Recently Published Content

1. AWS: Competitive Positioning & Shift from Horizontal to Vertical Selling

2. Carvana: Finance GPU Challenges and Potential Solutions

3. Investor Dialogue: Judges Scientific

4. Elastic Positioning & Growth Strategies

5. Victoria, Carpetright, & UK Flooring Market

6. Tremor International: SSP & DSP Vertical Integration

New Weekly Update Format

We’re introducing two new content formats:

Screenshot 2022-07-23 at 18.31.44.png
Screenshot 2022-07-23 at 18.31.44.png

Weekly Update

At the start of each week, we will publish a Weekly Update listing the content we published in the last week and highlights of the most interesting quotes. We will provide a short piece of commentary for every quote explaining why we think it’s important.

This Weekly Update is the first of this format.

IP Analysis

Over the last 18 months, we’ve been writing ~1,500 word pieces of analysis on companies we’ve covered. This analysis was sent in the Weekly Update along with a list of all content published that week.

We plan to significantly increase the non-interview content published at In Practise. This will include IP Company Profiles, ad-hoc company updates, and deeper pieces of analysis on quality companies.

We plan to leverage our existing interview library but also hire a small team of investment analysts internally to publish quality research.

It’s our aim to provide quality, boutique research on top of our interview library. We will only publish analysis when we can add value and so we expect a somewhat irregular cadence for IP Analysis pieces.

Going forward, most IP Analysis write-ups will be behind the paywall. Periodically, we will include full write-ups in the Weekly Update similar to Costco’s Changing Culture.

AWS: Horizontal to Vertical Selling, SNOW, and Cloud Competition

Former Head of Strategy for AWS, West-US on the 3 waves of the cloud industry:

The way I see the market broadly, just for some context here, is really in three waves. The first wave is done. This was brand new, in circa 2006 to 2008. Compute and storage, the original category, and that was when everything sold itself. It was self-service that pioneered the market. That was all about telling people what the cloud was and then convincing them to migrate. That ship sailed. That is done. At the end of that, you were starting to go from pure IaaS to PaaS. - Former Head of Strategy at AWS, North America
Wave two was all about blowing out and expanding the PaaS category. That’s when we went from two services to 100+ and everything beyond. That’s when we also started to realize this is very complex, and we needed to do some of the things I was talking about; simplifying with bundles and starting to engage other parts of the customer organizational chart. We are right at the end of wave two right now, and other parts of the organizational chart need to get addressed. Microsoft is doing it. Google is doing it. Oracle is doing it. IBM is doing it. AWS is not doing it, at least not scalably. As we go into wave three, it is going to be all about that and a few other things. What’s also going on in this wave is it’s a maturing market in terms of pricing power in certain categories. The need for innovation is greater. As we move into this wave three, this is the kind of stuff that Adam has on his plate. - Former Head of Strategy at AWS, North America

Although we’re still early in the cloud migration, the executive argues that Wave 3 of the cloud industry is the evolution in the go-to-market strategies of hypers-calers: a shift from a horizontal to vertical selling process. This means the cloud players need to be focusing on selling solutions to specific end verticals rather than simply letting developers adopt micro-services horizontally like the early days of cloud.

The executive believes the shift from horizontal to vertical selling is a big challenge for AWS. Take how they work with partners vs MSFT, for example. Improving the Partner program has been a focus for AWS for the last few years but still has room to improve.

AWS has always worked with partners. AWS intentionally leaves the SaaS layer in the stack to partners. They don’t have a SaaS layer like every other hyper scaler does and Oracle does. That is the purview of the partners. That is baked into their model. There is a ton of partner activity going on and the partner attach rate is that, theoretically, 100% of deals would have a partner attached to them. That is what is they say internally. That was already happening, but the engagement with the partners was purely technical, for the most part, and maybe focused on cloud migration versus a specific type of workload or one of those 31 categories or certainly not a cross-multiple category simultaneously. Contrast that with Microsoft which engages with the very same partners, also at that technical level, but also at industry specific level with different teams, engagement with different groups within the same partners. They did it; we didn’t, at least didn’t do it in a systematic way. That distinction is pretty critical. If we fast forward to now, and fast forward for the next five to 10 years, that’s something that is still not happening adequately now, and they really need to solve that for the next five to 10 years. - Former Head of Strategy at AWS, North America

Interesting comment that all the cloud offerings are somewhat converging and the main competitive advantage for AWS is the scale of its data center footprint.

You see a maturing market in terms of customer understanding, in terms of market offerings, though it’s not yet a race to the bottom for pricing. In terms of differentiation though, at a technical level it’s very similar. Let’s look beyond the market offerings and what distinguishes them, the points of presence in terms of the network infrastructure and datacenters, and this is not articulated, by any of them, in terms of how many data centers they have, the capacity, the build plan. They don’t do that, but you can infer it in some cases. AWS has a fairly pervasive infrastructure and who knows what the capacity is. They just released their capex, by the way. You probably saw in Q1. Now at least we know how much they’re spending on it. Azure is catching up, but it’s less penetrated. GCP is much more episodic. They are much more patchy. I cannot speak to Oracle, but that’s a key point of differentiation. So, more coverage, more redundancy, and the ability to port workloads across, which also impacts capex needs. If you have a certain part of your datacenter for structure that’s under capacity, you can theoretically, as long as it doesn’t impact latency, port workloads there with full transparency and differentiated pricing, or not. That’s another competitive advantage of AWS and I think Azure as well, GCP far less so, and Oracle far less so. - Former Head of Strategy at AWS, North America

How Snowflake is capturing value from the cloud value chain amongst the hyperscalers.

Snowflake is unique because they’re not only multi-cloud, but they’re also allowing you to leave the data where it is, even if it’s on prem. What it doesn’t do is subsume the brands of the hyper scalers. It says, we’re going to play with your data wherever it is. We’ll let you access and get the full value of it. The AWS brand and the Azure brand and such still exist in the Snowflake paradigm, but what Snowflake is doing is facilitating customer choice, flexibility, and cherry picking value out of a value chain. - Former Head of Strategy at AWS, North America
Where we see things collapsing in terms of what’s driving value for the hyper scalers, storage category, very commoditized, hard to do anything new with storage, I would argue. Compute was on its way down, then custom silicon and reserved instances, bare metal, that has propped up pricing power there. There is still hope for compute for a while. And then database doing super well because it’s the front of the data category next to analytics and AIML. That’s the value bundle, and that’s where Snowflake is. By the way, Snowflake has the same network effects of dragging all the nice storage revenues as well. Yes, they are protected in AWS, but they are going in and just grabbing where the value is and increasingly, they are capturing that industry verticalization nugget that database has been struggling to capture. I think that is a strong set of things that is driving them. - Former Head of Strategy at AWS, North America

One question we're asking is how durable are AWS’ returns on capital?

Who knows what the cloud computing stack looks like in 10 years time.

One bull case is that few companies have the capital or ability to build the physical infrastructure to offer similar quality cloud computing infrastructure as AWS. It's estimated that the IaaS (EC2 and S3) offerings for AWS is ~70% of revenue; if this is an oligopoly, what does pricing and margins look like in the long run?

Some of the questions we're exploring include:

- If the PaaS layer converges and becomes more commoditised, how does this impact AMZN EBIT in long run?

- How should we look at the risk of SNOW-like services capturing value; could this happen across other verticals? If so, where does this leave AWS?

- How can AWS improve Partner Network collaboration and Marketplace adoption?

We plan on producing much more content specifically on AWS and SNOW over the coming weeks.

Carvana Finance GPU

Comment from a Chief Risk Officer of a subprime auto dealer competing with CVNA:

The investment bank who is helping us with securitization told us to expect a 200 to 250 basis point increase in the overall yield expectation [cost of funds] we are seeing from 2021 to 2022 at this point. - Chief Risk Officer of subprime auto dealer

As interest rates increase, the cost of funds will increase for CVNA’s securitisations. Put simply, investors are demanding a higher yield on the ABS than last year.

The question is how can CVNA adapt to maintain excess spreads on its securitisations?

CVNA could increase the APR’s to offset the higher cost of funds. It could also manage the mix between prime and non-prime loans that go into the pool.

Big players like Carvana or CarMax can control which accounts go into their overall pool, because that impacts the loss expectation which drives the overall ABS pool. They would look at the weighted average cycle, concentration across geography and the term, and the current challenge globally is the shortage of inventory which increases the inventory cost. That obviously drives the amount financed, from a customer standpoint, higher. Companies like Capital One or Ally can leverage their deposit to fund a high-risk customer or portfolio, so if they want to get better yield in the market, they can change what goes into their overall pool. - Chief Risk Officer of subprime auto dealer

CVNA management said Q1’s decline in finance GPU was mainly due to the rapid change in rates, rather than the absolute level of interest rates. This meant the cost of funds at origination was far below the cost at the time of securitisation.

We previously discussed how CVNA can manage finance GPU. It’s currently too difficult for us to understand given the many variables at play; higher rates, declining ASP’s, higher cost inflation impacting subprime customers, etc. We will be following CVNA closely over the next few quarters.

Investor Dialogue: Judges Scientific

David Cicurel, founder and CEO of JDG, is probably one of the best capital allocators in the UK over the last 10 years.

One of our subscribers once asked him how?

His answer: because he is scared.

If you go to the website under the acquisition section and David's name is on there. He is the one, I guess, still negotiating the deals. I think he just gets it. His background is a turnaround guy and, building this company from the beginning, he just understands. What they have done has worked; the formula has worked. He found the broker, who introduced him to the first company and found this little niche that is phenomenal. In my first conversation with him, back in 2018, I asked him how they had been able to create so much value via acquisitions when a lot of companies just vaporize value. The wording he used stuck in my head. He said, because I'm scared. Those were his exact words. He said, every time I sign on the line, I'm scared; I've got all my net worth tied up in this. I've got friends and family who are shareholders, and I'm scared we are going to do a bad deal. He said, I know you are going to ask me about succession. He said, the biggest question for me is, can I find somebody who is going to be as scared as I am. I think he just wants to make sure they don't do a deal, just to do deal, that they don't overpay. They've done 20 deals; that's not a lot. I just think he's very, very careful. - Professional Investor, IP Subscriber

He has made 20 acquisitions in the last 17 years. This is discipline. It takes a certain type of character with significant skin in the game to be this patient.

JDG is an excellent business with an excellent capital allocator at the helm. And it's still relatively small at ~£500m market cap.

The benefit of early serial acquirers is that the hurdle for inorganic growth is low. If we assume 3% organic growth (the 10 year average is ~5%), JDG only needs to acquire £1.5m - £2m in EBIT per year for the next 5 years to grow revenue 10%. This is roughly 1 company per year, similar to its historical record.

At 5x EBIT acquisition multiple and constant leverage, this will return ~12% including dividends.

What makes the thesis complicated is the recent Geotek acquisition, the largest in the company’s history. We recently discussed Geotek and the risks involved in more detail.

Elastic

ESTC has closed the feature-gap between the on-prem and cloud product which is driving greater cloud adoption which drives higher paid conversion.

That’s why Elastic Cloud is so powerful and why Elastic has been pumping a lot of money into the cloud. Because if you're using Elastic as a service, you pay. Everybody pays. The downloaded version is very hard to monetize, and I don't know if you're aware of this, but up until about six months ago, there was a big gap between the features you get when you download Elastic versus the features you get when you use it as a service. That was a big hurdle. Even if you understood that moving to the cloud is easier because you don't have to muck with downloading it and keeping it up to date, the lease, patching it, tuning it, you still had this feature gap, and Elastic has closed that gap. That hurdle is gone, which I think is a huge plus. It's not that hard to walk into a company that's using Elastic on-prem and make up a very strong business case that they will save money if they move to the cloud. Elastic is not the easiest product to download. Every four or six weeks, there is a major new release, so it becomes a full-time job for a small team. That’s a big sales motion for Elastic. All they have to do is convince the customer using elastic on-prem to move to the cloud. They've been doing that, it's a very dominant sales motion for Elastic, and that's why the cloud becomes key to monetize and spread. - Former VP at ESTC

An interesting observation that buying decisions that are today separate between security, observability, etc will converge. This could put Elastic in a stronger position.

the market plan to watch is the convergence of DevOps and SecOps into DevSecOps. If you go to LinkedIn, there are already job postings for DevSecOps managers from some companies that are early adapters. In other words, if those two converge, then Elastic's story becomes much more urgent. Over time – and it will take some time – purchasing decisions and buying centers that are separate today will converge. If the DevOps and SecOps teams merge and security becomes part of development from day one and coding, you want a tool that can speak both languages. You want a common data plan; you want a common visualization plan. That makes Elastic’s story that much more urgent and powerful. - Former VP at ESTC
In my opinion, Elastic has to do everything it can for that convergence to pick up the pace, and the way to do that is what you said. Find early adopters in financial services or healthcare, and show social proof. That's what Elastic has to do, more social proof. We need to see recognizable names embracing Elastic. The other big part of it was Shay stepping down as CEO. He was a reluctant CEO; he didn't want to be the CEO. He loves technology; he's a nerd, so he did it until he could find somebody and hand the job to them. Ash is top-notch. I had the chance to work with him almost daily. He has a strong security background, having worked at Akamai and McAfee. He's very customer-obsessed, which Shay was not. - Former VP at ESTC

An interesting observation that IT buying decisions across security, observability, etc may be converging. This could put Elastic in a stronger position versus competing with DDOG / SPLK and other vertically-focused best of breed SaaS cos.

Victoria, Carpetright, & UK Carpets

In a recent interview with the former Procurement Director at Balta, a Victoria competitor, we discussed the flooring supply chain and how raw material prices have increased over the last 18 months.

This has put pressure on retailer margins. The Former Regional Manager from Carpetright believes this could pose a risk to VCP if retailers decide to go direct to end manufacturers (VCP is both a manufacturer and distributor) and miss out VCP or get VCP to take some margin pressure with the retailer.

The retailer might go direct. They might hypothetically say, these are my top 10 volume lines; how close can you get to these in specification? What's the cost. It's well known in flooring what Victoria ranges are supplied to people like Carpetright. So new competitors in the market, such as Likewise and Furlong, they’ll all be looking at that and probably leaning on an open door with Carpetright to say, look, let’s have a conversation about alternatives to those products. That’s probably the biggest challenge for the Victoria Group at the moment. - Former Regional Manager at Carpetright

There are two interesting characteristics of the soft flooring industry that make this a potential risk:

- End customers typically don’t know the manufacturer brands

- With advances in technology, all product specs can be closely replicated by any manufacturer

Retailers can easily change the spec of products without customers knowing. Manufacturers have the ability to produce any type of product at the request of the retailer.

The big challenge is distributing it on time. And it can’t be late because the end customer has moved all the furniture around ready for installation.

This is VCP’s real competitive advantage: it owns distribution. It’s the only UK manufacturer that owns a distribution business internally.

A competitor to Victoria would be the AW Group. The AW Group doesn't have the logistics set up that Victoria does. It's a good brand, all produced in Belgium, and their problem is they can’t get the stock in the UK to the customer as quickly as required. They don't communicate with the retailers, and AW as a brand is probably losing market share because they don’t have the distribution network that Victoria Group does. I’ve seen it first-hand where it’s a great brand, product, and commercial proposition for retailers, but without logistic support, it will fall flat. - Former Regional Manager at Carpetright
You're now finding people replicating AW to offer better service with the same spec. Competitors of AW are now saying to retailers, look, you've got an AW stand in your store, it looks great. I get it's a brand in flooring but, realistically, if you were to swap that to a like-for-like replacement with better servicing and better margin, are you interested? The answer is yes. So Victoria's still got that challenge, but Victoria still has the advantage that they're seen as a good service provider within flooring and established products. - Former Regional Manager at Carpetright

Tremor International

Almost 50% of advertising spend is retained by middle-men in the ad-tech value chain. Tremor looks to reduce the cost by integrating a DSP and SSP in one platform.

Let's ignore what Tremor specifically focus on which is the video side, and answer your question purely from a vertical integration opportunity perspective. I always struggled with the number of hops an ad buyer takes in the ecosystem. Before a single dollar of buy from a brand eventually hits the publisher, it takes many hops, so there is an advantage to having an integrated platform because you eliminate several hops and present a single bid-driven environment to buy opportunities. If an advertiser has to find both an SSP and DSP to do this, that's at least two bids so that presents opportunities for scale. The technology works in such a way that if you have an integrated platform, you don't have to match IDs so you don't lose fidelity as your platform scales.
If I buy directly from Twitter using their platform, or if I have Twitter plugged into another DSP, I will see a different scale in those two places, whereas a vertically integrated player can offer advertisers better scale and performance. Integration also offers pricing advantages as they will see better bid returns which ultimately help them find better economies in terms of how that ecosystem works. I see SSPs and DSPs as fairly comparable technologies. An SSP is a bidding platform to bring various publishers together, whereas a DSP uses a bidding platform to allow ad buyers to choose from that supply. That is effectively the same concept in terms of, find the highest bidder and place the ad opportunity. - Former VP at MediaMath
If you look at the other media types, whether it's search, social or display, the winners/losers have largely played out. Google lead search and Facebook won social, so having this integrated platform in an emerging media channel is where they have a real opportunity. eSports, gaming and VR/AR and the metaverse means a lot more new video inventory will be available, and in the ad ecosystem there will be much more fragmentation. Google won search and Facebook won social because they are much less fragmented in the current market. If you look at TV, there are four large studios, so if you take that same analogy towards the video marketplace, you can put Snapchat, Twitter and TikTok in the same spectrum of video inventory as you can syndicated video from NBC or ABC, so there is a need for an aggregation platform like Tremor. If they come to the table with an integrated platform which brings DSP and SSP together, that will present real advantages while the ecosystem is still maturing - Former VP at MediaMath

However, a DSP’s aim is to get the highest ROAS for the buyer which relies on a good price, and the SSP aims to monetise the inventory of the publisher at the highest price. An integrated platform seems to have a fundamental conflict of interest?