A couple weeks ago, in a comment to an MBA Mondays post, Dan Lewis wrote "LTV has to be higher than your CPA or you're not going to make it." I asked Dan to elaborate in a MBA Mondays guest blog post on this topic and he agreed. Here's Dan's post:


LTV stands for “lifetime value” of a customer.  CPA stands for “cost per acquisition” of a customer/subscriber.    LTV has to be greater than CPA or you won’t be able to scale – or, for that matter, survive. To demonstrate, let’s go to the video tape:

Monday through Friday, I publish a free daily email newsletter (which you can, and should, sign up for), putting to words the notion that you should – and can – learn something new every day.  Oddities, like the fact that Abraham Lincoln created the Secret Service the day he was shot, carrots used to be purple, there is only one Jewish person in all of Afghanistan, etc.  It has about 4,000 subscribers.  It’s basically a blog, sure, but I send it as an email newsletter.  Long story; one for another day.

The main costs involved – the costs of writing and editing the email – are fixed costs.   (At this point, it’s also entirely a time cost – mine to write it, and a volunteer who very generously edits it before I send it out.  But even if I hired writers and editors, the cost of producing the content is, at least in this context, a fixed cost.)  Write it once, send it to many – as many as subscribe.  It costs the same amount to write the email to one person as it does to one million (God willing!) subscribers. 

The marginal costs are, relatively speaking, minor, and consist of the fees paid to the email service provider I use, Mailchimp.  At the number of emails I send and the number of subscribers I have, it’s about a twentieth of a cent per email, or $0.05 per 1000.   Let’s say I can monetize the emails pretty easily at $1 per 1000 (or $1 CPM, as in “cost per mille” from a fictitious advertiser’s perspective), and my margin is 95%.   Your blog, twitter account, etc.?  It’s probably even better, unless you’re somehow paying on a cost-per-post basis.

Let’s say that an average subscriber sticks around for 10 months, and that I publish 20 issues a month.  And, while I’m making up pie-in-the-sky numbers, let’s also assume I can get $5.00 CPM from advertisers, on average, over the course of those ten months.  The lifetime value of each subscriber?   $1.  Each subscriber receives 200 emails, and at $5 per 1000, that’s a buck.

That LTV – $1 – is my upper limit.  Spend any more to get a subscriber and I am losing money with every additional subscriber, and that’s bad news.  So while word of mouth, Tweets, Facebook posts, guest blog posts (did I mention you should subscribe to the newsletter?), etc. are all very good ideas — that’s how I’ve made it to about 4,000 people signed up thus far — are also pretty close to the only viable ways to grow the subscriber base.   

Other ideas?  AdWords or Facebook ads would cost about $0.25 per click to advertise on “learn something new every day,” requiring that I convert a quarter of clicks to subscriptions just to break even – a tall task for email newsletters, akin to getting 25% of visitors to subscribe to your blog’s RSS feed.  A lot of list-building (cue ominous scare quotes) “best practices” – co-registration, contests/incentivized signups, etc. – will hurt the CPM level and likely lead to a much lower LTV.   And buying lists?  No thanks.

This inability to turn money into new subscribers, customers, etc. is a huge limit on the speed and potential for growth.  Sure, things can break well and the newsletter can grow by itself, via the methods noted above.  But those methods are a mix of elbow grease, dumb luck, a solid product, and a lot more of that dumb luck stuff – the hallmarks of the “secret sauce” which somehow, some way turns a hobby into a viable business, in a manner and fashion beyond repetition.

LTV > CPA applies, of course, beyond newsletters and beyond media, digital or otherwise.  It’s generally easy to apply, as even estimations based on guesses and conjecture are enough to make sure that you’re not making a huge mistake.   After all, it may be a great idea to spend a ton of capital and spend it to grow your customer base, but not if you’re spending more money to get each additional customer than that customer is worth.

#MBA Mondays

Comments (Archived):

  1. William Mougayar

    That’s a great example,Was there supposed to be a video attached?

    1. fredwilson

      no, that’s just a warner wolf line “let’s go to the videotape”

      1. William Mougayar

        Ok. Thanks. You fooled me 🙂

    2. Dan Lewis

      Thanks!And as Fred said, yeah, it’s a reference to Warner Wolf. He was the sports guy for CBS News in NYC during the 1980s, a formative point in my life, especially now as a Mets fan. That was his catchphrase, throwing to the details of the events at hand — albeit literal in his case!

      1. fredwilson

        oy. the curse of being a met fan. and a jet fan. and a knick fan. i guess itbuilds character

        1. Dan Lewis

          I managed to be infused with New York Football Giant DNA. Not surehow you all cope.

        2. matthughes

          You have recourse in New York you know…It’s called the Yankees.

          1. fredwilson

            I’m a sucker for the underdog. I can’t root for the Yankees Lakers orCeltics

        3. falicon

          I must have *a lot* of character…

      2. William Mougayar

        I hesitated thinking it was figurative given that Fred’s posts often havevideo.

  2. LIAD

    sweet post Dan.Another great primer on LTV & CAC is David Skok’s blog – http://www.forentrepreneurs…His great use of graphics really helps drive home the point.

    1. Scott Barnett

      yes, David’s blog is awesome for us “old-fashioned” B2B’ers – so if you’re selling to a business vs. consumer, it is a must read. Particularly the exponential jump in CAC as you “complicate” the sale.

    2. PhilipSugar

      That is a really well done blog, especially for us BtoB’ers as Scott says.

    3. Mark Essel

      +1 LIAD, props to the guest post Dan.

    4. Youssef Rahoui

      Yep, great blog indeed!

  3. awaldstein

    OK…you got me Dan.4001 subscribers. I’ll stick around if you start putting in wine facts ;)LTV gets wacky and insanely complex as the number and types of products you sell increases.In fact, there are many (if not most) smaller companies who sell catalogs of stuff, or get paid by referrals who struggle all the time with parsing data at the back end data to figure out where the click goes, does it come back, does it refer forward, blah blah blah.The math at the top is easy, the data collection to feel comfortable with that math when the model has a number of variables gets complicated.

    1. Dan Lewis

      Yeah, as the type of customer changes, so does the analysis (and complexity thereof). That’s also true for, e.g., a varied monetization strategy. For example, I’ve gotten a couple requests to push my newsletter to Kindle, which I’m glad to do, but apparently cannot do for free (per Amazon). I’ll probably end up doing it for $1.99/mo, which nets (I think) $0.60/mo to me. And I’ve had a couple requests to cobble a group of emails together and turn them into a book or ebook. Totally changes how I have to approach things.And thanks for subscribing! We’ll see about wine… maybe I’ll do some digging.

      1. awaldstein

        I’m wondering whether a distribution widget for your ‘factoid’ wouldn’t work. Bloggers are always looking for data to publish so someway to share the factoid forward and have it live in a box/button on other blogs might be interesting as a channel.

        1. Dan Lewis

          Interesting. Hard to do, especially as a non-coder (me). But interesting. Thanks!

          1. awaldstein

            This is done in the deal space all the time. Building an affiliate network to distribute the deals forward.People sweat over the payment system for this. I believe that providing information to pass forward to bloggers/alpha sharers is as powerful.Remember those desk calendars with an idea a day? That’s what you may have but the desks are the URLs of bloggers.

    2. Mark Essel

      Dan can count me in as well. Always happy to tune into a regular at AVC, it’s my favorite channel with countless spin offs.

  4. Sean Weinstock

    Good corollary to the post on marketing (… from a while ago. A related interesting topic is the best-guess approach to LTV. It seems like a lot of people thinking about starting up businesses get caught in worry about not knowing, and/or some start worrying too little.Also – I’ll be making references to Kentucky Bend and purple carrots throughout the day.

    1. awaldstein

      Managing to LTV is a slippery slope.You broaden acquisition to find new customers bringing up cost. You broaden what you sell to bring up LTV. Sometimes your core value and your focus get lost in this shuffle.

      1. Sean Weinstock

        Interesting. I guess a question for me is (with relation only to websites) – how important is initial core value to the end game? And/or, does core value transition to being defined as ‘whatever it is the user base wants to see/do’? This is especially a question with media businesses where your content (your initial core value) tends to cater to a niche, and that niche has many other associated interests.

        1. awaldstein

          Sean…all great questions.They point to a broader conversation on what you are building, for whom and how to connect with the initial market. At the early stage these are of course more assumptions than actuality but important to noodle over and understand as a baseline.These posts may be of some value to help think this through:

    2. Dan Lewis

      I like to play a game where I ask someone to give me a topic — something more specific than “Earth” but more general than “Fairfield, Connecticut” — and I’ll shoot back a random bit of trivial nonsense. I rarely use Kentucky Bend, honestly, but maybe that’s a new challenge 🙂

      1. Sean Weinstock

        Haha, sounds great. I’ll keep that in mind as I get more versed in the realm of one-off facts. Looking forward to the emails.

  5. Fernando Gutierrez

    One more subscriber, I bet you’re gonna get a bunch today.I use this approach myself in my business, but I make an adjustment when calculating LTV: I take away the top 10% customers. When they differ a lot in value (my case) a few of them can change the average significantly, so I prefer not to consider them when calculating LTV. This way I know I may not get a couple because I wasn’t willing to pay a high enough CAC, but I prefer to be conservative with my margins.

    1. Dan Lewis

      My focus is, actually, on my top 10% — not because of the monetization opportunities, but because I think in building a “fan base” (to use a crude term), they’re the best. Read Seth Godin’s stuff on the topic.

      1. falicon

        Awesome thoughts/post…and totally agree with you (and Seth) on the fan base …that’s how I try to build all my projects too.BTW – apparently this is what happens when I go offline for a couple of days…awesome to come back and see you and N.I.K. get a little love/attention…your little ‘hobby’ is now on fire…soon I’ll be bragging about how I was one of the first subscribers to N.I.K. (and people will actually know what I’m bragging about!) 😉

    2. John

      If you’re seeing a large standard deviation on your range of LTVs, it might make more sense to use the median instead of the mean as your measurement of ‘average’.That’ll reduce the effect of high outliers without being as crude as ‘ignore the top 10%’.

      1. Fernando Gutierrez

        You’re right, I like it more that way. In my case the result is quite similar using the median, but that’s because I’ve chosen that 10% analyzing my customer base (fairly uniform except for a small percentage that are quite different). As a general rule using the median looks more reasonable.

      2. Mark Essel

        I’d take it a step further and look at it as two separate distributions/businesses (multi-modal).Robust statistics is less sensitive to outliers but also is less sensitive to rapid changes in distributions over short time periods.

  6. Dan Ramsden

    This post is a perfect complement to yesterday’s bubble talk, for added perspective. Not arguing one way or the other, but valuations should be run through LTV/CPA calculator as back-of-envelope sanity check.

  7. Pat McGee

    Re 4th paragraph: 1 / 20 cent times 1000 = half a dollar.Either you meant 1 / 200 cent, or your margin is 50%, not 95%.

    1. Dan Lewis

      1/200, sorry. $50/mo to send 1MM emails.

  8. Charlie Crystle

    I’ll bring you some purple carrots when they’re in season. Guaranteed never to go orange.

    1. Dan Lewis

      I’m eating “baby” orange ones right now (which, by the way, aren’t really baby carrots — they’re aesthetically rejected, full-size carrots whittled down into more palatable shapes), but have never eaten a purple one. Would love to.

  9. Charlie Crystle

    Advertising–it depends on cost vs LTV, yes, but it also depends on each subscriber’s word of mouth value.For every new subscriber you get, how many new ones over what period of time does the subscriber attract through word of mouth? If it’s 5, then you’re in pretty good shape. If it’s .5, then not so much, but it can be as low as .25, right?Once you hit critical mass, you can ratchet the investment in growth back (assuming CM in this case means it grows on its own at a fast enough rate to pay the bills..I think I need more coffee)

    1. Dan Lewis

      The problem feels like a chicken and egg one. I don’t know how many new subscribers the average subscriber brings in via word of mouth (as in, I can’t really see that data so easily), and at the relatively small numbers I’m at anyway, it’s hard to say if that’s statistically significant.If I hit, say, 100k subs, different story. But at that point, I’ve already hit scale, so the “is this scalable” question because a tad moot.

      1. Charlie Crystle

        You can do it through a survey as long as you get a valid sample, whichshould be more than the minimum. You can define the questions so theyeffectively match what an automated system would show: number of people,frequency of sharing, methods for sharing, number of people you think signedup because of you…might be asking too much but maybe it will reveal a rateyou can compare against the knowns.After today, though, you’ll see a significant spike.

        1. Dan Lewis

          If I were running this more as a business than an hobby, I definitely would be running surveys to learn more about the readership. It’s great advice.

          1. Mark Essel

            You could do the surveys and pass on lessons learned to your subscriber audience. I’d read it.

          2. Jim Goodlett

            @DanDotLewis, good scribe indeed…the bit that seems to be missing on the discuss that you have started cottoning onto in your last reply to @CharlieCrystal, has to do with hard metrics on:a) audience engagement b) audience trending on a discussc) audience tools utilized to access contentd) number of views of content and embedded links on contente) best time of day to engage audience based also on type of content f) best platforms for delivery of contentg) reason for audience churnh) ad infinitumBy knowing more about these from a hard metric standpoint, engadgement and ROI can be increased exponentially if you measure (and utilize) data on whether your message is resonating past the simple look at number of subscribers that one holds…this will help you get more value out of your LTV and CPA numbers as well…

        2. awaldstein

          Or you can start simply by focusing shareability through discrete channels. Tweets and Shares (ala FB) through a stream or a page.Word of mouth = links and that is somewhat measurable

  10. Aaron Klein

    Dan, awesome post and very approachable. It’s amazing how many entrepreneurs, from high tech to my 16-year-old brother-in-law’s lawn maintenance startup, need a better understanding of this.To me, one of the most compelling things are tech products that inherently multiply the monetization potential of users. Facebook is a great example because you give it so much data, the monetization rates skyrocket.Eyeballs are one thing to monetize. Qualified eyeballs multiply that revenue.My goal with my new startup is to hit a LTV:CPA ratio of 2.5 to 1. I think we’ve got the product, technology and target industry to make that happen.(But to tie this into yesterday’s AVC post, we decided against doing a pre-product $41MM financing round and hiring 27 people to find out… :>)

    1. Mark Essel

      I messed this up badly with early advertising effort. The estimation process for lifetime value was too squishy. Now I see how important it is to get a good ball park for this.

      1. Aaron Klein

        So true. That may be the goal but you start slow and test extensively, while putting a ton of focus on the best kind of user acquisition – the $0 CPA!

  11. Guest

    This was literally taught in my MBA entrepreneurship class last week (by the guest speaker, a venture capitalist). MBA Mondays are the best deal on the internet – MBA content without the tuition.

  12. Denis

    There is one thing that I would add to the explanation on how to calculate LTV & CAC ratio: any discounts on prices must be taken into account by subtracting them from the LTV while the costs of running those promotions should be added to the CAC.Many small businesses are oblivious to this rule, and group buying sites take full advantage of the situation.

  13. David Quiec

    This not as easy as you think:1) At LTV=CPA upper limit, people usually miss out a lot of other expenses. Esp. in bigger companies where its harder to find all associated costs, both fixed and variable.2) Fixed costs/unit goes down as you acquire more users. So your CPA can go down as you scale.3) Cash flow is not addressed. If you have a 3 yr LTV, you don’t really want to spend all of it today. You’ll run out of money.

    1. Dan Lewis

      Yeah, it’s very blocking-and-tackling. All of what you note is very important.

    2. fredwilson

      great point about the cash flow aspects of this

    3. Guest

      David, thanks for bringing some logic to this. That is all.

    4. Alex Murphy

      You will only run out of money if you don’t have 3 years worth of reserve.

  14. ShanaC

    It sounds like this is arbitraging advertising….

    1. Dan Lewis

      Could be, but that’s more of a tactical flavor derived from the overall concept.

    2. John

      Couldn’t you easily say that any business is arbitrage (ie. selling something for more than it costs you)?

  15. Steve Hallock

    The problem with that analysis is that CPA is fixed in a unit of time. LTV is an estimate that can change (for better or worse).I would imagine that the estimated LTV for a facebook user in 2004, turned out to be dramatically lower than the actual LTV, especially taking into account the value of network effects created by the initial base of users.

    1. Dan Lewis

      It’s never going to be (anywhere near) perfect. But that doesn’t mean it’s not worthwhile. And chances are, your errors in estimation will complement each other. While the annual value of a Fb user are probably lower than expected, the length of time that user is on the site (in years, I mean) is probably higher; and the cost to acquire that user (due to how viral it is) is also probably lower. Net net, you may even come out ahead.

      1. awaldstein

        This goes to another point Dan.In your sample model, cost out for traffic in is negligible mostly so squishiness in the model doesn’t matter that much.But if the outlay (let’s say for PPC) is high for a small biz even if the return is good, you want to build your model on a shorter return to avoid spreading the model beyond comfort cash flow.

      2. Dan

        This is my first time reading a guest post on AVC, and I just wanted to say thanks for replying to the comments ala Fred. This is such an important piece so thanks for taking the time.

        1. Dan Lewis

          My pleasure! Really.

  16. ErikSchwartz

    LTV > CPA + cost of goodsYou can still go broke if LTV > CPA

    1. Dan Lewis

      That’s him.

  17. Tim Geisenheimer

    When calculating LTV, how do you factor indirect revenue accrued from various deals/partnerships you’ve been exposed to via your newsletter? Just as all CPA costs need to be considered, I would think as complete a picture as possible of the true LTV of your subscribers would be important for this model as well.

    1. Dan Lewis

      I don’t, in part because there haven’t been too many, but mostly because the value of the newsletter (beyond the hobby aspect) is the odd networking side effect it has.

  18. paramendra

    The pickle story was great, but the newsletter story is pretty good too.

  19. John

    Good writeup. Of course, there’s lots of interesting ways to increase the LTV and descrease the CPA. That’s the key to success.Although, one of the beauties of the internet is the free (see sweat equity) methods of advertising that are available. If the CPA approaches zero because you’re using the free methods, then it makes the equation work out really nicely.Yes, you could value your time in doing the free marketing methods, but when the alternative is watching a little more TV or spending some time using the free marketing methods, then that cost isn’t counted the same way. Plus, this is why I encourage choosing something you’re passionate about. Then, the TV won’t entice you and it’s hard to calculate the cost for you to do something you enjoy doing.

  20. Harsh Patel

    Dan, This is some good info, but at what point do you know when you’ve established your baseline?What about the case in which you’re building a company and deploying functionality as its built? The full product is there, but the revenue capabilities haven’t been built out yet.In our case, the monetization on our site is minimal until we release functionality that will encourage interaction and consequently increase revenue. Would you still approach this as costs < revenue? Or would you spend based on forecasted costs < forecasted revenue? In other words, if we’re earning $0.50 per user right now, are there circumstances in which it makes sense to spend more than the revenue / user?Or…would you save the acquisition costs and use free channels until you met your forecasted revenue / user?

    1. Dan Lewis

      If you’re building a business, you should probably base i off forecasted value of the user, not the current value. But there are other questions there, e.g. risk tolerance.Let’s say that you’re building a product with an estimated LTV of $100, but you’re not ready to monetize for some reason or another. You feel that if you can get to 10k users, you can go out and raise a few million dollars to go from 10k to 1MM pretty easily. So you decide to focus on building the product while you build that first 10k set. Without monetization tools built in, LTV is zero.If you’re relegating yourself to free channels, you’re probably being pennywise and pound-foolish. But if you’re spending $50 CPA, that’s $500k! You could almost certainly spend the money better elsewhere. I can say with the utmost of certainty that it’s something in between — and I realize that answer is 100% worthless to you.You really need to find that spot to taste. It’s going to be a function of available acquisitions channels, your risk posture, your available capital, and I’m sure a bunch of other factors I’m not seeing right now.

      1. Harsh Patel

        Hey Dan, Thanks for the reply. It seemed to me that there has to be a balance, and that you have to take risk with customer acquisition – especially upfront as a young company. That risk balance of course is dependent on the startup and its founders. Given that we’ve bootstrapped for a couple years now, we have a fair tolerance for risk.Based on your experience, at what point should a startup begin to have a good grasp on their user acquisition cost and LTV?

        1. Dan Lewis

          Never: It changes as your company and product mature.

  21. awaldstein

    Dan…I forgot to say “Thank you” for this post. Well done. 65 comments to date means you’ve struck a chord.Congrats!

    1. Dan Lewis

      Thanks 🙂 It helps that 20 of them are mine!

    2. fredwilson

      I’d like to thank dan for the post and the active engagement in thecomments. That’s what the author of a post on AVC is expected to do.



  23. jfccohen

    This is one of the most helpful, interesting and applicable guest posts I’ve read on here. Thanks, Dan, for the insight and thanks, Fred, for the foresight. I will be revisiting this one.

  24. Douglas Crets

    here’s how I work it out in my head, because I am not much of a financial whiz, though I really appreciate your thoughts on this:Great contentMany voicesSincerityAbsence of marketing BS disguised as contentTruthGreat network of people who can tell you what’s up and give feedbackAn aggressive editor who can interview people, get real news and information, and be accurateEqualsGrowing an audience incrementallyAnything you do that strays from that doesn’t get much support. The web audience can sniff out BS from Denmark. And if they are in Denmark, they certainly knows what it is that’s rotten.

  25. Rick Wingender

    Dan, Fred – Interesting post. Customer LTV is something I’ve paid a lot of attention to in my current and “past lives.”. I find that the most problematic thing with LTV is in defining the formula. If you have a simple business, then great. But if you have a complex business, like wireless phones or home TV shopping, it becomes MUCH more difficult. Everyone has a different idea on what costs should be included. Even COA (cost of acquisition) is debateable. How do you calculate “Lifetime” value when your business in 2 or 3 years old? Or when you anticipate significant product line changes / additions. I can come up with a bunch of reasons why it’s difficult to accurately calculate. And if you’re trying to compare two companies, you’d better be intimately familiar with how each does it – it won’t be the same, and it’s not something you’ll find in a Annual Report. Lastly, I have a book that I recommend which broaches this topic fairly well: “Marketing Metrics: 50+ metrics every executive should master” by Farris, Bendle, Pfeifer, and Reibstein.

  26. Clay Schossow

    This simple math is actually how many SaaS companies do their rounds of funding. They can demonstrate to investors that for every $15 in CPA, we generate $45 in LTV. Pretty sweet, right?! The problem is, we’re hitting a ceiling in what we can spend for CPA and still run the business, so give us $3 million to plug into this formula, hire some support for it, and we’ve already proven the type of returns you can see.If you can approach investors with a proven CPA/LTV correlation, you’re in a great position to easily explain the type of returns they can expect and the amount of $$ you need to make it happen.

  27. James M Cooper

    Hi Dan,Nice post, a straight forward explanation, for a very important point for all businesses.LTV – at first glance I was thinking ‘Loan To Value’ ratio. I had not used Life Time Value as an acronym before. Now, I have two LTV acronyms to deal out at parties.Thanks,James

  28. Mike Kijewski

    This assumes a fixed LTV of the customer. In nascent industries it seems intuitive that LTV would increase over time while CPA would go down as the industry matures. Therefore, an instantaneous comparison of LTV and CPA very early in a company’s / industry’s life may be misleading.What is the LTV of a Twitter user?

    1. fredwilson

      I’m expecting it to be a lot!

  29. Youssef Rahoui

    That’s where social media may come in in a powerful way: if people that receive the email are somehow enticed to share the content, CPA decreases while profits increase.BuzzFeed is building something really interesting and promising regarding this. You can’t just take a look at their advertising page:…. Their concept of viral lift that they sell to advertisers is really brilliant!

  30. BuyGiftsItems

    I guess the LTV estimate for a user of Facebook in 2004, proved to be much lower than the current LTV, especially taking into account the value of network effects created by the initial base of users.Kamagra