Interview: Flywheel + Arb Insights
See what ChatGPT calls an "illuminating discussion"
Last week I sat down with James Griffiths, Davide Vetrale, and Dan Elbaz to discuss 2023 holiday shopping data, and the collaboration between Flywheel Alternative Data and Arb Insights. Below is a brief summary (yes, I used ChatGPT here) as well as the full transcript of our discussion.
Both teams will be in attendance at BattleFin Discovery Day Miami, January 24-26th. If interested in learning more about the partnership, or either companies’ offerings, please contact the individuals directly or reply to this email and I can connect you.
If this email was forwarded to you, please subscribe using the link below to receive the weekly It’s Pronounced Data newsletter of M&A, fundraising, people moves, articles, podcasts, job opportunities, and upcoming events.
Disclaimer: the content of this discussion and transcript is for informational purposes only and is not investment advice.
ChatGPT Summary (because why not?)
Dive into the dynamic world of alternative data and analytics with industry leaders from Flywheel Alternative Data and Arb Insights in this illuminating discussion. Explore recent developments, successful collaborations, and key insights into the trends shaping the data landscape. Gain valuable perspectives on solving challenges in understanding gross margin, the outlook for 2024, and the pivotal role of collaboration in driving innovation. A must-read for investors, data enthusiasts, and industry professionals seeking a comprehensive overview of the current landscape and future possibilities in the data and analytics sector.
The discussion features James Griffiths, GM of Flywheel Alternative Data Davide Vetrale, CEO of Arb Insights; and Daniel Elbaz, CTO of Arb Insights. The conversation is facilitated by Dan Entrup, and they discuss various aspects of their companies, recent developments, and trends in the data and analytics industry.
Key Points:
Introduction and Company Overview:
James Griffiths introduces Flywheel Alternative Data, which serves investors and consulting teams by repurposing data sources from its parent company, Flywheel.
Davide Vetrale and Daniel Elbaz introduce Arb Insights, a data platform integrating alternative and traditional datasets to provide investors with unique insights.
Recent Developments:
Flywheel recently rebranded from Ascential and completed a significant acquisition by Omnicom, becoming Omnicom's dedicated digital commerce practice.
Arb Insights partnered with Flywheel to leverage their data for building a gross margin model.
Problem Solving:
Flywheel addresses challenges in understanding gross margin trends, especially for smaller funds lacking resources for comprehensive analysis.
Arb Insights focuses on providing actionable insights by integrating multiple datasets to understand various aspects of a company.
Collaboration and Partnership:
Arb Insights approached Flywheel with the idea of leveraging their data for building a gross margin model.
The collaboration has been successful, with Flywheel providing data, and Arb Insights building the model.
Gross Margin Model and Customization:
Arb Insights' model considers factors like discounting rates, underlying costs of goods, and a company-specific COGS Index.
The model is ticker-specific, considering variances across companies and industries. It requires a fundamental understanding of the underlying company.
Holiday 2023 Trends:
Consumers showed sensitivity to discounts, with an increase in discounting rates, but higher average final prices due to inflation.
Sales growth varied across companies, with apparel doing well and e-commerce outperforming brick-and-mortar.
Outlook for 2024:
Flywheel anticipates focusing on providing more proof points for research teams, emphasizing the need for compelling cases.
Arb Insights hopes for increased collaboration among data providers and less competition [amongst data providers], with a focus on effective integration into the investment industry.
Data and Analytics Industry Predictions:
The industry may see increased use of alternative data in private equity.
Collaboration among data providers is encouraged, and a need for prioritization and understanding the investor landscape is highlighted.
More and Less for 2024:
More private equity data use and partnerships are anticipated.
Less entry of new data providers without an understanding of the investor landscape is desired.
Personal Note:
James Griffiths mentions expecting a second child in the summer and humorously suggests doing business early to ensure a comfortable July.
Full Transcript
Entrup: James, Davide, Dan, it’s great to be hear chatting with you today. To start off, can you please introduce yourselves and provide an overview of your companies?
Griffiths: I’m James Griffiths, GM of Flywheel Alternative Data. Our wider business serves thousands of consumer brands with their e-commerce intelligence needs. My team gets to repurpose some of the same data sources to create products for investors and consulting teams.
Vetrale: I’m Davide Vetrale, I’m the CEO of Arb Insights, a data platform that integrates alternative and traditional datasets to give investors access to unique insights and forecasts.
Elbaz: I’m Daniel Elbaz, I’m the CTO of Arb Insights.
Entrup: James, it’s been an exciting year for Ascential moving over to the Flywheel brand. Tell us about what’s new. And can you talk a little bit about the data and where it comes from?
Griffiths: Yeah, definitely. It’s been a very busy few months. You might have followed that a few months ago Ascential rebranded and integrated 11 businesses serving consumer brands and my business serving investors, into one new company called Flywheel.
Flywheel has got some fantastic innovations particularly around how brands buy retail media and advertising services on hundreds of digital commerce platforms. And with retail media being such a fast growth segment, Omnicom decided it wanted to make Flywheel its biggest-ever acquisition in a deal that was completed a couple of days ago for around $835M. So Flywheel is now Omnicom’s dedicated digital commerce practice.
For our investor clients, the main dataset we've become well known for in the alt data space over the last five years, it's business as usual - the same data points, coverage, etc, same compliance and data collection processes. But really this gives a significant opportunity to my team now to access so many more data sources from within Flywheel that we can look to productize and bring out to the investor audience.
Entrup: Great, and congratulations on the acquisition by Omnicom and all the success. James, if I can stay with you for a moment, can you talk a little bit about the problem that you're solving for with your data? I'm curious, before Ascential and Flywheel, what did the world look like without this data? What were investors and business decision makers reliant upon before having access to your data?
Griffiths: From my perspective, there are many great data vendors out there that have helped buy-side investment managers create really good predictability capabilities for revenue. But I think the further that you go down the financial statements, the more that it's been troublesome, especially in recent times, to really get a handle on what's going on around gross margin.
So over the last few years, my teams have been delivering a pricing and discounting dataset to lots of earlier adopters. Some of those bigger funds with lots of resources have been able to take our data inputs and enrich their own models to come up with some good outcomes. But when you get beyond those really big tier 1 funds, you quickly get to a lot of organizations that would like to have the intelligence but don't have the resources or the bandwidth. And so there's been this gap on gross margin capability that I think has been a big problem to solve for the last few years.
I'm very conscious that I've got one input to the answer. So I've been looking for partners that can come together, collaborate, and do more by working together. I think it serves the longer tail of the investor audience, but I think it also gives us some ammunition to go to funds of all sizes to say “here's some better pre-sales tools. Here’s some proof points in our sales cycle that can give you confidence that you should commit the resources to evaluate us through a trial process.”
Entrup: Dan, Davide, you guys both have backgrounds at some of the tier 1 funds. Can you tell us a little bit about the problem that Arb is solving for, and similar question, what were people doing before and how are you solving that problem?
Vetrale: Absolutely. Our goal is to provide investors with deeper and unique insights. What I mean is that we don't offer isolated trends or forecasts which will then require further analysis by a team of data scientists to become valuable for investors. Instead, our focus is on providing actionable insights, which sometimes require the integration of multiple datasets to gain a holistic understanding of the company. It’s essential to understand every aspect, including all of a company’s revenue segments, distribution channels, and geographical performance. That’s the challenge Arb Insights aims to address. Our platform was conceived based on our experiences at major multi-strategy hedge funds where we worked with a diverse group of investors from various sectors and different investment styles. We aimed to build a platform that is flexible enough to serve both discretionary and quant investors.
Entrup: And who's your ICP at some of these funds? Is it a portfolio manager? Research analyst? The data team? Is it multiple personas? Who is the ideal customer?
Vetrale: Our primary targets are investors, specifically research analysts and portfolio managers. However, our platform is also highly suitable for data teams. We’ve designed it with various functionalities to serve a broad spectrum of sophistication levels. This ranges from teams seeking deeper customization of our forecast models to those aiming to construct portfolios and backtest strategies.
Entrup: I'd love to hear about how this collaboration came about. Who reached out to how? How did you guys think about this? Tell me a little bit about where the collaboration is at today and some of the things that you guys are excited about to work on together.
Vetrale: I can start and James can jump in. During the pandemic we realized that gross margins were a critical issue. I was a portfolio manager at that time, and I remember that during earnings calls company management was spending most of their time discussing commodity prices, freight costs, pricing strategies, and their impact on gross margins. We wanted to build a gross margin forecast but soon we realized that to do that we needed to integrate a number of datasets, quite different from each other.
For many companies a big component was promotions and pricing. To build a good model, you always want the best datasets. We've known Flywheel data for a long time now and we know that they have probably the most comprehensive dataset for promotional activity and pricing. We reached out to James and we shared our idea of leveraging the Flywheel data to build a gross margin model. This was something that James really cared about, so we decided to start this partnership where Flywheel provided us the data, and we built the model. After a few months here we are.
Griffiths: I get approached by many organizations that say they can help me. But when I met the two guys beginning of last summer, their credibility from their backgrounds was so strong. They were so clear about what they wanted to achieve, It was a really easy decision for me to make our dataset available to them.
As they were able to show strong progress through the subsequent months, it really lent itself to start a strong partnership. And then you get into the exciting world of starting to show it to clients and prospects and you start to see the questions coming back and forward and just the appreciation for the lengths that Dan and Davide have gone to go further than potentially anything that anyone in the market has done before. So it's been a great start to the partnership.
Entrup: Dan, Davide, can you talk a little bit about some of those other inputs? You mentioned other alternative datasets and traditional data inputs. What are some of the other metrics, KPIs, or even whole datasets that you guys are incorporating into the gross margin model?
Elbaz: So kind of like we were mentioning, it really has to be a comprehensive model that takes into account a lot of the different factors that affect gross margin. So it's not as simple as, for example, a revenue model that a lot of people are used to. The biggest components are obviously going to be the components within the Flywheel dataset, the discounting and promotions. When we look at the discounting rates of a company from the Flywheel dataset, we look at both how deep those discounts are and how broad those discounts are across the product range of the company, as well as how average prices are changing on a year-over-year basis. Those are some of the biggest components from the Flywheel dataset. On the other side, we take into account the underlying costs of goods of a company. That’s the other big factor.
We construct factors that measure how prices are increasing (or decreasing) versus the underlying costs. That’s what’s going to affect gross margins at the end of the day. So to gauge that, we built what we call the Arb COGS Index, which is a ticker-specific indicator of how the underlying costs are trending by using a combination of different macroeconomic indicators. For example, let’s say apparel companies are highly sensitive to pricing of cotton and different synthetic fibers. Nike and other footwear companies might be more sensitive to rubber. Freight costs are big factors for a lot of e-commerce names. Those are some of the kinds of indicators we look at to create our COGS Index. Once we have our Flywheel promotional discounting and pricing factors, and account for costs, we create a time series model to forecast a point estimate of gross margins.
Entrup: It sounds like this is not one size fits all, and a lot of customization goes into this. How much customization is there though? Is it every company? An individual ticker or product level needs its own specific model? Are you able to group say footwear manufacturers, separate from electronics, separate from clothing & apparel? How granular is the customization? Is there a model that fits multiple companies?
Elbaz: You really have to be company specific in terms of the underlying dataset and features. So there's quite a bit of variance and discrepancy across tickers, even within the same sector or industry. The Flywheel factors are ticker-specific so that accounts for differences in pricing and promotions. Our Arb COGS Index Is also ticker-specific in the sense that the weights across commodity and macro indicators are different by company.. The way we do that is with a semi-systematic approach where we use NLP models that crawl the management commentary and public statements of a lot of these companies and pull out themes. For example Nike is talking about rubber prices affecting their underlying margins, and inflation, we’ll pull those out, map those indicators to specific macroeconomic variables, and then combine those at each ticker level. So it is definitely ticker-specific and we think that's quite important. The model itself we do generalize, and the biggest thing is not overfitting. So we do have a general modeling process where once we have those company-specific indicators and factors, we feed that into a systematic modeling approach that we use across all of our KPI forecasting.
Entrup: You mentioned not overfitting, which is a term that that's come up for years around data science in investment research in particular. Not about your prior firms, but putting your investor hats back on - and you both worked at very sophisticated shops - what are some of those challenges that investors are running into or mistakes that you guys are seeing that people are still making in the industry as they use data to analyze individual tickers?
Elbaz: There's a few things. I think the big thing that we always talk about is using point-in-time data. A lot of alt datasets and vendors will do some restatements. You need to make sure you are taking those restatements into account and that your models are realistic in the sense that that’s what's going to happen when you're actively trading. So we always look at point-in-time data.
The other big factor is making sure you have a fundamental thesis for your underlying variables. This is more of an issue we see sometimes with quants where they might not have a fundamental understanding of the underlying company. So they might include some variable that doesn't exactly make sense or forecast a KPI that doesn't make sense for the dataset they're working with. One thing we do on the Arb Insights platform, is we do a lot of mapping of the datasets to the KPIs, and figuring out which datasets actually make sense for the specific KPI. We’re doing a lot of filtering and mapping on our side. It eliminates some of that [challenge/issue], but we think that having a good fundamental sense of which variables you’re using and which KPI you’re forecasting is a really good start.
Entrup: I see how it's unique and you're solving a real pain point for some of these funds out there. I want to switch and chat Holidays 2023. What have you guys seen in the data? Any interesting takeaways overall? And if there are any specific companies or trends that you want to highlight - is anyone performing particularly well or poorly? Let's talk holiday 2023
Elbaz: We've seen some interesting trends, particularly when we're looking at the Flywheel dataset in combination with some other data points. With Flywheel we get a really good sense of the promotional environment of this holiday season versus 2022. We also look at transactional data to see how the consumer is spending and reacting to these discounts. And also looking at the macro data we were talking about to see how costs are trending versus prices as well. If you're looking at discounting and promotions, there's a lot of variance across tickers, but for most companies we're seeing that discounting was up slightly from last holiday shopping season both in terms of how steep those discounts were and how broad they were, but despite having a year-over-year increase in discounts to consumers, the average final prices were still higher year-over-year because of inflation. Full prices have already been higher year-to-date that have more than offset a small increase in the discounting rate, so consumers were still dealing with higher prices at the end of the day, even when getting a small discount. At the same time, despite price increases for consumers, we think most companies probably started to see some inflation relief in the last couple months of the year versus the same period last year. A lot of raw material prices started stabilizing or even coming down. Shipping costs were lower year-over-year. Even packaging materials and products were lower year-over-year. So it should be overall a positive environment for improving margins. That's going to be really interesting to see.
We’re also tracking how these discounts and promotions are affecting consumer spending, especially across companies. If you just look on aggregate, sales and consumer spending was up a bit from last year - around low to mid single digits. Most of that sales growth is probably even just due to inflation and rising average prices, but there's a lot of variance in holiday sales growth across companies. Some of that variance can be explained by industry and product. For example, apparel has done particularly well. Also, it varies a lot by channel - e-commerce has grown significantly more than brick-and-mortar, for example, which is not really surprising. The interesting thing is that a lot of the variance in terms of sales growth across companies can actually be explained by the pricing and promotion data from Flywheel. If we were to rank companies in our universe from the lowest discounts to the highest discounts during the season, there's a clear pattern that consumers are spending more at companies with steeper discounts, and that's happening at a higher rate from even the last several years. It’s pretty clear that consumers have been particularly sensitive this season to promotions and discounting.
Entrup: James, anything you want to add in terms of trends that you're seeing? A lot of what we just learned is based on your data, but is there anything else you’re seeing from your seat?
Griffiths: I'll go for one a few steps away from discounting. I'm always fascinated by new product ratios. Generally, I believe they are a really telltale sign of how confident a brand or retailer is right now - for instance, if their sell out is high then they are often excited to increase the flow of new products to match that higher demand. Earlier today, with data for the full month of December, I looked at 350 retailers and around 50 of them had neutral or positive new product ratios. So, nearly 300 retailers in my sample set chose to launch fewer new products in the recent period than they did in the prior year. We often see new product ratios being reduced so that we can get to that natural cycle of clearing out excess inventory.
Entrup: Looking at the holiday season, you have Singles Day, Black Friday, Cyber Monday, and for me personally, Christmas Eve when I've forgotten to buy presents for everyone and I venture out in the cold to do that. Were there any notable trends or notable standout events in particular? Are you seeing more Cyber Monday versus Black Friday, which was historically more brick-and-mortar, but obviously a lot has geared towards e-commerce. Are consumers still getting excited about those holidays? Or is it from my anecdotal, N of 1 perspective a lot of these sales seem to persist throughout or some started even earlier. What kind of trends are you guys seeing in terms of earliness of sales starting, and discounting, or if there are any notable standout events or notable trends whether positive or negative.
Elbaz: I’ll mention a couple of things. We mentioned this earlier, but it’s not surprising that e-commerce is continuing to grow at a much higher rate than brick-and-mortar, even more so than prior years. In terms of the timing, a lot of these sales and promotions are happening earlier. So well before Black Friday. And finally, we’ve seen consumers being very sensitive to discounting during the 2023 holiday shopping period so it’s going to be interesting to see if that trend continues into the new year as well.
Entrup: Actually on that note, that leads right into my next question. What does this trend mean for 2024? Whether Q1 or full-year or going into next holiday season or even the next 1-3 years. Anything notable that you expect to continue? Anything that you expect was maybe a short-term event that will go back to the way it was?
Griffiths: I touched on the clearance cycle. I'm looking to see how realistic the retailers and brands are being about their inventory levels. Did they get the right mix of breadth of discount versus depth of discount? And on the whole, what happens over the course of the next 3-5 weeks when stock needs to be rotated, especially in those seasonal names that need to clear the decks before the next set of inventory comes in. I think there's a real tell tale period to come and that ties back to what I mentioned earlier about slowing down the new product ratios, probably to tweak the levels of inventory that is still left.
Entrup: Alright, thanks for the great insights on Holiday 2023. Let’s chat about the overall data & analytics industry. Any predictions for what's going to happen this year? I published a post this morning highlighting a bunch of the things I came across on LinkedIn in particular and other newsletters of different people's predictions. Some talk about M&A, some talk about private equity activity picking up, IPOs being down - that's more about the general investment world. I'm curious what you guys see or if you have any predictions for what's going to happen in the overall data, analytics, and investment research space.
Griffiths: I'm gearing up my business to provide substantially more proof points for everything we take to research teams. Long gone are the days of experimental data budgets where tier 1 funds would splash a load of cash around on things that looked good, that they hadn't really got their teeth into, but they had some spare money and they wanted to play with it. My sense from how the market has evolved is everyone has to work that much harder to present a really compelling case, at every stage from introduction through the early stages of securing trial time and through the subsequent phases up to hopefully buying the dataset. We're really focused on our side around going that much further than what we have done in the past, to showcase why we believe we're sitting on unique value. And that leads itself nicely into partnerships like what we're teeing up with Dan and Davide. And then the second part is that I imagine that we're going to see some innovation somewhere around the conference circuit. I'm a huge advocate and financial supporter of the conference circuit, but marketing budgets are tight, travel is really expensive, and I think that data vendors and data buyers are going to look for interesting ways to keep coming together, and it might need to partly evolve somewhat from what we've been used to over the last five years.
Entrup: So low-key happy hour? Free, pay for your own drinks.
Griffiths: I think that we've got to find a way to keep the engagement and the Innovation flowing back and forth between both sides of the table, but I'm not so sure that it's going to be flying around the world at the frequency that we've all done in the past.
Vetrale: I totally agree with James. I also think that over the last few years there’s been an explosion of data providers. Sometimes data providers come out with small panels that have no history, and they just claim that they have the best data out there. I think that this is actually creating a risk for investors because they are getting confused about the alternative data environment. They see too many dataset options, and they don't know on which datasets they should invest their time and money. It is hard also for data teams, because when you have 50 datasets in the pipeline, you have to assign priorities, and if you work on the wrong dataset, you waste a good amount of time. That's something that I see more and more. I think that the integration of alternative data into the investment industry will be more effective if data providers start collaborating rather than competing, especially when they provide different types of data.
Another trend we have seen over the last few years is the growing use of alt data in private equity. This is expected to continue this year too.
Entrup: More and Less for 2024. I think you guys answered some of it. It sounds like we'll see more private equity data use. Hopefully we'll see more partnerships. I'm a huge advocate for partnerships. James, you're hoping to see less needs for travel and budget, except everyone's open to Miami in January, so we'll see you guys in a few weeks. Anything else you want to see more of or less of from the data world, or even from the client side from the investors that you guys are working with?
Vetrale: For the less, I mentioned less new data providers that come to the market without an understanding of the investor landscape.
Entrup: I bet those providers’ one pagers say it’s an absolute need for investors everywhere.
Vetrale: Yes.
Entrup: Awesome. James, Dan, anything to add?
Griffiths: We're having a second baby in the summer, so there's going to be a lot more diapers in my house and a lot less sleep. So if anyone wants to do their business with us really early this year and give me a comfortable July, I'd really appreciate it.
Entrup: That's really exciting. And so we're anticipating diaper sales are going to be up next year?
Griffiths: Diaper sales in Central London are going to fly!
Entrup: Excellent.