Weekly Update: Amazon Retail Unit Economics, Databricks vs SNOW, Texas Instruments, Sonova, WANdisco

Content Published Last Week

1. Amazon: Deconstructing the Retail P&L

2. IP ANALYSIS: Amazon Retail Unit Economics

3. Databricks: Melting the SNOW

4. Analog Devices vs Texas Instruments Positioning

5. Sonova, Hearing Aid Fittings, & OTC Channel Growth

6. WANdisco & IoT Growth Opportunities

Amazon Retail Unit Economics Breakdown

Last week, we interviewed a Former Finance Manager at Amazon who was partly responsible for collating separate country and retail product category financials into one consolidated AMZN P&L for Brian Olsavsky, AMZN’s CFO.

We also published our analysis, where we share our learnings from the interview and breakdown AMZN’s Retail P&L to better understand the underlying e-commerce unit economics.

We’ve shared a few snippets of both below. The full interview and write-up is available for members only.

The first insight from the interview is that the S-Team looks at the retail P&L excluding ads and Prime:

If you include ads in the core consumer business, you are unable to assess the health of the business, at its fundamentals. When you think about core ecommerce, because the fulfilment network and those investments are so massive, there is a relentless focus on optimizing the opex/capex outlays associated with those operations…if you include the advertising numbers in the core consumer business, it entirely obfuscates how that business is operating independently of the advertising. - Former Finance Manager at Amazon

The S-Team is focused on optimising the operating cost per unit shipped. If you include ads and subs revenue, it hides the underlying retail operating efficiency. This could lead to the dreaded 'Day 2' mentality.

In our full analysis, we breakdown the various expenses in the consolidated cost of sales and net out Whole Foods COGS and content licensing costs to find the underlying ecommerce margin. AMZN’s pure ecommerce gross margin has been stable at ~22% but the contribution margin has plunged from ~8% to -3.5% in the last 7 years.

close

Zoom InZoom OutReset

This is due to the significant increase in fulfillment costs: from 2015-2022, fulfillment as a percentage of revenue increased from 14% to 25.8%. AMZN’s relentless focus on same-day delivery has made it difficult to cover the variable costs of shipping a package for the last 2 years.

However, if we add back both subscriptions and advertising revenue, gross and contribution margins look very different.

close

Zoom InZoom OutReset

The full non-AWS business, including ads and Prime subscriptions, yields a ~40%+ and 16%+ fully-loaded gross and contribution margin, respectively.

Decoupling the consolidated opex lines between AWS and Retail is somewhat trickier. During the interview, the executive walks through how to split marketing, tech and content, and G&A between the two businesses.

We use these estimates to build an underlying EBIT margin for the e-commerce business excluding and including subscription and advertising revenue. Members can read the full write-up and model here.

One interesting question to understand AMZN's normalised e-commerce EBIT margin is: what percentage of the tech and content line is pure venture spend?

This was the executive’s answer:

15% to 20%; maybe more. Let me help you view the business in the way that Brian Olsavsky views the business; it is called the contributions and investments page. The first half of this page is all the contributions, in terms of operating income, and it's broken out one column North America, one column international, one column stores, one column total…The way you think about the business is, we're building some profitable businesses and we're going to make a bunch of investments. We're structured as the world's greatest entrepreneurial organization. Every team is no bigger than two pizzas can serve and they're all inventing all the time. - Former Finance Manager at Amazon

15-20% of the tech and content line is potentially VC-like expenditure.

This highlights one major takeaway from the interview: Bezos’ genius is in orchestrating a scalable, innovation machine. Two pizza teams, Day 1 philosophy, 6-page memos, etc. It may seem cheesy, but this has laid the foundation to build probably the most entrepreneurial company that has ever existed.

The company is engineered to spend every dollar it earns. To invest in AMZN, you have to believe management will allocate retail and AWS FCF effectively because it's organised to spend it all.

And this is what worries some investors today.

Has AMZN overbuilt its network? Is NA e-commerce mature? Can International margins replicate NA? And if e-commerce penetration is more mature than we expect, how long may it take to recoup the required return on the recent significant capex?

The decline in the 3-year rolling ROIIC underpins such worry:

close

Zoom InZoom OutReset

Members can read the full interview and our write-up and see the line-by-line assumptions and AMZN Retail P&L.

Databricks: Melting the SNOW

Databricks executives talk about ‘melting the snow’, jargon for taking market share from Snowflake.

This interview with a Former AWS and Current Databricks executive is interesting on many levels for anyone looking at cloud computing.

Databricks is picking off SNOW workloads as customers look more carefully at costs:

We can quickly and easily analyze Snowflake real estate with the customer's buy-in and sponsorship. We can target specific workloads and say okay, you're spending X amount on this workload, and we don't know your discount, but share your discount, so we have to be however much money for this workload. Okay, let’s mirror that workload in Databricks and see. So basically, we can get a blueprint of the environment extremely easily and then chip away and say okay, we did three POCs on three different workloads and consistently proved that we are 50% cheaper and 30% more performant. It’s a no-brainer. - Account Executive at Databricks

ETL spend seems to move over relatively easy before a more tactical approach to win a greater share of workloads that impacts business users. As Databricks’ ease-of-use improves, this could prove a greater problem for SNOW:

A lot of it was the ETL, and you could move at least 50% quite easily, and then the remaining 40% to 50% requires a more tactful approach because it impacts the business user and what their preferred solution is. They don’t know about upstream data transformations; they don't care about that, so you can move that pretty easily. Data engineers prefer a superior technical solution, which Databricks can provide. - Account Executive at Databricks

The executive believes there are early signs that a recession will favour Databricks given customers are seeing up to 50% cost savings when switching from SNOW. Check out the full interview for a deeper-dive into how Databricks aims to 'melt the snow'.

The recession is a great fuel for this, and those conversations are being had with some of these major Fortune 500 companies; the economic downturn is helpful to those discussions. I don't think it's incredibly difficult. It could be a several-year plan, a two-year plan, or a one-year plan, depending on all the technical debt in the existing system, but it's a conversation worth having if you can save 50% of your costs. - Account Executive at Databricks

Analog Devices and Texas Instruments

Fundamentally, the analog semiconductor business is great:

semiconductor wins are very sticky. You could shut the lights off and stop answering customer phone calls tomorrow, and you'd still have a long tail of revenue stream coming because many designs last 10, 15, 20 years, if not longer. That was another advantage to ADI, by the way. TI would end-of-life many products, and we promised to end-of-life only on very rare occasions. We still have DSPs that are 40 years old and even older than I am. - Former VP at Analog Devices

Texas Instruments seem far more aggressive to win new business vs Analog:

We would lose on price. And that was always the fear, excluding DSP. On DSP, TI was a formidable competitor in many ways. In my later years there, DSP gradually died on its own. In my first few years at Analog, half of my design opportunities probably had some DSP component, but as ARM soared and took over, very few conversations were around DSP. For the non-DSP stuff, we were always afraid of TI coming in with high volume and low prices that we couldn't compete with. They also had claims of accuracy that weren't always true. They were very aggressive, and they had a huge sales force. - Former VP at Analog Devices

And TI’s culture lead many to move over to Analog:

TI also had a bit of a revolving door, so it wasn’t the happiest place to work; I never met one person who said they loved working at TI. We had a lot of ex-TI people. The management was aggressive, expectations were unreasonable, and they were high. During QBRs, it would be your ass if you didn't perform or you didn't have an answer for something - Former VP at Analog Devices

This interview explores how TI compares to Analog and on what basis each company competes.

Sonova Group and Hearing Aids

A Former Sonova Director explores the recent Sennheiser acquisition and compares its fate to a previous Consumer venture by Sonova:

So many things go along with that device, meaning the consumer doesn't want the device in the first place. The only person who is keeping it in that person's ear and fitted is that audiologist or dispenser. There were other direct to consumer products such as Whisper and Crystal Ear, and Johnson & Johnson had Songbird, and the studies proved it took us more time and effort to sell and fit those products for a lot less margin than it did to fit one with the stronger margin to keep that practitioner with a roof over their head and food on their plate. With Sona, you had this device in the office which they gave to practitioners on consignment. It was at a low price point, similar to what they're trying to do with Sona and Sennheiser, but they didn't sell. I think Sennheiser will go the same way as Sona. - Former Sonova Director

WANdisco & IoT Product Market Fit

WANdisco is a UK-listed data migration technology platform that has a new avenue of growth with its Edge-to-Cloud tool. Its original data migration tool struggled to gain mass adoption due to customer inertia and ‘good enough’ free tools like Hadoop:

I think the challenge that we had was just getting customers to say yes and to use it. The competition was, do nothing, which is true for a lot of technology. The competition either do nothing or use free tools out there that did require shutdowns but they were willing to accept that for not spending money on a tool that was a more elegant solution. - Former WANdisco Director

WANdisco’s Data Migration technology appears well-suited for migrating data from IoT devices stored on the edge to the cloud. The recent customer wins and expansions of the contracts suggests WANdisco may have finally found real product-market fit. This interview explores the IoT opportunity for WANdisco.

There's features inside of the clouds that can do that kind of replication and synchronization, but I don't know that it's necessarily real time like the WANdisco solution, where you're effectively copying to multiple places. The technology, in its simplest form, copies data and it copies, copies, copies while people are using the data at one location and then at some point all the data's copied, now it goes into a synchronization mode, so you're effectively writing to both places at the same time when something changes. So they're identical. That's the premise of the technology. I don't know that anybody else really does it that way, so it probably is fairly unique. - Former WANdisco Director