A lot of interesting data points from this @theinformation article by @amir.

This gives us a hint as to the contribution margin and CLV of @OpenAI paid subscribers, how much is lost per free user, and the freemium conversion calculus.

5 things we need to account for:
1) CLV is a variable profit figure, so we care about variable costs
2) Not all of OpenAI's costs are variable
3) Variable costs are driven by usage
4) OpenAI has a lot of free users, who consume a lot of the variable costs
5) But free users probably use a lot less than paid users on average

Some quick back of the envelope calculations...

(Spreadsheet below)

bit.ly/3WGZz1h 2024 costs:
-Inference costs: $4B (assume 100% variable)
-Training costs: $3B (80% variable?)
-Salaries: $1.5B (15% variable?)

Training includes paying for data. Assume not all contracts are pure usage-based.

Those server costs are likely purely variable.

$6.6B in variable costs.

🚨New Paper Alert!🚨
The overall churn rate is one of the most popular disclosures for subscription businesses.

When it falls, executives will say this is a sign that business is going well. We show that in general, it tells us almost nothing about how a business is doing.

bit.ly/3HWjj8I Case in point: Daniel Ek, CEO of $SPOT. Overall churn is down 30%, ergo what we are doing is working.

Including the overall churn rate in the CLV calculation is, to our knowledge, what is always done when specific numeric calculations of CLV are included in public filings and equity research reports. Here are a few reports. $PTON did the same thing.

1/11🚨"NEW" PAPER ALERT!🚨
New version of my paper with Shin Oblander proposing a method to estimate the long-term impact of major events on consumer purchase behavior. We now analyze the impact of COVID in 12 categories (see below) through June 2022.

bit.ly/330SUSJ 2/11 Main question we answer: what was the impact of COVID on spending (and how it breaks down into active customers, purchase frequency, and AOV) in those 12 categories, and how did it evolve over time?

Cool to see CLV focus in @Clear's filings:
-~90-95% annual retention in '21-22 -- wow! But they are adding winbacks/renewals. That won't last. Normalized more like 85%?
-They say 16x LTV to CAC at IPO. Is it *really* that good?

bit.ly/3ifR7nP --On one hand, was based on '19/20 retention, with more "normal" 79-86%.
--On other, CAC = *airport-related* S&M / *total* new subscribers. What?

They say 72% of acquisitions were through airport.

No info about what % airport-related S&M is of total S&M. Deflates numerator.

We did a new deep dive into $W and just don't see any value in the equity. Fair value of $0.

Highlights:
1: CAC has gotten a lot worse -- more than doubled vs 2018. $104 assuming 7% of repeat spend = ads for repeat orders. $186 front-loaded.

Link: bit.ly/3A3cH4Q 2: Spiking CAC means that much higher of a hurdle to break even on the customer, when the economics at the much lower CAC were already challenging.
3: I assume Apple privacy changes are at least partly to blame. Doubtful this is just a COVID thing.

1/7
Scary churn figures from $PTON this Q. After vacillating .5%-.9%, it spiked to 1.41%.

Yes, Canada, But c'mon. A lot of people were canceling because of the subscription price increase.

https://twitter.com/laurenthomas/status/1562757064400654337

2/7
If $PTON is in cash preservation turnaround mode, why is it still spending > 12% of revenue on R&D?

Alright, finally read the @jdroege thread. While there are some points I agree with, there is a lot that I don't. A few quick hits... To me, false! If all customers are acquired organically, and your CAC is very low, that's awesome! From a CBCV perspective, there is no presumption that contemplating CAC means myopically focusing on FB/Google ads.

https://twitter.com/jdroege/status/1540391779324657669?s=20&t=17ETMP-wgOtMiUghn3isPQ

1/8
All costs can be attributed as being driven by revenue/orders (variable), customer acquisition (CAC), or uncorrelated (fixed), contemporaneously or with a lag.

[1] Some costs are purely variable -- those should always ding the contribution margin (direct labor/materials). 2/8
Funnily enough, some of these are left out of CLV calculations (e.g., with gross margin CLV calculations).

It seems to me like the pendulum is swinging too hard for $PTON (weird for me to say given I could be considered kind of negative on their retention).

Let's take a deep breath and go through pre-pandemic economics and how they've changed.

Spreadsheet: bit.ly/3qiEgmV CAC in Fiscal 2019 was ~$806. They disclose Net CAC. Add back GP on Bike gets us to gross CAC by their definition (although yes further adjustments should be made).

FY 2019 had high S&M as a % of sales. I just finished talking about it here myself:

https://twitter.com/d_mccar/status/1457756869855006730?s=20

1/7
One of my hobbyhorses is disclosure, and a lot of the CBCV work I do (and through @ThetaCLV) is as companies go public. The current state of company disclosure is a mess.

Well, I just co-wrote a paper with @AswathDamodaran and @maxccohen on this: bit.ly/3Af7KmK 2/7
We talk about what the problems are, why they are that way, and what we can do about - specifically pre-IPO. It was a ton of fun pulling this together, and by pulling together the world's leading experts on this stuff (myself notwithstanding), makes for a really good read.

1/10
I was reading @profgalloway's note on @Aspiration and could barely believe what I was reading. Two things...

It takes guts to celebrate adjusted EBITDAM, or adjusted EBITDA minus marketing expenses. 2/10
Yes, companies with good retention will see marketing go down. But to suggest that it is extraordinary and should be "backed out" is ridiculous. Esp when, at Aspiration, marketing is > 100% of revenue, expected to continue for at least 2 years *under their own projections.*

1/7
Our analysis of $WRBY's unit economics is live!

Main takeaways...

bit.ly/3i46Kfh 2/7
1) Good but not amazing unit economics - they spend ~$55 to acquire customers, make back $70-120 in profit post-acquisition => marketing ROI of 130-220%. Better than $LYFT, not as good as average company we have diligenced (but margin assumptions were somewhat conservative).

1/6
$TOST drops their S-1 (h/t @zwinst)! So interesting reading through the customer disclosures. All sorts of nuances/questions from their CAC, NRR, and payback disclosures.

(No cohort disclosures - no C3. Grr!)

A few thoughts...

bit.ly/3nrCeiG 2/6
CAC:
(1a) As with $PTON, they bake hardware profit into CAC. Unlike $PTON, $TOST loses money. Subsidy = CAC (something @trengriffin talks about a lot). Makes sense.
(1b) CAC = spend *2 months ago* / adds. Never seen this, but makes sense. Acquired => live takes 2 months.

UPDATE: Their CAC is probably significantly higher than $27. Someone spotted that their definition of CAC is acquisition costs divided by active customers and not new customers acquired.

What on earth definition is this? We could still back it out from the retention figures...

https://twitter.com/d_mccar/status/1430265892739354631

A lot of comments about this - man! This is my take:
1) Transactional businesses generate operating leverage in that they spend less for a repeat order than a new one. The extent of it depends (less for $CSPR, more for marketplaces, etc), but it's a well-known reality.

Lots of great customer data in the @WarbyParker S-1. A few interesting tidbits:
--Blended CAC of $25 is pretty low, but then again, 2/3rds of their sales came through physical stores pre-COVID (=> lower CAC but more store-related overhead).

bit.ly/3sLr5du --It is weird that they would bake CAC into their contribution margin. This would only be true if they had to spend that much every time any customer ordered, which hopefully is not the case...

[1/8]Given all the uncertainty swirling around it, I put together a detailed analysis of $DASH's unit economics. As it turns out, a relatively simple model captures most of the customer dynamics (great fits!).

Key insights:

bit.ly/3lqUAfk [2/8]
1/ Exeptional repeat buying -- e.g., customers do ~2.6 orders in yr 1, 3.5 in yr 2, even more thereafter. Very back end loaded.

2/ Variable margin is thin but expands over calendar time and tenure: < 0% in yr 1 to 10% in yr 3 => valuation sensitive to margin assumptions.

@patrick_oshag Golly - there's a lot that one could say, but let me try:
1/ The paradox of great NDR: great for valuation, but leaves you extremely sensitive to changes in long term competitive position and the discount rate, a variant of duration risk. @patrick_oshag 2/ Some of the best opportunities in B2C could be those for which we see a "retention smile" -- NDR that falls then rises again. During the growth phase, aggregate-level retention measures *must* by construction look really bad, yet sales growth will be surprisingly resilient...

[1/7]"The Rise of ‘Retention Rates,’ the Software Metric Investors Are Obsessing Over"

Great, timely article. A few comments:
1-I lament that there is no universal standard for calculating NDR. We need standards for comparability/adoption. Regulators!

bit.ly/316guNq [2/7]2-Weird how they discuss importance of revenue growth versus NDR. The latter drives the former. Would be more conceptually distinct to compare gross customer acquisition [units or acquired MRR] vs NDR.
3-While good, NDR isn't great for many reasons:

$W earnings!!

A few thoughts:
1/ This was not a good Q. Adj EBITDA margin was -5.5%, FCF performance similarly poor, weak margins, rev growth slowest ever?

Little/no CAC leverage despite March pop. No leverage on variable headcount despite cuts. See images. 2/ The stock is up because QTD sales are +90% vs +60-80% consensus from crcd panel data. Very strong but unsurprising. As per my comments to @laurenthomas, 86% of the market was shut down, $AMZN de-prioritizing furniture, and WFH purchasing: cnb.cx/2YB2zi3

1/Working paper thread! “Assessing the Role of Customer Equity in Corporate Valuation”. TLDR: customers deserve a role in valuation, CE was prescient but ~no uptake outside marketing. We show why and propose CBCV as solution. bit.ly/2tXVOth (LI: bit.ly/2QQFVxL) 2/Traditional financial data is getting less relevant while customer data is getting more relevant. Investors have a blind spot and insights through customer data could be the solution. I think this is our field’s biggest opportunity to expand the role of marketing by far.

A few quick thoughts on $CSPR:

1) As mentioned earlier, no cohorted data, but no surprise there (bit.ly/30e9DQu).
2) Only repeat purchase data points: 14% of customers returned within a year of original purchase, 20% of customers in DTC were repeaters. 3) Not surprised they quoted repeat figures in terms of customers, not sales. Repeat orders at dramatically lower AOV's (mattresses => pillows). As per @second_measure, their annual *dollar* retention is 6% (h/t @ataussig). They do not have repeat business right now.