9.8). Can you help me with the intuition behind that?

  • Hi Mike, without knowing what math you used in your simple amortization table that would be hard to do. Are you able to share this with me? Thanks, David

  • Mike C.

    Hi David – absolutely, I just emailed you the file. Again, it’s very simple logic to cross check the LTV formula built into your spreadsheet. Doing this to better understand the logic/intuition of your model. In my simple amortization file, I assume a beginning ARPA GM value that grows at 25% per year while I lose customers at the rate of 20% per year (start with 10 customers to make the math simple). I then discount the annual values at a 10% discount rate. The sum of that should theoretically give me the same answer as your LTV model using similar simple numbers. However, your model is producing a value that is 40% lower than mine ($5.8 vs. $9.8). Would certainly appreciate any insights you could share as to why that would be. Thanks!

  • Mike – if you were using the “real world tab” there was an error in the original spreadsheet that might have been the cause of this. Were you using that? If so can you please re-download and try again?
    Thanks, David

  • Jieming C.

    Hi Jason,
    I think the reason David said he might consider adjusting the CAC/LTV downward might be that the number 3 is obtained by using the simplified definition of CAC. (The “CAC/LTV>=3 is good business” verdict is based on observation of obtained data from companies.) Therefore, if you continue to use ratio of 3 after factoring in DCF, when you look back at the data of all successful SAAS companies you might suddenly find that they all failed the successful or not test. Please correct me if I’m wrong @dskok:disqus

  • Hi Jieming – you are exactly right. Thanks for answering this!

  • Sam Handfield-Jones

    Great article. Quick question, in the Real World tab it does not multiply row 64 by row 65, as far as I can tell the gradual decay of the customer count isn’t factored into the calculation, potentially over estimating LTV ?

  • Lek D.

    I noticed that too. Part of it is that you stop at 24 years, whereas the formula is meant to calculate the value of the cash flow till it’s negligible due to discounting. I think there’s a simpler way to express the formula, using geometric series (https://en.wikipedia.org/wiki/Geometric_series). Try this formula instead: ARPA * GM / [1 – (1 – churn) * (1 + ARPA growth rate) / (1 + discount rate)]. You should get fairly close to your amortization table (just add more rows beyond 24 yrs). Cheers

  • Nick Franklin

    We’ve built an interactive “LTV calculator” modelled after David’s work in this post, if you’re interested you can play with it here https://chartmogul.com/metrics/ltv/

  • Matthew Buckley

    Hi David, great post! How has your thinking evolved on the appropriate LTV:CAC ratio for companies who are calculating LTV using this DCF methodology?

  • Hi Matthew, after giving it more thought and talking it over with various SaaS CEO’s, I am still of the view that it should be 3:1 as it is near term cash flows that are so crucial to a successful SaaS business. I also think that the metric “Months to recover CAC” is the more important metric of the two, as it focuses on that issue of near term ROI. I’d relax my recommendation of less than 12 months there to less than 18 months for companies that have access to lots of capital. What are your thoughts?

  • MBATalentForHire

    Hi David – apologies if this is already addressed in the article or the Q&A, but is there a general reason gross profit is used instead of net profit? Wouldn’t it be useful to also deduct opex, taxes, and debt interest (in addition to just COGS)? Are both methods widely accepted, with one being favored in particular circumstances? Thank you in advance!

  • The goal of measuring LTV is to look at what each customer contributes before you add in the expenses of the company. Just the pure contribution margin they bring. So Gross Profit, which is revenue minus cost of goods sold is the best way to get that number.
    If we mixed in all the other expenses, we would not be able to isolate the customer contribution, as the numbers would be polluted with rent, travel, and all sorts of other expenses that have nothing to do with what a customer contributes.
    One of the reasons why I then go on to recommend that LTV is greater than 3x CAC is to take into consideration the need to cover all these other expenses.
    Please let me know if this explanation makes sense, or if I need to try again. Thanks, David

  • MBATalentForHire

    This makes perfect sense, and is the explanation I was hoping to hear. Thank you very much – your articles are very helpful.

Send this to a friend