Category Archives: SaaS

Video of My SaaStr 2020 Presentation: Churn is Dead, Long Live Net Dollar Retention

Thanks to everyone who attended my SaaStr 2020 presentation and thanks to those who provided me with great feedback and questions on the content of the session.  The slides from the presentation are available here.  The purpose of this post is to share the video of the session, courtesy of the folks at SaaStr.  Enjoy!

 

Appearance on the CFO Bookshelf Podcast with Mark Gandy

Just a quick post to highlight a recent interview I did on the CFO Bookshelf podcast with Mark Gandy.  The podcast episode, entitled Dave Kellogg Address The Rule of 40, EPM, SaaS Metrics and More, reflects the fun and somewhat wandering romp we had through a bunch of interesting topics.

Among other things, we talked about:

  • Why marketing is a great perch from which to become a CEO
  • Some reasons CEOs might not want to blog (and the dangers of so doing)
  • A discussion of the EPM market today
  • A discussion of BI and visualization, particularly as it relates to EPM
  • The Rule of 40 and small businesses
  • Some of my favorite SaaS operating metrics
  • My thoughts on NPS (net promoter score)
  • Why I like driver-based modeling (and what it has in common with prime factorization)
  • Why I still believe in the “CFO as business partner” trope

You can find the episode here on the web, here on Apple Podcasts, and here on Google Podcasts.

Mark was a great host, and thanks for having me.

SaaStr 2020 Session Preview: Churn is Dead, Long Live Net Dollar Retention!

Capture

Reunited with old friend Tracy Eiler on the speaker page

The SaaStr Annual conference was delayed this year, but Jason & crew know that the show must go on.  So this year’s event has been rechristened SaaStr Annual @ Home and is being held in virtual, online format on September 2nd and 3rd.  The team at SaaStr have assembled a strong, diverse line-up of speakers to provide what should be another simply amazing program.

The purpose of this post is to provide a teaser to entice you to attend my session, Churn is Dead, Long Live Net Dollar Retention Rate, bright and early on Wednesday, September 2nd at 8:00 AM.

“I eat SaaS metrics for breakfast,” he thinks.  Or at least, “with.”

In this session, we’ll cover:

  • Separating a SaaS business into its two component parts
  • What makes SaaS companies so interesting for PE buyers
  • The SaaS leaky bucket of ARR
  • SaaS unit economics 101:  CAC, LTV, LTV/CAC, and CAC payback period
  • The three, fairly lethal problems with churn rates
  • Why “ARR is a fact and churn is an opinion”
  • Cohort analysis basics and survivor bias
  • Net dollar retention (NDR) rate definition and benchmarks
  • Explanatory power of NDR vs. ARR growth and the Rule of 40 in determining valuation multiples
  • The NDR implications of Goodhart’s Law
  • Applying Goodhart’s Law to NDR
  • The next frontier:  remaining performance obligation (RPO)

While the topic might seem a little dry, the content is critically important to any SaaS executive, and I can assure you the presentation will be fast-paced, fun, and anything but dry.

I hope you can attend and I look forward to seeing you there.

Are We Due for a SaaSacre?

I was playing around on the enterprise comps [1] section of Meritech‘s website today and a few of the charts I found caught my attention.  Here’s the first one, which shows the progression of the EV/NTM revenue multiple [2] for a set of 50+ high-growth SaaS companies over the past 15 or so years [3].

meritech saas multiples

While the green line (equity-value-weighted [4]) is the most dramatic, the one I gravitate to is the blue line:  the median EV/NTM revenue multiple.  Looking at the blue line, you can see that while it’s pretty volatile, eyeballing it, I’d say it normally runs in the range between 5x and 10x.  Sometimes (e.g., 2008) it can get well below 5x.  Sometimes (e.g., in 2013) it can get well above 10x.  As of the last data point in this series (7/14/20) it stood at 13.8x, down from an all-time high of 14.9x.  Only in 2013 did it get close to these levels.

If you believe in regression to the mean [5], that means you believe the multiples are due to drop back to the 5-10 range over time.  Since mean reversion can come with over-correction (e.g., 2008, 2015) it’s not outrageous to think that multiples could drop towards the middle or bottom of that range, i.e., closer to 5 than 10 [6].

Ceteris paribus, that means the potential for a 33% to 66% downside in these stocks. It also suggests that — barring structural change [7] that moves baseline multiples to a different level — the primary source of potential upside in these stocks is not continued multiple expansion, but positive NTM revenue surprises [8].

I always love Rule of 40 charts, so the next fun chart that caught my eye was this one.  meritech r40 score While this chart doesn’t speak to valuations over time, it does speak to the relationship between a company’s Rule of 40 Score and its EV/NTM revenue multiple.  Higher valuations primarily just shift the Y axis, as they have done here, uplifting the maximum Y-value by nearly three times since I last blogged about such a chart [9].  The explanatory power of the Rule of 40 in explaining valuation multiple is down since I last looked, by about half from an R-squared of 0.58 to 0.29.  Implied ARR growth alone has a higher explanatory power (0.39) than the Rule of 40.

To me, this all suggests that in these frothy times, the balance of growth and profit (which is what Rule of 40 measures) matters less than other factors, such as growth, leadership, scarcity value and hype, among others.

Finally, to come back to valuation multiples, let’s look at a metric that’s new to me, growth-adjusted EV/R multiples.

meritech r40 growth adjusted

I’ve seen growth-adjusted price/earnings ratios (i.e., PEG ratios) before, but I’ve not seen someone do the same thing with EV/R multiples.  The basic idea is to normalize for growth in looking at a multiple, such as P/E or — why not — EV/R.  For example, Coupa, trading at (a lofty) 40.8x EV/R is growing at 21%, so divide 40.8 by 21 to get 1.98x.  Zoom, by comparison looks to be similarly expensive at 38.3x EV/R but is growing at 139%, so divide 38.3 by 139 to get 0.28x, making Zoom a relative bargain when examined in this light [10].

This is a cool metric.  I like financial metrics that normalize things [11].  I’m surprised I’ve not seen someone do it to EV/R ratios before.  Here’s an interesting observation I just made using it:

  • To the extent a “cheap” PE firm might pay 4x revenues for a company growing 20%, they are buying in at a 0.2 growth-adjusted EV/R ratio.
  • To the extent a “crazy” VC firm might pay 15x revenues for a company growing at 75%, they are buying in at a 0.2 growth-adjusted EV/R ratio.
  • The observant reader may notice they are both paying the same ratio for growth-adjusted EV/R. Given this, perhaps the real difference isn’t that one is cheap and the other free-spending, but that they pay the same for growth while taking on very different risk profiles.

The other thing the observant reader will notice is that in both those pseudo-random yet nevertheless realistic examples, the professionals were paying 0.2.  The public market median today is 0.7.

See here for the original charts and data on the Meritech site.

Disclaimer:  I am not a financial analyst and do not make buy/sell recommendations.  I own positions in a wide range of public and private technology companies.  See complete disclaimers in my FAQ.

# # #

Notes 
[1] Comps = comparables.

[2] EV/NTM Revenue = enterprise value / next twelve months revenue, a so-called “forward” multiple.

[3] Per the footer, since Salesforce’s June, 2004 IPO.

[4] As are most stock indexes. See here for more.

[5] And not everybody does.  People often believe “this time it’s different” based on irrational folly, but sometimes this time really is different (e.g., structural change).  For example, software multiples have structurally increased over the past 20 years because the underlying business model changed from one-shot to recurring, ergo increasing the value of the revenue.

[6] And that’s not to mention external risk factors such as pandemic or election uncertainty.  Presumably these are already priced into the market in some way, but changes to how they are priced in could result in swings either direction.

[7] You might argue a scarcity premium for such leaders constitutes a form of structural change. I’m sure there are other arguments as well.

[8] To the extent a stock price is determined by some metric * some multiple, the price goes up either due to increasing the multiple (aka, multiple expansion) or increasing the metric (or both).

[9] While not a scientific way to look at this, the last time I blogged on a Rule of 40 chart, the Y axis topped out at 18x, with the highest data point at nearly 16x.  Here the Y axis tops out at 60x, with the highest data point just above 50x.

[10] In English, to the extent you’re paying for EV/R multiple in order to buy growth, Zoom buys you 7x more growth per EV/R point than Coupa.

[11] As an operator, I don’t like compound operational metrics because you need to un-tangle them to figure out what to fix (e.g., is a broken LTV/CAC due to LTV or CAC?), but as investor I like compound metrics as much as the next person.

 

The Pipeline Chicken or Egg Problem

The other day I heard a startup executive say, “we will start to accelerate sales hiring — hiring reps beyond the current staffing levels and the current plan — once we start to see the pipeline to support it.”

To mix metaphors, what comes first: the pipeline or the egg?  To un-mix them, what comes first:  the pipeline or the reps to prosecute it?  Unlike the chicken or the egg problem, I think this one has a clear answer: the reps.

My answer comes part from experience and part from math.

First, the experience part:  long ago I noticed that the number of opportunities in the pipeline of a software company tends to be a linear function of the number of reps, with a slope in the 12-18 range as a function of business model [1].  That is, in my 12 years of being a startup CEO, my all-quarters, scrubbed [2] pipeline usually had somewhere between 12 and 18 opportunities per rep and the primary way it went up was not by doing more marketing, but by hiring more reps.

Put differently, I see pipeline as a lagging indicator driven by your capacity and not a leading indicator driven by opportunity creation in your marketing funnel.

Why?  Because of the human factor:  whether they realize it or not, reps and their managers tend to apply a floating bar on opportunity acceptance that keeps them operating around their opportunity-handling capacity.  Why’s that?  It’s partially due to the self-fulfilling 3x pipeline prophecy:  if you’re not carrying enough pipeline, someone’s going to yell at you until you do, which will tend to drop your bar on opportunity acceptance.  On the flip side, if you’re carrying more opportunities than your capacity — and anyone is paying attention — your manager might take opportunities away from you, or worse yet hire another rep and split your territory.  These factors tends to raise the bar, so reps cherry pick the best opportunities and reject lesser ones that they’d might otherwise accept in a tougher environment.

So unless you’re running a real machine with air-tight definitions and little/no discretion (which I wouldn’t advise), the number of opportunities in your pipeline is going to be some constant times the number of reps.

Second, the math part.  If you’re running a reasonably tight ship, you have a financial model and an inverted funnel model that goes along with it.  You’re using historical costs and conversion rates along with future ARR targets to say, roughly, “if we need $4.0M in New ARR in 3 quarters, and we insert a bunch of math, then we’re going to need to generate 400 SALs this quarter and $X of marketing budget to do it.”  So unless there’s some discontinuity in your business, your pipeline generation doesn’t reflect market demand; it reflects your financial and demandgen funnel models.

To paraphrase Chester Karrass, you don’t get the pipeline you deserve, you get the one you plan for.  Sure, if your execution is bad you might fall significantly short on achieving your pipeline generation goal.  But it’s quite rare to come in way over it.

So what should be your trigger for hiring more reps?  That’s probably the subject of another post, but I’d look first externally at market share (are you gaining or losing, and how fast) and then internally at the CAC ratio.

CAC is the ultimate measure of your sales & marketing efficiency and looking at it should eliminate the need to look more deeply at quota attainment percentages, close rates, opportunity cost generation, etc.  If one or more of those things are badly out of whack, it will show up in your CAC.

So I’d say my quick rule is if your CAC is normal (1.5 or less in enterprise), your churn is normal (<10% gross), and your net dollar expansion rate is good enough (105%+), then you should probably hire more reps.  But we’ll dive more into that in another post.

# # #

Notes

[1]  It’s a broad range, but it gets tighter when you break it down by business model.  In my experience, roughly speaking in:

  • Classic enterprise on-premises ($350K ASP with elephants over $1M), it runs closer to 8-10
  • Medium ARR SaaS ($75K ASP), it runs from 12-15
  • Corporate ARR SaaS ($25K ASP) where it ran 16-20

[2] The scrubbed part is super important.  I’ve seen companies with 100x pipeline coverage and 1% conversation rates. That just means a total lack of pipeline discipline and ergo meaningless metrics.  You should have written definitions of how to manage pipeline and enforce them through periodic scrubs.  Otherwise you’re building analytic castles in the sand.

Why You Should Eliminate the Title “Implementation Consultant” from Your Startup

I’ve worked with several startups that fell into the following pattern:

  • Selling a SaaS application at a healthy price (e.g., $100K to $200K ARR)
  • With low, fixed-cost implementation packages (e.g., $25K)
  • But a product that actually takes maybe $50K to $75K to successfully deploy
  • Resulting in an unprofitable professional services business (and wrecking the market for partner services)
  • High adoption failure
  • And, depending on the initial contract duration, high customer churn [1]

For example, one company had a CAC of 4.0, churn of 25%, and services margins of negative 66% when I started working with them [2].  Ouch.

Before proceeding, let me say that if you have a low-touch, high-velocity, easy-adoption business model — and the product to go with it — then you don’t need to read this post [3].  If you don’t, and any of the above problems sound familiar, then let’s figure out what’s going on here and fix it.

The problem is the company is not charging the appropriate price for the services needed.  Perhaps this is because of a zero-sum fallacy between ARR and services.  Or perhaps they feel that customers “just won’t pay” that much for implementation services.  Or perhaps their product takes more work to deploy than the competition and they feel forced to match price on services [4].

This under-pricing usually triggers a number of other problems:

  • In order to work within the self-created, low-cost implementation services model, the company “hires cheap” when it comes to implementation consultants, preferring junior staff and/or staff in offshore locations.
  • The company’s “implementation consultants” are overloaded, working on too many projects in parallel, and are largely focused more on “getting onto the next one” than getting customers successfully implemented.
  • Once a certain number of hours are clocked on any given project, the consultants go from “in a hurry” to “in a big hurry” to finish up and move on.
  • Customers are left high-and-dry with failed or partial implementations that, if left unfinished, will likely lead to churn.
  • Customer success, whose job is to prevent churn, is left holding the bag and is pulled away from its primary mission of adoption, renewal, and expansion into the implementation-completion business, potentially changing its hiring profile from more sales-oriented to more product-oriented and/or complementing CSMs with customer success architects (CSAs) or technical account managers (TAMs) to try and fill the implementation void.

I sometimes consider fixing this corporate chiropractor work, because one maladjustment results in the whole organization being twisted out of shape [5].  The good news is that, as with chiropractors, one adjustment can pop the whole system back into alignment.

Now, before we move onto fixing this, there’s one more problem we haven’t discussed yet — and give yourself ten pats on the back if you figured out before I got here:

Who ever said the customer defined success as getting the software implemented?

Oh shit.  We were so tied up trying to deliver a $25K services package that costs $40K to deliver that we forgot about the customer.  What customer equates implementation with success?  None.  Zero.  Nobody.

“Hey, it’s all set up now, you can login, gotta go!” is not the credo of a success-oriented consultant.

But what do we call our consultants again?  Implementation consultants.

What do implementation consultants think they do?  Well, implementations.

When an implementation consultant reads their own business card, what does it tell them they their job is?  Implementations.

Are implementations what customers want?  No.

So why do we have implementation consultants again?  I have no idea.

What do customers what?  Overall they want success, but what’s a good proxy?  How about attaining their first business objective?  If you sell:

  • A recruiting app, running your first recruiting campaign
  • A financial planning app, it’s making your first plan
  • A demandgen marketing app, it’s running your first demandgen campaign
  • A customer service app, it’s your first day running the call center
  • A deflection app, it’s deflecting your first cases
  • A sales enablement app, it’s training your first reps
  • An IT support app, it’s handing your first tickets

So, what’s the fix here?  While not all of this will be possible or recommended in all situations, here’s the long list:

  • Re-frame services as in the success business, not the implementation business
  • Eliminate the job title implementation consultant in favor of consultant
  • Get services to make plans that end not with implementation, but with the achievement of an agreed-to first business objective.
  • Increase your services pricing, if needed, so they can both deliver success and break even.
  • Hire more experienced consultants who can better make customers successful and don’t be afraid to charge more for them.  (They’re worth it.)
  • Agree to an ARR price before negotiating the services price; refuse to trade one off against the other.
  • Involve your services team in the sale well before the contract is signed so they propose the right prix fixe package (e.g., small, medium, large) or create an appropriately-sized bespoke statement of work.
  • Modify your product so it is not at a competitive disadvantage on required implementation work.

# # #

Notes
[1] With one-year contracts, a failed implementation that takes 6-9 months to fail typically results in churn, whereas with three-year contracts, you will often get another swing at the problem.

[2] These horrific unit economics result in an LTV/CAC of 1.0 and make the company totally uninvestable.  The CAC would be even higher if hard-ass investor added the services losses back into the CAC on the theory they were subsidizing sales.

[3] Product-led growth business models are great, but when companies that are not designed for them try to emulate pieces of the business model, they can get into trouble.  Implementation is an area that quickly goes awry when companies not built for PLG attempt bottom-up, try-and-buy, viral go-to market strategies.

[4] In which case, an obvious solution is to reduce the deployment workload requirements of the product.

[5] Put differently, the sales bone is connected to the services bone, and the services bone is connected to the customer success bone.

Measuring Ramped and Steady-State Sales Productivity: The Rep Ramp Chart

In prior posts I have discussed how to make a proper sales bookings productivity model and how to use the concept of ramped rep equivalents (RREs) in sales analytics and modeling. When it comes to setting drivers for both, corporate leaders tend to lean towards benchmarks and industry norms for the values.  For example, two such common norms are:

  • Setting steady-state (or terminal) productivity at $1,200K of new ARR per rep in enterprise SaaS businesses
  • Using a {0%, 25%, 50%, 100%} productivity ramp for new salesreps in their {1st, 2nd, 3rd, 4th} quarters with the company (and 100% thereafter)

In this post, I’ll discuss how you can determine if either of those assumptions are reasonable at your company, given its history.

To do so, I’m introducing one of my favorite charts, the Rep Ramp Chart.  Unlike most sales analytics, which align sales along fiscal quarters, this chart aligns sales relative to a rep’s tenure with the company.

You start by listing every rep your company has ever hired [1] in order by hire date.  You then record their sales productivity (typically measured in new ARR bookings [2]) for their series of quarters with the company [3], up to and including their current-quarter forecast (which you shade in green).  Reps who leave the company are shaded black.  Reps who get promoted out of quota-carrying roles (e.g., sales management) are shaded blue.  Future periods are shaded grey.  Add a 4+ quarter average productivity column for each row, and average each of the figures in the columns [4].

Here’s what you get:

full

Despite having only a relatively small amount of data [5], we can still interpret this a little.

  • The relative absence of black lines means we’re pretty good at sales hiring.   I’ve seen real charts with 5 black lines in a row, usually down to a single bad management hire.
  • The absence of black lines that “start late”  — for example {0, 25, 75, 25, 55, black} — is also good.  Our reps are either “failing fast” or succeeding, but things are not dragging on forever when they’re not working.
  • Over average 4Q+ productivity is $308K per quarter, almost exactly $1,200K per year so it does seem valid to use that figure in our modeling.
  • Entering $300K as target productivity then shows the empirical rep ramp as a percent of steady-state productivity, exactly how sales leaders think of it.  In this case, we see a {10%, 38%, 76%, 85%, 98%} empirical ramp across the first five quarters.  If our bookings model assumed {0%, 25%, 50%, 100%, 100%} you’d say our model is a little optimistic in the first two quarters, a little pessimistic in the 3rd, and a little optimistic in the fourth.  If we had more data, we might adjust it a bit based on that.

I love this chart because it presents unadulterated history and lets you examine the validity of two hugely important drivers in your sales bookings capacity model — drivers, by the way, that are often completely unquestioned [6].  For that reason, I encourage everyone to make this a standard slide in your Sales ops review (aka, QBR) template.  Note that since different types of rep ramp differently and hit different steady-state productivity levels, you should create one rep ramp per major type of rep in your company.  For example, corporate (or inside) sales reps will typically ramp more quickly to lower productivity levels than field reps who will ramp more slowly to higher productivity.  Channels reps will ramp differently from direct reps.  International reps may need their own chart as well.

You can download the spreadsheet I used here.

# # #

Notes

[1] Sales management may want to omit those no longer with the company, but that also omits their data, and might omit important patterns of hiring failure, so don’t omit anyone.  You can always exclude certain rows from the analysis without removing them from the chart (i.e., hiding them).

[2] New ARR bookings typically includes new ARR to both new and existing customers.

[3] You’ll need as many columns to do this as your longest tenured rep has been with the company, so it can get wide.  Let it.  There’s data in there.

[4] Ensuring empty cells are not confused with cells whose value is zero.  Excel ignores empty cells in calculating averages but will average your 0’s in when you probably don’t want them.

[5] In order to keep it easily and quickly grasped

[6] Particularly the ramp.