Kellblog covers topics related to starting, leading, and scaling enterprise software startups including company strategy, financing strategy, go-to-market strategy, sales, marketing, positioning, messaging, and metrics
I’m always looking for better ways to distill strategy. My favorite strategy author is Richard Rumelt, who wrote Good Strategy, Bad Strategy and the more recent but less acclaimed follow-on, The Crux.
I love Rumelt’s work for two reasons:
He takes a wrecking ball to the garbage that is often passed off as strategy. Aspirations are not strategy. Goals and OKRs are not strategy. Financial projections and forecasts are not strategy. SWOT analyses and five forces analyses are not strategy. Driving results is not strategy. Deciding to be a butcher, baker, or candlestick maker is not strategy. You may, like me, find reading these takedowns not only educational, but therapeutic.
He whittles strategy down to the head of a pin. First, by defining strategy as identifying and planning to overcome a company’s most important challenge (aka, challenge-driven strategy). Then, by capturing what he calls the kernel of strategy: a diagnosis, a guiding policy, and a set of coherent actions.
Much as I love the kernel idea, in one assignment a few years back we tried to apply this framework and stumbled into a problem. We arrived at a diagnosis fairly easily, but got stuck trying to create a guiding policy. We found that the diagnosis alone wasn’t enough to arrive at a guiding policy. We kept needing to insert a few assumptions (or beliefs) about the future before we could agree on a guiding policy. We drifted to a modified framework that looked like this:
Given diagnosis X,
And beliefs Y,
We choose guiding policy Z,
And coherent actions 1-5 to implement it.
I was so excited with this discovery that I emailed Rumelt. While he kindly did reply, I don’t think my point landed. He directed me to his then-upcoming book and suggested it would be addressed there. The Crux was subsequently published and I don’t think it was. Never meet your heroes, as Flaubert wrote, a little gold always rubs off when you do.
Undeterred, I continued to use Rumelt’s framework, but added beliefs as an explicit part. I’ve always felt that diagnosis was by far the hardest part of strategy, as I believe does Rumelt, given this excerpt from his first book:
“After my colleague John Mamer stepped down as dean of the UCLA Anderson School of Management, he wanted to take a stab at teaching strategy. To acquaint himself with the subject, he sat in on ten of my class sessions. Somewhere around class number seven we were chatting about pedagogy and I noted that many of the lessons learned in a strategy course come in the form of the questions asked as study assignments and asked in class. These questions distill decades of experience about useful things to think about in exploring complex situations. John gave me a sidelong look and said, “It looks to me as if there is really only one question you are asking in each case. That question is ‘What’s going on here?’ ” John’s comment was something I had never heard said explicitly, but it was instantly and obviously correct. A great deal of strategy work is trying to figure out what is going on. Not just deciding what to do, but the more fundamental problem of comprehending the situation.”
I believe Rumelt would say that what I call beliefs are simply part of the diagnosis. For example, he said, “Netflix’s overall challenge (in 2018) was that it could no longer count on contracting for existing good TV and studio films at reasonable prices.” I’d argue that Rumelt’s Netflix diagnosis is actually two statements in one. Writing from the viewpoint of Netflix:
That today we find ourselves increasingly hit with large price increases and/or a non-desire to renew distribution agreements for content.
We believe the vast majority of the content producers will enter the content distribution business via streaming services in the next few years and ergo will not want or need to work with us.
First, that’s one hell of a “gnarly challenge” as Rumelt likes to call the crux issue. Second, I like splitting it because, particularly when working with a good-sized group to build strategy, it helps to distinguish between what we are seeing right now versus what we anticipate in the future. The former are facts, the latter are beliefs — and most of the interesting debate is not about the facts, but the beliefs.
I was happy with this modified framework until a strange thing happened the other day. I was talking with a founder and — lightbulb moment — I realized I could further distill strategy simply by looking only at beliefs. Not a laundry list of them (which can easily get generated in such a process), but what I call the primary belief, the big one, the one that resolves the crux issue and drives all the rest.
I immediately tried to apply this idea to my experience at Business Objects where, for nearly a decade, I worked as one of the top executives as we grew the company from $30M to $1B. I found it was pretty easy to divide 16 years of history into four eras categorized by primary belief:
Era 1 (5 years). We believe customers will pay 5x the price of commodity query and reporting (Q&R) tools for an enterprise solution.
Era 2 (4 years). We believe that Q&R and online analytical processing (OLAP) tools should be integrated in one product.
Era 3 (4 years). We believe the Internet will require a wholesale rewrite of business intelligence (BI) and enable both existing internal and new external use-cases.
Era 4 (3 years). We believe that customers will increasingly want to buy an integrated suite of BI tools, including Q&R, OLAP, and enterprise reporting.
These beliefs were largely heretical at the time. $500/seat for a Q&R tool? Insane. Integrating Q&R and OLAP? Can’t be done (and “they” were nearly right). Extranet BI? Never, corporate data is highly proprietary. BI suites? No, customers still want best of breed!
But those four beliefs took us from $0 to $1B in revenues. The beliefs alone are not enough, of course. You need to build strategies (e.g., product, go-to-market) and execute against them. In era 1, we needed a highly targeted strategy to break into the market with this radical idea. In era 2, we needed to build and market the integrated product. In era 3, we needed to devise the right web product strategy, a task that befuddled several of our competitors.
But I can and will argue that it all flowed from the underlying primary belief.
I worked with Alation in various capacities for many years, so I feel I know their evolution pretty well. Let me try the same exercise, as an outsider looking in, separating Alation’s history into three eras and assign a name to each:
Era 1 (search and discovery). We believe that companies will need a centralized data catalog to help people find the data they need, and that machine-learning can help with that finding.
Era 2 (data governance). We believe that data catalogs (almost surprisingly) turn out to be an ideal tool for data governance, particularly the non-invasive variety.
Era 3 (data intelligence platform). We believe that customers will increasingly want to buy a data intelligence platform that includes data search & discovery, governance, and lineage.
I’m probably missing the company’s strong commitment to cloud platforms as part of era 3 and there may be a new era 4, but you get the idea. Again these beliefs were often heretical at the time. A lot of people didn’t believe data catalogs were even needed. Most people believed data governance was a distinct category and that the “prevent access” ethos of data governance ran strongly counter to the “enable access” ethos of data catalogs. Until recently, many people didn’t believe in data intelligence platforms (but with help from IDC and Databricks that debate has been put to bed). Again, beliefs alone are not enough. There are numerous also-ran data catalog companies who presumably shared some of these beliefs, but built the wrong strategies in response or lacked Alation’s relentless drive in execution.
I often say that strategy is best analyzed in reflection. Meaning that somehow everything is clearer and simpler when you look back 10 or 20 years to reflect upon what happened. In fact, I often encourage people to do a future look-back when formulating strategy: “imagine it’s ten years from now and your company won in the market — now tell me why.”
Add beliefs to the framework. More precisely, separate the diagnosis into present truths and future beliefs.
Work to find the one primary belief for your current situation. If you’re a new startup, that belief is probably embedded in the answer to, “why did you found the company?” If you’ve been around for a while, start by analyzing your history and trying to break it into belief-driven eras.
Once you’ve found a potentially era-defining primary belief, resume the Rumelt exercise: define guiding policy and coherent actions around it.
“If a bird walks like a duck, swims like a duck, and quacks like a duck, I call that bird a duck.” — James Whitcomb Riley
Many marketers are in such a hurry to talk about topical issues that they forget the duck test: if it walks like a duck, swims like a duck, and quacks like a duck, then most people will conclude it’s a duck. Philosophers teach that such abductive (or should we say, abducktive) reasoning can lead to incorrect conclusions — and it can.
But here in marketing, we draw a different conclusion from the duck test. It’s how most peoples’ minds work so we shouldn’t fight against it. There are two common ways that marketers fail the duck test and we’ll cover both of them — and what to do instead — in this post.
Deny Thy Father and Refuse Thy Name Many marketers are eager to pretend that their product is the latest in-vogue thing (e.g., AI), and can get so busy dressing it up in the latest tech fashion, that they generate more confusion than sales opportunities.
It’s like a replay of the clichéd movie scene: Man: Who are you? Woman: Who do you want me to be, baby?
When someone asks your company the equivalent of “who are you?” [1], you need to answer the question and that answer needs to be clear.
Remember, the enemy for most startups isn’t the competition. It’s confusion. The easiest thing for a prospect to do is nothing. If we talk and I leave confused, then I’ll just write off the wasted half-hour and go on with my day.
Consider an answer like this [2] [3], to the question “what is MarkLogic?”
I mean great question. We ask ourselves that all the time. It’s actually hard to answer because there’s nothing else like it. Answering that is like trying to explain the difference between a Cessna and a 747 to someone who’s never seen an airplane. Our marketing people call it an XML Server, but that’s not a great description.
What is it really? Literally, it’s what you get when you lock two search engine PhDs in a garage for two years and tell them to build a database. You know, it looks like a database from the outside, but when you pop open the hood — surprise — you find that it’s built from search engine parts. Search engine style indexing. And it’s schema-free like a search engine so it can handle unstructured, semi-structured, and, of course, structured data as well. Let’s get into those exciting distinctions in a minute.
This thing — whatever you want to call it — it’s the Vegomatic of a data: it slices and dices and chops in every conceivable way. In the end, I think what makes it hard to understand is that it’s basically a hybrid: half search engine, half content application platform, and all database.
Is that clear?
As mud. What’s wrong with that answer?
It’s confusing
It’s long
It’s navel-gazing (let’s talk about me)
It’s bleeding on the customer (sharing internal troubles)
It’s a horrible, horrible answer.
Now before you stop reading, perhaps thinking that this is one specific, dated case study, let me say that I could easily write such a parody for about a quarter of the few score of startups I work with today. This is not some ancient example from another world. This is a current problem for many startups, but I’m not going to parody any of them here [4]. Might you suffer from this problem? Go listen to some Gong or Chorus recordings, particularly high funnel (e.g., SDR) and/or discovery calls, and see if anything resonates.
Now, let’s contrast the previous answer with this one:
It’s an XML database system, meaning it’s a database that uses XML documents as its native data modeling element. Now, what did you want to do with it again?
What’s nice about this answer?
It’s short
It’s clear
It’s correct
It leaves an opportunity for follow-up questions [5]
But the really nice part of this answer is that it puts focus back on the customer. The direct cost of all the previous blather is confusion. The opportunity cost of all that blather is you waste precious time you could have spent listening to the customer, learning more about their problem, and trying to decide if you can solve it.
So why didn’t some of our sellers want to give the second answer? They didn’t want to say the X word. XML was cool for a while, but that quickly passed and XML databases were always distinctly uncool. So, some sellers would rather spend five minutes tap dancing around the question rather than directly answering it.
What followed was almost always a difficult conversation [6]. But the flaw in tap-dancing was simple: the customer is going to figure it out anyway [7]. Customers are smart. If it:
Stores data like a database
Builds indexes like a database
And has a query language like a database
Then — quack, quack — it’s a database.
That’s the first way marketers fail the duck test. They’re afraid to say what the product is for fear of scaring people off. But there’s another way to fail the duck test.
Confusing Products and Solutions The second way to fail the duck test is to rotate so hard to solutions that you basically refuse to say what the product is. You end up dodging the question entirely.
Customer: So, what is it? Vendor: You can use it to build things, like a deck. Customer: That’s great, but what is it? Vendor: You can use it to assemble things, too, like a bed. Customer: Sure, but what is it? Vendor: And you can use it for disassembling things too. Customer: Wait, it’s a drill isn’t it?
Here we have the prospect playing twenty questions to figure out what the product is. Yes, we all know that customers buy solutions to problems [8] and Theodore Levitt’s classic example of customers buying 1/4″ holes, not 1/4″ bits.
But don’t take that in a fundamentalist way. If the customer asks, “what is it?” the answer is not, “a thing that makes holes” but, “a power drill with a 1/4-inch bit.” If they ask why ours is better, we say that our bits are titanium and don’t break. “Feature” need not be a four-letter word to remember that the purpose of the drill is to make a hole and, transitively, that the purpose of the hole is to build a new deck with the ultimate benefit of quality family time.
The point is: knowing what solutions (or use-cases) we want to target does not eliminate the requirement to have strong product messaging. Particularly in unexciting categories, we will need to lead with use-cases, not product superiority, category formation, or market leadership. But, inevitably, even when you lead with use-cases, you will get the question: what is it?
And a short, clear answer – as we discussed above – not only gets the customer what they want, but it lets us have more time for listening and discovery. I see many companies where they rotate so hard to use-case marketing that their product messaging is so weak it actually interferes with discussions of the use-case.
For example, say the product is a data streaming platform (DSP) and the use-case is industrial monitoring for manufacturing facilities. Let’s assume that data streaming platforms are not a hot category, so there aren’t a lot of people out shopping for them. That means we’re not going to target DSP shoppers with a product-oriented superiority message, instead, we are going to target people who have a problem with industrial monitoring.
But when one of those people asks what it is, we’re not going to say, “a thingy that helps you do industrial monitoring.” Instead, we’re going to say, “it’s a data streaming platform, many of our customers use it for industrial monitoring, and here’s why it’s such a great fit for that use-case.”
That is, we map to the use-case. We don’t redefine the product around the use-case. We don’t try to use the use-case to avoid talking about the product. Doing so only confuses people because eventually they figure out it’s not an industrial monitoring application, but a data streaming platform that can be used for industrial monitoring. Unless we are clear that it’s a platform being used for a use-case, then we fail the duck test.
In the end, you will get the right answer if you always remember three things:
Customers are smart
Time spent in hazy product explanations confuses customers and robs time from discovery
If it walks like a duck, swims like a duck, and quacks like a duck — then, for marketing purposes at least — it’s a duck.
# # #
Notes
[1] That is, “what is it?”
[2] I swear this is only partially dramatized, and only because I’ve assembled all the fragments into one single response.
[3] This is circa 2008. Presumably much has changed in the intervening 15 years.
[4] I obviously don’t use more recent examples as a matter of both confidentiality and discretion.
[5] An obvious one might be, “so if it’s a database, does it speak SQL?” (To which the answer was “no, it speaks XQuery,” which could lead to another loop of hopefully tight question/answer follow-ups.)
[6] Because, simply put, nobody wanted to buy an XML database. Gartner had declared the category stillborn around 2002 with a note entitled XML DBMS, The Market That Never Was. The way we sold nearly $200M worth of them (cumulatively) during my tenure was not to sell the product (that nobody wanted) but to sell the problems it could solve.
[7] And when they do, they’re not going to be happy that you seemingly tried to deceive them.
[8] Or hire them to do jobs for them, if you prefer the Jobs To Be Done framework.
I’ve realized that one of things I do for (or should I say, to) early-stage startups is detect whether they have a real or a faux focus (pronounced fo-focus) — the latter being a focus that appears to be real at first, but is in fact fake.
But, alas, when you drill in, the conversation often goes something like this:
DK: So what are you focused on? CEO: Mid-market (MM). DK: Do you sell anywhere else? CEO: Yes, for sure. A lot of today’s SMB companies will be mid-market tomorrow, so we want to catch them on the way up. CEO: Oh, and today’s MMs will be enterprises (ENT) one day, so we sell to them too — so we can learn their product requirements –and our customers can grow without fear. I mean, we’d never want a customer to outgrow our platform. DK: OK, but what about our marketing, do we have any focus there? CEO: Well, you know, you can’t really control who clicks on a Google ad, so we can get inbound from SMB, MM, and ENT. We can control geography though. We don’t run ads in Florin or Genovia. DK: But those are fictional countries from princess movies. How about LinkedIn, you get a lot more control there? CEO: We’re not doing as much on LinkedIn. Leads are more expensive. DK: (Thinking that maybe they’re higher quality and more aligned to strategy.) How about industries? CEO: No, we’re a horizontal play. We have customers in banking, healthcare, pharma, insurance, energy, and retail/CPG. And many others. DK: Any focus on use-cases? CEO: Yes. But no. No, we’re a platform play. That’s the beauty. We have an integrated platform that can support dozens of strategic use-cases. So, really, yes and no. DK: So you’re focused on everything. All size segments, all geos, all industries, and all use-cases? CEO: Yes. We love focus so much that we have a whole bunch of them. Go big or go home.
At this point, I’m thinking three things:
My Latin teacher, Mr. Maddaloni, taught me that focus was singular.
As it turns out, a quick nod to the chasm gods is a lot easier than embracing them. In the rest of this post, I’ll share some tools I use to detect real vs. faux focus and that you can use to sharpen focus in general.
An ideal customer customer profile (ICP) with concentric circles. Sometimes it’s too binary to have ICP and non-ICP customers, with the result that everything gets equal treatment. Instead, treat your ICP like a bulls eye. Ring zero is credit unions of size X with use-case 1. Ring one is banks of size X with use-case 1. Ring two is insurance companies of size X with use-case 1. Ring three is financial institutions of size X with use-case 2. Ring four is everyone else. I find this increases focus, especially when the inner rings are variations on a core.
Define the idea of strategic vs. opportunistic revenue. Look, I’ve run startups. Cash is king. You want to give me money, I’ll take it. As long as there are no strings attached. Startups get in trouble when they draw-and-quarter themselves by selling roadmap (i.e., non-existing) features to a diverse set of customers. That’s why you should define strategic revenue (e.g., in the first three ICP rings) vs. opportunistic revenue and then religiously enforce this rule: if it’s oportunistic revenue you have to sell what’s on the truck. Don’t even bother asking for roadmap commitments. Maybe give those sellers lower quotas in return. But don’t let them ruin your future by selling your scarcest resource, R&D capacity, for non-strategic purposes.
Segmented metrics. Let’s say you’re strong in SMB and your growth strategy is a big up-market push into MM. All of your reported metrics quickly become a variably weighted blend of two different businesses. You’ll find yourself in board meetings saying things like, “well the average sales price isn’t that meaningful because it’s a blend of SMB deals at $10K and MM deals at $40K.” For that matter, neither are average sales cycle, close rate, win rate, loss-to, and other metrics. So, segment these metrics: present SMB, MM, and total (aka, “blended”) figures. The same goes for industries and use-cases. Sometimes you’re doing great on the new strategy but the core business is collapsing faster than you thought. Sometimes, the core business is going gangbusters and you’ve made no progress on the new strategy. Without segmented metrics, you can’t easily tell.
Not-on-list lists. Planning is an additive process at most startups. “Let’s do this and this and this. Forget anything? OK, let’s add that, too!” To sharpen your focus, add a subtractive element. When you discuss something and decide not to do it, capture that in a not-on-list list. Think: here’s the list of things we decided to do, and here’s a list of things we considered and decided not to do. It will both help your current focus and shorten subsequent debate (think of the asked and answered objection in court).
Split business units. If you’re constantly arguing it’s actually two different businesses that happen to share a go-to-market (GTM) team, then consider splitting the GTM team. Back in the day at MarkLogic, we had two unlikely bedfellows as businesses: intelligence and media (aka spies and publishers). It helped that our staff literally couldn’t attend meetings in the other segment (e.g., security clearances). So we split our business in two: media and federal. We didn’t have SCs, we had media SCs. We didn’t have consultants, we had federal consultants. We didn’t have a CRO, we had a VP of media and a VP of federal. While this is a pretty extreme approach, in certain situations — particularly when the businesses are pretty far apart — it might make sense. We had two different distribution businesses atop a shared product foundation.
I hope this post has given you a few ideas on how to test your own focus, how to sharpen it, and how to report on it.
The growth-adjusted enterprise value to revenue multiple is a personal favorite metric because it’s a quick way to determine if a SaaS stock is in the bargain basement, where I sometimes like to shop. Quick reminders before proceeding:
Any bargain basement shopper needs to heed Wall Street’s warnings about catching falling knives. (Something I’ve painfully done many times in my dabbling as an investor.)
I like this metric because it reminds me of one of the first metrics I ever used to evaluate stocks: the price/earnings to growth ratio, also known as the PEG ratio, popularized by Peter Lynch in his 1990’s book One Up on Wall Street.
The PEG ratio compares a stock’s price/earnings (P/E) ratio to its earnings growth rate. For example, if a stock trades at a P/E of 15x and its earnings growth is 15% a year [1], then its PEG ratio is 15/15 = 1. As it turns out, a PEG of 1.0 tends to be the norm. A PEG > 1.0 suggests a stock is over-valued (relative to its earning growth). And a PEG < 1.0 suggests a stock is under-valued. So, if you’re measuring the value of a stock by its P/E ratio and you’re looking for the bargain basement, you can screen for stocks with a PEG well below 1.0.
Note that — and this is foreshadowing — instead of calling it the PEG ratio, they could have called it the growth-adjusted P/E ratio. It’s the same thing; the latter just has 8 times as many syllables as the former.
Today, software investors don’t really value stocks based on price/earning multiples. Far more commonly, you’ll hear about enterprise-value/revenue (EV/R) multiples instead [2]. So how do we map this growth-adjusted concept to EV/R? It’s easy, do the same math, and just divide EV/R by growth:
Growth-adjusted EV/R ratio = enterprise-value/revenue/growth.
There are three potential complexities with this metric:
1. The name. Coming in at a whopping 13 syllables, the name is a prohibitive mouthful [3]. If we borrow the naming convention from the PEG ratio, we can just call this the ERG ratio — Enterprise value to Revenue to Growth. At a single syllable, ERG gets us an A+ in syllabic parsimony.
2. The details of the definition. EV is almost always a current snapshot, but you can use either forward or trailing revenue and revenue growth rates. This introduces potential confusion, so the most important part is ensuring you know what you’re looking at before making comparisons. Beyond that, we just need to pick a convention, and I’ll be happy to steal Meritech’s —
Growth Adjusted EV (Enterprise Value) / NTM (next-twelve-months) Revenue is calculated by dividing enterprise value over NTM revenue over LTM (last-twelve-months) revenue growth rate.
3. It’s non-intuitive. I embraced this metric somewhat reluctantly because, unlike PEG, I have no intuitive sense of what the value should be. Somehow, the center of PEG at 1.0 is both intuitive and convenient. For the ERG ratio, today’s median is about 0.3 which does little for me intuitively.
Let’s look at some numbers to try and build some intuitive sense around this metric.
Above is a set of companies I arbitrarily picked from Meritech’s public comps. The median [4] EV/R for all of the Meritech data is 6.2, median revenue growth is 21%, and median ERG is 0.31.
One rule of thumb I get from the above is that your EV/R mutiple should be around 1/3rd of your growth rate. To the extent the ERG ratio is much lower than one-third, it suggests a stock is cheap for its growth. To the extent the ERG is much higher than one-third, it suggests a stock is expensive for its growth. For example:
While Klayvio is trading at a lofty 9.4x revenues, it is growing at a rapid 54%. Hence an ERG of 0.17, making the stock appear cheap relative to its growth [5].
While C3 is trading at a even loftier 10.7x revenues, it is growing at only 6%. Hence, an ERG of 1.78, making its stock appear expensive relative to its growth.
In fact, Domo is also growing at 6%, yet trading at a low 1.3x revenues, about 1/8th as much as C3 [6].
Zoominfo and Walkme are both at the median, though they get there in very different ways. Zoominfo trades at 6.3x revenues and is growing at 20%. WalkMe is trading at only 2.0x revenues but also growing only 6%.
Perhaps the most interesting exercise is to sort this list by ERG and then sort it by EV/R — the more traditional way of determining whether a stock is cheap or expensive.
I’ve shown some of the bigger movers using blue arrows when they move down the list and orange when they move up. For example, Wall Street likes Snowflake relative to its growth with an ERG of 0.48, but Wall Street loves Snowflake when looking only a EV/R multiple. Put differently, Snowflake looks a lot overvalued when comparing its EV/R multiple to the median; it looks a lot less overvalued when also considering its zippy growth rate.
I hope this post has demystified this useful metric somewhat, and planted the seed that if this metric is ever going to be popular, it can’t have a 13-syllable name. If you’d like to hear my metrics brother, Ray Rike, and I discussing this ratio, you can listen to the SaaS Talk podcast episode on ERG.
# # #
Notes
[1] Technically, it’s the growth rate times 100 to convert the percent to a number.
[2] Back in the day, we’d talk about the price/sales ratio, the price of share divided by the sales per share. If you take that ratio and multiply the numerator and denominator by the number of shares, you get market-capitalization/revenue, which is one adjustment away from enterprise-value/revenue. In effect, making the EV/R multiple the rough equivalent, and simply a modernized version, of a P/S ratio.
[3] Los Angeles comes in at four syllables and people still abbreviate it to LA to save two.
[4] Nit, but the median EV/R divided by the median growth rate will not necessarily (or even usually) produce the median ERG. Hence the small difference between 0.295 (by dividing the medians) and the actual median value of 0.31.
[5] Which then begs the obvious question, why? For Klayvio, I don’t know the answer.
[6] This again begs the question, why? For C3, I’m guessing it’s probably because of their strong positioning around AI which is red hot right now.
Well, it’s that time of year again, time for my annual predictions post, now in its tenth incarnation. As per my custom, let’s review my 2023 predictions before presenting those for 2024. Please remember that I do these for fun and fun alone. See my FAQ for my terms, disclosures, disclaimers, et cetera.
2023 Predictions Review
1. The great pendulum of Silicon Valley swings back. Hit. I think Silicon Valley is driven by a master pendulum that in turn drives numerous sub-pendulums — and they all swung back in 2023. Valuations came down, structure regrettably came back, cashflow trumped growth, founder friendliness decreased, diligence generally flopped back from FOMO to FOFU, and companies again started to treat employees as, well, people they are paying to do work.
2. The barbarians at the gate are back. Partial hit. They’re there, but not quite buying with the frenzy I’d anticipated. The problem with buyer’s markets is that sellers can often wait — and it seems many have. PE software acquisitions were at roughly pre-pandemic levels in the first three quarters of 2023, though still well below 2021 and 2022 highs. Notable deals on the year include Silver Lake buying Qualtrics ($12.5B) and SoftwareAG ($2.4B), Thoma Bravo buying Coupa ($8B), Clear Lake and Insight buying Alteryx ($4.4B), Vista buying Duck Creek ($2.6B), Francisco Partners buying Sumo Logic ($1.7B), and Symphony buying Momentive ($1.5B). Expect more of this activity in 2024.
3. Retain is the new add. Hit. Customer retention came into sharp focus in 2023 and with it a new, balanced view relying on both NRR and GRR as key retention metrics. As I said last year, “while this bodes well for the customer success (CS) discipline, it does not automatically bode well for the customer success department.” Some found themselves blown up (aka Slootmanned), often in hasty lose/lose transactions leaving customers dissatisfied with reduced attention levels and sales unhappy with additional work without additional resource or pay. Blowing up customer success to save money is myopic. Re-organizing it, or simply re-chartering it, with a more business-aligned mission is the key to success. New technology (e.g., Hook) will help. Jason Lemkin predicts a slow reboot of the customer success function in 2024.
4. The Crux becomes the strategy book of the year. Partial hit. Two things went wrong here. First, I was manifesting this prediction – I wanted it to be the strategy book of the year. Second, I was late to the party. I bought my copy in December, 2022 so to me it was a brand-new book, but it had been released seven months earlier and had already won recognition from the FT, Forbes, and The Globe & Mail. Sales-wise, I don’t have access to great stats, but I can see its best ranking on Amazon is in Business Systems and Planning where it currently ranks 121st. It should be in the top ten with Good to Great, Blue Ocean Strategy, Thinking in Bets, The Art of War, and its older sibling Good Strategy, Bad Strategy. Popularity be damned, I think The Crux is a great book, better than its predecessor which does a great job tearing apart the garbage that passes for strategy, but a worse job of saying what to do about it.
5. The professionals take over for Musk. Hit. I almost downgraded this to a partial hit because “take over” may not properly describe what has happened with Linda Yaccarino. But she is nevertheless the CEO, if perhaps in name only. (And yes, I’m still reluctant to call Twitter X.) The question today is not how long Musk lasts, but how long Yaccarino lasts. Having withstood so much already, I think it’s unlikely that she’s gone in 2024, but I won’t waste a prediction on it this year.
6. The bloom comes off the consumption pricing rose. Hit. I’ve always felt the famous Warren Buffet quote applies to consumption-based pricing: “when the tide goes out you can see who’s swimming naked.” I’m scoring this a hit not because I think usage-based pricing (UBP) – as it’s also known — is bad, but because I felt it was overhyped and often pushed too hard on companies by investors chasing stratospheric (or Snowflake-spheric) net revenue retention rates (NRR). In reality, UBP has both pros and cons and is better applied to some products than others. While UBP companies were hit harder, as this slightly confusing slide from Iconiq demonstrates [1], they nevertheless grew faster than their subscription counterparts in 2023. Consumption models are here to stay, but hopefully the industry can take a more balanced, rational view on them.
7. The rise of unified ops. Partial hit. I think organizations increasingly realize that stovepiped ops functions generate inconsistency, conflict, and excess cost. Though here again, I was manifesting because I believe in unifying all go-to-market ops – e.g., salesops, servicesops, successops, and marketingops — into a single ops function. Some companies call that unified function revops, others use revops to mean only the unification of sales and successops. The big rock is to bring marketing into the unified team. While it’s impossible to know the revops job description from the name alone, a phrase search for “sales operations” versus “revenue operations” on LinkedIn jobs reveals 3x more listings for salesops than revops. We still have a long way to go, but I’m confident slowly and steadily these functions will integrate over time. Every time a unified revops team is created an angel gets its wings.
8. Data notebooks as the data app platform. Hit. This prediction is in large part a proxy for “Hex will prosper,” because I’m a big believer in their vision to create a collaborative analytics platform [2]. In a difficult fundraising environment they raised $28M from not just anyone but Sequoia, using my all-time favorite fundraising strategy — not looking for money. As of the round, they’d grown the business 4x over the prior year. Per LinkedIn, headcount is up 240% over the past two years. They continue to rapidly innovate on product. They support a wide variety of use-cases that go well beyond data apps. They’ve also expanded the personas they support. And, for the marketers out there, they’re the first data-oriented company since Splunk to have a distinctive voice in their marketing (e.g., the Hex 3.0 launch subtitle, “one arbitrary version number for Hex, one giant leap for data people.”) If you want to understand why I’m so excited about this company (and see concrete examples of what some of these data buzzwords mean), watch their latest product launch video.
9. Meetings somehow survive. Hit. I’m so glad the idiocy of companies are for builders, not managers was brief. Yes, companies need to focus on continuous productivity improvement. Yes, companies need to remain vigilant against unproductive meetings, particularly standing ones. And yes, we can always do better. But to suggest discarding the collaboration baby with the unproductive bathwater was always absurd. If you want better meetings, read Death By Meeting. But meetings were, and are, here to stay.
10. Silicon Valley thrives again in 2024. TBD. In a desire to end last year’s list on a positive note, I realize that I inadvertently included a 2024 prediction in my 2023 list. Thus, the score on this prediction remains to be decided. Despite a rough 2023, or more aptly, in part because of it, I remain optimistic that the Silicon Valley business environment will improve in 2024.
Kellblog Predictions for 2024
1. Election Dejection. No matter your political leanings, the 2024 presidential election will be divisive, distracting, and quite probably depressing. It will test our institutions, challenge supreme court legitimacy, and drown voters in higher-calling rhetoric about saving the country or saving democracy, as the case may be. There will likely be a constitutional crisis or two along the way, for good measure.
To stay in my wheelhouse of Silicon Valley, communications, and to a lesser extent, media, I think three things will happen:
The media will make a dog’s breakfast of coverage. Alternative facts. Improper framing. Narrative fallacy. Bothsideism and false equivalence. And many others. Worst of all, due to a lazy preoccupation with oddsmaking, the media will abdicate a key duty in its coverage, wasting the coming months endlessly handicapping the outcome. Instead of this horse-race journalism, the media should do what NYU professor Jay Rosen advises: focus on not the odds, but the stakes in its coverage.
The election will test the once-veiled political neutrality of Silicon Valley. For years, Silicon Valley was a place of quiet liberalism among workers and veiled libertarianism among overlords. The attitude towards Washington was leave us alone and let us work [3]. In the past two decades, that’s changed with lobbying dollars up about 10x, more VCs and celebrityCEOs openly expressing political views, and the rise of podcasts with strong political leanings. A16Z’s American Dynamism initiative has strong political overtones and is surprisingly nationalistic for an international firm [4]. Politics are coming out of the closet in Silicon Valley, for better or worse.
There will be a lot of infantile rhetoric. The rise of social media dropped the level of our discourse, with many politicians only too happy to follow suit. Today’s vile norms (e.g., name-calling) were unacceptable only 10 or 20 years ago. This debasement will continue during the 2024 election cycle. I refer readers to Graham’s hierarchy of disagreement as a framework for characterizing the quality of debate and to encourage everyone to climb, not descend, this ladder.
The ray of good news is that while the election will almost certainly be a mess, most Americans are exhausted by today’s politics and polarization. Eventually, this should percolate into votes and candidates, and ultimately result in a government focused on consensus and compromise [5]. One hopes so, at least.
2. A Slow Bounceback in Startup Land. There’s blood in the Silicon Valley water. 3,200 startups failed in 2023. Unicorns are turning into zombies. The predicted mass extinction event appears to be upon us. Those who can raise money face dilutive downrounds. Even among healthier unicorns, there’s a large backlog of over-valued private companies trying to grow into a contemporary valuation before running out of cash.
On the financing side, VC funding was down to pre-pandemic levels. OpenView surprised the industry with an abrupt shutdown in new investing. Some predict that 25% of VC partners will exit the business in the next few years. Silicon Valley Bank failed.
So, what will happen in startup land in 2024?
We will start to turn the corner. ARR growth stalled. Valuation multiples were hammered. But green shoots are emerging. I think the worst of it is over, particularly for those companies that responded quickly to the downturn by increasing focus, reducing burn, and increasing runway.
This will happen more quickly on the startup side. Net new ARR growth rates are already rebounding. David Sacks is calling an end to the software recession of 2022 and 2023. Gartner predicts software spend will grow 14% in 2024. Things will recover, but they won’t snap back.
And it will happen more slowly on the venture side. Everything happens more slowly on the venture side [6]. While public markets can turn on a dime, venture funds are decade-long, illiquid, limited partnerships where prices are reset more quarterly than daily [7]. This creates a damping effect whereby dramatic change needs time to percolate through the system [8].
3. The Year of Efficient Growth. If 2023 ended up the year of hunkering down, then 2024 will be the year of efficient growth. For the first time, an overall productivity measure, ARR/FTE, has crawled its way into the top 5 SaaS metrics [9]. See chart below for how it varies with scale [10].
The rule of 40 (R40) is back with a vengeance. R40-compliant companies currently command a 61% EV/R multiple premium over their non-compliant counterparts [11]. In a two-factor regression, the relative importance of growth to profitability in predicting EV/R multiples is currently around 2.0 [12] – so growth and profit both matter, but growth still matters more. Because of that, and because Bessemer believes that the relative impact should change as a function of scale, they have introduced a new metric, the rule of X, which is a variably growth-weighted rule of 40 [13]. Don’t read the article with the understanding that there will be no math. There’s plenty of it.
The ultimate sales pitch for the rule of X is its superior explanatory power of the EV/R multiple, as depicted in the chart below [14].
While I have several concerns about this proposed metric [15], the point is that Bessemer, a thought leader in SaaS metrics who to my knowledge defined and/or were early evangelists of CAC, CPP, and CCS, is spending time and energy on a growth/profit balance metric. That’s the point. GAAC is dead. Long live balanced growth and profit.
In 2024, expect emphasis on the usual go-to-market (GTM) efficiency metrics like CAC, CPP, and LTV/CAC, continued emphasis on both net and gross retention rates (NRR and GRR), new emphasis on overall productivity (ARR/FTE) and balanced growth measures (R40), and of course strong attention to cash burn efficiency (burn multiple).
4. AI Climbs the Hype Cycle. In 2023, artificial intelligence peaked on Gartner’s hype cycle. It garnered significant attention, particularly in sectors like healthcare, finance, and entertainment, promising personalized solutions and immersive experiences. However, amid this excitement, there was a growing awareness of AI’s challenges, including ethics and regulations. This marked a crucial juncture for AI, transitioning from hype to practical use, demanding responsible implementation.
Perhaps you noticed the change in voice — the prior paragraph was written by ChatGPT. While I think I’m still winning my John Henry battle with generative AI, I know my lead won’t last forever. Writers fighting ChatGPT are like mathematicians fighting calculators.
Last year was an amazing year for AI, one that both inspired and frightened us. While 40% of humans don’t pass the Turing test, ChatGPT can now pass as human about 40% of the time. Marc Andreessen, in his role as public intellectual, declared that AI will save the world (presumably after software has finished eating it). Some see Andreessen’s manifesto as visionary, others as self-serving, but it’s well worth reading as is this Stratechery interview. For extra fun, watch the techies debate on Hacker News.
Should we lean into AI as the e/acc movement believes, or should we pull back to avoid turning humanity into collateral damage from an AI all-consumed with making paperclips?
If you don’t have time for Marc’s philosophy, I recommend Ben Evans’ wonderful, more down-to-earth deck on AI. It hits all the key issues with a nice balance of insights, examples, and just enough Meeker-style trends data [16].
In 2024, I think AI will continue to blow our socks off as we climb to peak hype. Vendors will propose a wide variety of use-cases, some of which will stick while others will not. Some features will become companies and some products will become features [17]. What’s a technology consumer to do? Allocate time to experiment with a broad range of AI features and products. I expect many AI solutions to go from magical advantage to table stakes almost overnight.
In 2024, AI will continue to pose interesting questions in four areas:
Philosophical. The semantics of predicting vs. reasoning. See this amazing interview with Jensen Huang and Ilya Sutskever, in particular the part where Ilya presents his detective novel analogy. Goosebumps.
Practical. Are you getting quality answers that you can trust or generating botshit? Do generated answers include hallucinations, as a hapless lawyer discovered, or math challenges, as highlighted by Stephen Wolfram?
Legal. Copyright and fair use questions reminiscent of Internet 1.0. Will OpenAI have their Napster moment? Read the New York Times complaint. While not yet at the forefront of debate, my friend Anshu Sharma often highlights important privacy concerns as well.
Pricing. Much as SaaS moved the industry from perpetual to subscription (and then consumption) pricing, will AI move the industry to value- or results-based pricing? [18]
5. AI-Driven GTM Efficiency. We are experiencing a Cambrian explosion of enterprise AI tools. Here’s a part of Sequoia’s map to them. And these are just the leaders.
These things are everywhere. And we’ve not even discussed customer success, customer support (e.g., chatbots), or professional services.
My prediction is that this Cambrian explosion will continue into 2024 and by the end of the year things will start to sort out. What does that mean?
If you’re a vendor, you’re playing musical chairs and you should go all-out to ensure you have a seat when the music stops (i.e., the market starts to organize)
If you’re a customer, you should allocate real time to play with and explore these tools. Don’t be too busy fighting battles with swords to talk to the machine gun salesperson.
If you’re a GTM executive, you should understand that your investors expect real productivity gains from these tools.
In terms of gains, this slide from Battery argues that an AI-enabled sales team with 75 people can support the same number of sellers (and drive the same quota) as a traditional 110-person team. Are you ready for this board conversation? You should be.
6. Beyond Search. The traditional search business is in trouble. For decades, information retrieval people have pleaded for “answers, not links.” While Google has made progress over the years at providing answers (e.g., featured snippets, PAA) [20], generative AI clearly delivers the answers that many have sought for so long.
Search today is in roughly the same mess that it was in the pre-PageRank days of Yahoo and AltaVista. Bombed out. Gamed out. Loaded with clickbait. Over advertised. It’s just increasingly hard to find what you’re looking for. And that’s before the coming, widespread creation of more AI-generated, SEO-driven content. More cruft to jam up the system.
Well, Clayton Christensen to the rescue. We are watching the cycle of disruptive innovation play out. As Google continues to cater to its existing customers and is increasingly run by extractors as opposed to innovators, they create the opportunity for disruption. Now, since Google is a very smart company, they’re not flat-footed in response and are very much trying to disrupt themselves. But, regardless of which vendors win, I expect generative AI’s answers to largely replace traditional search’s lists-of-links going forward.
This will have a huge impact on SEO. For example, the question will no longer be “are you above the fold?” but instead, “are you in the answer or not?” Consider this example, where I asked ChatGPT to make a short-list of conversation intelligence tools to evaluate.
You’re either on that list or you’re not [21]. There is no next page — no consolation prize if you will. Perhaps that’s not really a change because few people clicked on subsequent pages anyway. But I think the stakes are going up in an increasingly winner-take-all race — where most of us currently lack the requisite knowledge and skills to even compete. I’m not talking about how to use ChatGPT for traditional SEO and generate more cruft. I’m talking about optimizing your content for inclusion in ChatGPT results. SEO is dead. Long live ChatGPTO.
For decades, information retrieval expert Stephen Arnold has written a blog called Beyond Search. In 2024, we’re finally going to get there.
7. From RAGs to Riches. Consider this now famous chat with Chevy of Watsonville.
That feeling when your chatbot is overqualified for the job.
General-purpose, large language models (LLMs) can suffer from three weaknesses:
Broad scope, in many applications far broader than is necessary or desirable.
Inability to inform them with specialized knowledgebases and/or supplemental information after the model has been trained.
No sourcing, making hallucination detection more difficult and limiting their use in environments that require sources.
A relatively new technology, introduced in 2020, called retrieval-augmented generation (RAG) solves these problems. This article provides a great technical overview of RAG. IBM Research also wrote a great high-level overview, including two nice analogies:
“It’s the difference between an open-book and a closed-book exam,” Lastras said. “In a RAG system, you are asking the model to respond to a question by browsing through the content in a book, as opposed to trying to remember facts from memory.”
And
“Think of the model as an overeager junior employee that blurts out an answer before checking the facts,” said Lastras. “Experience teaches us to stop and say when we don’t know something. But LLMs need to be explicitly trained to recognize questions they can’t answer.”
From what I understand of RAG, I like it because it’s a practical approach for eliminating problems with LLMs that adds enterprise features like use of existing knowledgebases and references to sources.
In 2024, I think we’ll be hearing a lot more about RAG. Salesforce has added it to Einstein. Glean has raised over $150M from Sequoia and others to reinvent enterprise search using RAG. Cohere has raised over $400M from Index and others to build conversational apps with RAG. Many more will follow.
8. Outbound Finds Its Proper Place. Debates about outbound heat up faster than honey in a microwave oven. Particularly when companies (often quite prematurely) think they have picked all the low-hanging inbound fruit, outbound becomes a religious issue, fast. Here are some of the reasons I’ve heard for this:
The great hope. It must succeed because other methods are topping out or failing (and execution quality couldn’t possibly be the reason).
It worked before. Five years ago at my last company, even if it was in a different situation, with a different strategy, in a different time.
I was brought here to make it work. It’s why the CEO hired me. I know how to build it.
Sales wants control of its own destiny. Even if it’s inefficient, I don’t want to be so dependent on marketing.
I need outbound SDRs to groom into sellers. They’re my funnel for filling AE headcount.
I want a club to beat sellers. When sellers complain about lack of leads, I need to be able to say: “So what have you done to help yourself?”
The last point is true only in cases where sellers are required to generate a certain amount of their own pipeline, which, with the exception of account-based marketing (ABM) models, I don’t think they should do. Remember the quote: “sellers are like airplanes, they only make money when they’re in the air.”
Recently, I’ve heard more and more CEOs abandon this religious belief in outbound. That’s good. Standalone outbound [22] is a low-conversion rate activity. Stalk someone. Twist their arm to agree to a meeting. See if they show up. (Often, they don’t, so repeat the stalking process.) Try to convince them they need to buy in your category and then to buy from you. See what happens.
If you were a seller, which would you prefer?
The stalked, arm-twisted lead above, or
Someone who found us through an organic search, downloaded a white paper, attended a weekly demo session, rated it highly, and asked to speak to a seller
Conversion rates usually reflect this [23]. Partner- and inbound-generated leads often convert at double or triple the rate of outbound. I expect standalone outbound effectiveness to only get worse because of the AI-driven tools arms race. Every SDR will be sending AI-generated, personalized email sequences. And that’s not to mention the new Gmail anti-spam rules that go into effect in February.
What’s the glaring exception here? ABM, done properly. When a company targets a small number of accounts, focuses sellers on penetrating them, and aligns both marketing and outbound SDRs as part of the effort. In effect, the whole company stalks the customer, not just an SDR. Does this work? Yes, absolutely. What’s the catch? That’s simple:
Is the juice worth the squeeze?
ABM is a lot of work. You shouldn’t bother trying it to win a $10K or even a $50K deal. But when you can do $100K to $500K+ deals and have a few strong references in a vertical to which your company has strategically committed, that is when you should do ABM.
Outbound isn’t Santa Claus. It’s just a nice old man with whiskers. In 2024, I think many companies will figure that out.
9. The Reprise of Repricing. Compressed valuation multiples and reduced growth mean lower stock prices. That’s no surprise. However, this creates real problems with equity-based compensation, greatly lowering or entirely eliminating its value. Let’s look at two common equity-based compensation methods.
RSUs which are typically granted in terms of value. For example, if you’re granted $400K worth of RSUs over 4 years when the stock is $50, you get 8,000 shares over 4 years or 500 shares/quarter. If the stock falls to $20, you’re now vesting $10K per quarter instead of $25K. That’s a big compensation hit.
Stock options which are the right to buy shares at a fixed price, typically the stock’s value on the day the option is granted. For example, you are granted 8,000 shares over 4 years when the stock price is $50. If the stock falls to $20, your option is “underwater,” meaning it’s basically worthless because the market price is well below your strike price [24]. That’s an even bigger compensation hit [25].
Now, let’s imagine that we at GoodCo have a similar competitor across the street called NiceCo, and that NiceCo’s stock has suffered similarly. I can stay at GoodCo and vest equity compensation at a reduced or zero rate, or I can quit, cross the street to work at NiceCo, and get a new grant.
For RSUs, I might get a new grant of 20,000 shares and vest at my original $25K/quarter rate. And feel like there’s upside because the stock may appreciate from there.
For options, I might get a new 20,000 share grant at a strike price of $20/share, a no-brainer compared to my existing grant of far fewer shares at a far higher price [26] [27].
How can GoodCo retain its employees in this situation? The short answer — barring soft factors like superior management, culture, and perks — is they can’t. This is a major problem and left unsolved, GoodCo will lose a lot of employees to NiceCo [28].
Enter repricing. While I won’t get into the details, the basic idea for stock options is that in return for some modest consideration (e.g., a reduction in share count), the company will reset the strike price on the options in the example above from $50 to $20. While the concept is simple, the rules are different for public and private companies and, unsurprisingly, public companies are more restricted in what they can do.
For RSUs, it’s slightly different. Technically speaking you don’t need to reprice anything. The company can simply grant more RSUs to make up the difference in reduced value. Or, it seems they can run a sort of repricing where they, e.g., redo the initial grant math to produce a new higher vest rate, but in exchange for a vesting reset.
After that long introduction, my prediction is simple. In 2024, repricing will be back. If your company has a greatly reduced valuation and is not talking about repricing or its equivalents, then you might want to ask them. I’d advise some patience because these things can take time. And bear in mind these rules often vary a lot by country.
As always with financial and career matters, make your own decisions, consult your own advisors, and ensure you understand Kellblog terms and disclaimers. You can also read a book like Consider Your Options for more information.
10. Peak Podcasting. For years, podcasts have been on the rise, with the pandemic driving a massive peak in podcast creation. One of the better-kept B2B marketing secrets was that starting a CEO podcast could serve as a structured way to help CEOs, particularly introverted ones, get out there and meet new, important people. Creating a CEO podcast was the ultimate three-fer, improving:
Communications, driving company messages and positioning the founder/CEO as a thought leader.
Customer relationships, gaining access to and/or reinforcing relationships with next-level executive contacts as invited guests.
Partner relationships, interviewing fellow CEOs, greasing the skids for many kinds of partnerships, potentially including the one that eventually sells the company.
If you like the sound of that and haven’t started one yet, I still think it’s a good idea. But start fast. It takes a long time to build an audience and I think in 2024 we will hit peak CEO podcast, for the simple reason that the word is getting out. My feeling is largely intuition-driven – podcast advertising forecasts still paint quite a rosy picture – but I think the software market will tire of B2B CEO podcasts over the next few years. If you don’t believe me, ask the podcast police.
If you want to create a podcast, make sure everyone understands why you’re doing it, get buy-in for a long-term, high-frequency commitment [29], and start now. That should keep the podcast police away.
Thank you for reading to the end, and I wish everyone a happy and healthy 2024.
# # #
Notes
[1] The two bars on the right make the point – they compare top-quartile growth rates of UBP and subscription companies. This slide doesn’t do the world’s best job of making this point, but I’ve seen it in other studies as well.
[2] Note that while I’m an angel investor in Hex, I do not work closely or actively with the company so my conclusions about their progress are based entirely on external observation.
[3] With the major exceptions of government-funded research projects (e.g., DARPA) and, usually as companies gain in scale, the embrace of government as a customer.
[4] Though I suspect they’d argue it’s a US-focused practice more than a firmwide initiative, but that hasn’t been 100% clear to me in reading about it.
[5] Which I believe was the subtitle of my American Government textbook back in college.
[6] Even OpenView’s “abrupt” shutdown was not an overnight closing of the doors; it was a cessation in new investment. As the firm noted, it will continue to exist and support existing investments until the existing funds reach their eventual conclusions – which can be years, even for growth investors.
[7] While new investments and valuations can turn on a dime, the rest of the business is focused on the long-term task of building companies and delivering TVPI, DPI, and IRR over a decade or so.
[8] IMHO, this damping is a good thing because it damps out irrationality as well. You can’t see a bank run on a venture fund, because investors generally don’t have the right to demand their money back.
[12] Bessemer State of the Cloud, slides 14-16. This shows nicely the growth at all costs era (6.0x), the trough after the peak (0.8x) and return to normal (2.0x). While growth is still twice as important as profit in predicting valuation, the balance still matters.
[13] SEG’s growth-weighted rule of 40 is double the 2x growth-weighted-average of growth and profit (i.e., it’s like a weighted average that isn’t averaged because they don’t take the last step and divide by 2). SEG does this to stay consistent with R40 which is the sum of profit and growth, not the average. In the rule of X, the relative weight (i.e., the “multiplier”) varies – over time and across stage. This makes the metric more complex, less comparable across stage and time, and produces a wider ranges of outcomes. For example, a company whose (growth, profit) is (100%, 50%) scores 150 on R40 and scores 950 on RX when the multipler is 9x, 130 when the multiplier is 0.8x, and 280 when the multiplier is 2.3x.
[14] Frankly, this argument strikes me as circular. If you’re getting the weight multipler from a regression of the current market, it seems obvious that you’d expect a higher R^2 compared to any fixed weighting of growth and profit, including the default weight of one in the rule of 40.
[15] My concerns: bad name (if rule of 40 abbreviates to R40, this abbreviates to RX), hard to interpret scores, incomparability across stages and time, and seeming circular logic (see prior note). Their ultimate point is correct: growth matters more and blind adherence to an unweighted rule of 40 may take you to the wrong place. But this metric needs some more work.
[16] Meeker was legendary for drowning the audience in nevertheless interesting data in her annual tech trends reports. As my dearly departed father might have said, “there’s enough here to gag a maggot.”
[17] Ben Evans covers these ideas, starting on slide 39 of his deck.
[18] The argument in favor is that AI will create a lot of value, vendors want to capture that value, and vendors are certain enough that they’re willing to take the downside risk to get the upside. The argument against is that value creates an upper bound on pricing, but the lower bound is determined by the price of alternatives. At Host Analytics, I could replace a Hyperion system that cost $500K/year with a SaaS app that cost $50K. That’s a lot of value to tap. But if Adaptive Insights were willing to do the same deal at $25K, then the price of alternatives, not value, became the focus of the conversation and differentiation the focus of the sales cycle.
[19] Please note that none of these references are endorsements, I don’t know many of the companies, and I’m sure many would be unhappy with my chosen one-word label. The point is to show the breadth and depth of the market.
[20] Front-running content producers in the process – e.g., featured snippets provide answers that leverage content producers’ content while eliminating and/or reducing traffic to their sites.
[21] This example also shows the problems with ChatGPT’s cut-off date, e.g., it doesn’t seem to know that Chorus is now part of Zoominfo.
[22] By standalone outbound, I mean outbound not done as part of a bigger ABM program.
[23] Unless they’ve been gamed to over-credit outbound as is sometimes the case when a company has “outbound fever.”
[24] Technically, even an underwater option has value because of its time value and the chance the stock price may rise above the strike price at some point in the future during the life of the option. In my example, it needs to go up by 150% before the option has any intrinsic value.
[25] Though these days an increasing number of tech workers are jaded with stock options, may value them at zero, and see them as pure upside – e.g., lottery tickets on top of their cash compensation. In that case, there is no “hit” per se to compensation, because they were expecting zero value anyway.
[26] If the company derives option grants from value, they’d say: we’ll grant you $400K worth of value, so at $20/share, that’s 20,000 shares. Even if they don’t work this way and simply offer to match the number of shares, the job-switcher is still offered a far better deal — 8,000 shares at a $20 strike price, as opposed to $50.
[27] Note that other factors come into play here, including the fact that grant sizes tend to decrease over time. For example, if you’ve been at GoodCo for four years with an initial grant of 8,000 shares, the going rate for your job might have dropped to 2,000 shares. Thus, crossing the street to NiceCo might result in a grant at a lower strike price, but with a much smaller number of shares. I think this is somewhat less true of RSUs (because they feel more a part of annual compensation as opposed to gravy on top), but I’d need to think more to be sure.
[28] That said, in this example, they can presumably hire NiceCo employees in the same situation. That aside, neither company benefits from the mass rotation of employees.
[29] Because that’s what it takes to climb the charts. And some advertising spend doesn’t hurt either.
I’m Dave Kellogg, advisor, director, blogger, and podcaster. I am an EIR at Balderton Capital and principal of my own eponymous consulting business.
I bring an uncommon perspective to enterprise software, having more than ten years’ experience in each of the CEO, CMO, and independent director roles in companies from zero to over $1B in revenues.
From 2012 to 2018, I was CEO of Host Analytics, where we quintupled ARR while halving customer acquisition costs, ultimately selling the company in a private equity transaction.
Previously, I was SVP/GM of the $500M Service Cloud business at Salesforce; CEO of MarkLogic, which we grew from zero to $80M over six years; and CMO at Business Objects for nearly a decade as we grew from $30M to over $1B in revenues.
I love disruptive startups and and have had the pleasure of working in varied capacities with companies including Bluecore, FloQast, Gainsight, Hex, Logikcull, MongoDB, Pigment, Recorded Future, Tableau, and Unaric.
I currently serve on the boards of Cyber Guru, Light, Scoro, TechWolf, and Vic.ai. I have previously served on boards including Alation, Aster Data, Granular, Nuxeo, Profisee, and SMA Technologies.
