Fastly & Cloudfare - Usage & Models

Now bear with me, I only joined my company less than a year ago, and I was new to the industry. So I am by no means very experienced (I must stress that, also in relation to investing and the companies and industries we’re talking about on here). I’d also like to add that this board has been the most useful source of information I have found since I started out, I have other sources too, but this has been the most valuable for me, learning about investing and the companies discussed here. I read most to every post, and without Saul there would be no board, so all thanks to him.

Despite my inexperience, I try to offer some insight or perspective where I feel like I have something to offer (which in the last few days has spiked with Fastly posts - previously Zoom is where I’ve felt most comfortable talking about, maybe because it’s one of the companies whose product I’ve actually used and is therefore more tangible, other than Peloton).

So in my company we do have a usage based model, which I might be able to offer some more insights on.


Yesterday RainyDayFund posted this:

Matthew Prince "The vast majority of our revenue, more than 95% we’ve built upfront on a subscription basis that generally gives us good visibility into our future results.
Another consequence appears to have been and we saw the peak in customer concession requests earlier than other companies that fill in arrears on a more volatile usage basis. One thing we are seeing increasingly is customers who were surprised by their large usage based bills that other vendors now coming to us for predictable consistent pricing. No one likes to be surprised by a bill. And we believe the consistency of our staff approach is not only more predictable for us, but also builds trust and wins loyal customers over the long-term." - END

I’ve thought more about this overnight.

I hadn’t seen this before. The inference does concern me. I’m just not sure how much to apply my own business experience to Fastly, because they are very different companies, in different industries and with very different products. And I just don’t have much insight into Fastly’s usage based model.

But what I HAVE seen from my own company, is that customers don’t like to be surprised by a spike in invoices. They try and negotiate new commercials, to get more predictability in their invoicing. So I can concur with that sentiment first hand.

When I apply this to Fastly though, I’m not sure how much an Amazon or a Shopify will contest this - more likely their smaller customers. I think this is where Fastly building close relationships with the customer, and a developer focused platform comes in. Does the value of Fastly’s proposition mean it’s unlikely that customers churn over to Cloudfare? Or are the differences not enough, that the customer would rather more predictable invoicing.

There are ways of making invoicing more predictable with a usage based model. One way we do this is by negotiating a minimum revenue commitment. So the customer knows they will pay ‘x’, and any overage is upside (for us). Again - I don’t know if this is applicable at all to Fastly.

There is perhaps a possibility that Fastly surprised some customers with a spike in invoices in Q2 due to Covid usage, who negotiated a refund in Q3. This could be their ‘usage impact’ from a ‘few customers’. I’ll look out for this in the earnings call.

If a customer is pushing back about an invoice for unexpected usage, it may be better to take this approach (eg. refund/discount), to maintain the relationship and foster goodwill. This is one of the challenges of the usage based model.

However, lest we forget the benefits of a usage based model, or why would any company have one? Why wouldn’t every company just be subscription based? As Bear says, it’s a double edged sword.

There is arguably more upside to a usage based model. Again, I don’t have any great insight into Cloudfare’s model, and I don’t have any insight into their pricing but stay with me - a hypothetical. If a customer spikes their usage, which presumably goes through their infrastructure and which there will be a corresponding cost for, how is Cloudfare getting compensated for that? Cloudfare therefore needs to actively grow by finding new customers, renogotiating subscriptions with existing customers, or making new products to upsell to them. Cloudfare’s model is more sales intensive, requires more R&D and network costs and so eats at their operating margins.

Compare Fastly and Cloudfare’s operating expenses in the most recent quarter (Q2):

**As a % of revenue**                  Cloudfare            Fastly
**Sales and Marketing                   52%                 33%**
**Research and Development              28%                 22%**
General and Admin                     21%                 24%

I had noted in a previous post that Fastly has been accelerating it’s S&M expense as a % of revenue, to drive it’s Enterprise customer growth (as per their management reports), and suggested that due to the slow sales cycle of the size of their customers, it may take a while to see the benefit of this.

So what does this look like in terms of margin? Well while Fastly’s losses as a % of revenue in Q2 of 19% appears at first glance appears better than Cloudfare’s 25%, Cloudfare has improved its losses as % of revenue from 40% in Q1 - while Fastly’s has stayed at 19% each quarter. This suggests that Cloudfare’s margin is more rapidly improving, evincing a clearer path to profitability perhaps.

Now back to usage, Fastly will get upside for increased data usage through its platform/servers. And it doesn’t have to actively chase for this upside, meaning it can focus its R&D more on improving its proposition. Perhaps Cloudfare’s higher R&D can also be explained by the launch of various products in quarter such as Cloudfare Workers and more recently Cloudfare One for Teams. However, while Fastly is now increasing its S&M each quarter to try and stimulate its Enterprise customer growth, a high S&M seems a necessity for Cloudfare’s current business model. With the launch of these new products, perhaps we can expect Cloudfare’s S&M expense to start to come down, as its more rounded proposition begins to sell itself increasingly.

Again, I am speculating somewhat - but I don’t believe it’s as simple as saying a subscription model is superior to a usage based model, or vice versa. There are implications to both, benefits and disadvantages to both. And I have equal positions in both Fastly and Cloudfare right now.

I know there is a focus in growth investing on top line growth, because that’s what growth investing is about, right? Nail the land first, move on to the expand. So perhaps people would rather a higher S&M spend and higher revenue growth. And where’s Fastly’s R&D going, since Cloudfare has launched all these new products? Compute@Edge? Well, what I’ve learned from the likes of Peter Offringa and his foundational thesis on Fastly, is that Fastly has invested a lot on improving its points of presences and its infrastructure (with much more technical analysis than I’m able to offer), so that it is well poised for the advent of edge computing - a market which should ultimately drive this top line growth we’re looking for. This helps drive my long term perspective, although Cloudfare seems to be executing excellently at the moment. I saw AnalogKid70 ask what Peter Offringa’s view was about Fastly. Well, he has a Twitter account, and he has said yesterday that he hasn’t sold any shares, as his foundational thesis has not changed, he wants to wait to hear management comments on Q3 report, and to get an update on 2021 initiatives. In other words, he maintains an outsized position.

We do already know that Fastly hasn’t seen the usage upside it did in Q2. While initially I didn’t expect it to have this one off upside again in Q3, because we’d had a clue about this with the false flag of increased expectations going into Q2 earnings, we still don’t yet know the reasons why.

I was at first optimistic that Fastly usage in Q4 was going to have some tailwinds (namely in its e-commerce customers). Why its customer usage was impacted in the latter part of Q3 seems indicative of what the run rate is going into Q4. It’s guided to be +18% from Q3-4, so it needs to be going some. However will the incremental Signal Sciences revenue more than offset this, and lead to raised FY guidance. And if so, how will the market react.

Usage seems like such a variable element, that without understanding the driver behind it, it is difficult to make a decision based on this alone, for me. Let’s see what management have to say.


When a customer’s revenue is primarily usage based, then they often would prefer a usage based model for major expenses over a flat subscription model. When a customer’s revenue is primarily user or subscription based, then they would prefer a per user or per subscription model. Clients also must manage their business and their investors’ expectations. That is much easier when their major input costs match their revenues. It’s not necessarily one size fits all. Although for most expenses, a flat user/subscription fee is probably preferable because it is more predictable for businesses.


ATF, I agree with everything in your last post. There is so much to there there. I do however keep coming back to the FSTLY CC. I recall thinking like wpr101 stated here before, ‘Fastly’s last conference call had the shortest prepared remarks sections I have ever seen for a public company. I believe it was roughly 7 minutes, before they started taking questions. Something felt off about that, in addition to adding only 3 new enterprise customers.’

I think Fastly’s best is yet to come also (struggling against investIng in hope, Compute@edge, I was at 15% port last week). There is soo much greenfield opportunity;but, I can’t help comparing Fastly to Cloudflare. Compare the conference call of Cloudflare which gave a ton of extensive detail on their business, both technical, financial, and sales side.

But, to your main point about usage, like wpr101,’wondering why the usage based model is suddenly hurting Fastly when it’s helping Twilio still? Twilio had to make the opposite move of revising guidance up mid quarter.’.

Thanks for your consideration,



I assume that usage or subscription, the way of paying can be easily changed to fit the market demand. Or maybe offer both.

1 Like