TikTok improved their data compression

This may explain (some of) TikTok’s reduced usage: TikTok changed their data compression algorithm resulting in a better compression ratio (compressed data/uncompressed data), which resulted in a reduced bitrate, which implies less (usage) fees to FSLY, as reported by Robert Majek, analyst at Raymond James, who covers FSLY.

Announcement and link below

… Raymond James analyst Robert Majek repeated his Outperform rating on the stock. Majek theorizes that the issue might involve TikTok switching to a higher level of data compression that could have reduced their bitrate consumption “and could provide some reassurance for investors.” He adds that he “[hasn’t] lost any confidence in Fastly’s level of performance and product differentiation that should drive sustainable share gains and growth over time.”

link: https://www.barrons.com/articles/fastly-stock-tumbles-25-her…

The (my?) takeaway: if TikTok can reduce their usage fees by improving their compression implementation, other customers likely did it, or will do it. This is where the usage model is at a disadvantage to the subscription model. So FSLY’s growth story moves (further) away from Content Delivery and more toward Edge Computing, which we all knew but maybe didn’t think Content Deliver revenue would wane so quickly, knowing FSLY’s Edge Computing play will take some time.

As such, I think my outlook on FSLY is changing. Not a complete sell like some others, but some reduction and no new money going into FSLY until the earnings call, at the very earliest.


Thanks Gary.

This new information concerns me. This is only one analyst’s opinion, but I would really like to know if there is anything supporting this (I have not been able to find anything so far). If there is any truth to it, I would find this far from ‘reassuring’(!!) but I agree with your take that this would only portend trouble for Fastly.

I would also like some more insight into how Fastly’s usage based model actually works. I won’t pretend to have much technical insight there, and so previously I have avoided tagging on my own experience of a usage based model on to Fastly, because in truth - I just don’t know how much it applies, if at all.

I work in finance for a Telecoms company, my product is essentially IOT connectivity to the company’s network for Enterprise customers. It is one of the highest growth but smallest areas of the business by far, and I pretty much have financial ownership of growing it (c.£50m revenue this year, 10 year business plan £1bn p/a - 20% YoY growth industry benchmark). We have a usage based model. Last year (when I joined the company), we underwent a migration from various legacy platforms into a single Cisco platform. The unforeseen migration impact was a severe ‘optimisation’ of customers’ data usage - because they could now get better visibility into their SIM estate and deactivate non-optimal rate plans. Great for customers, not for us.

We try and manage this through signing customers onto a minimum revenue commitment, in effect mirroring a subscription based model, while maintaining the usage upside. This protects against the downside risk that seems to be starting to show in Fastly’s model. If indeed that is the case.

Now, we are selling SIMs to IOT devices and connectivity, which runs through a single third party platform. Fastly is signing customers up to its own platform - this is fundamentally different. However, I can’t see how a customer optimising its data usage can be anything but problematic for Fastly, like it has been for us.

I was determined not to do anything with Fastly until I understood the WHY behind their usage drop. This theory would explain why TikTok dropped significantly despite being the most downloaded app in the world in August and I think it could spell a world of trouble if Fastly’s other customers’ usage had been impacted the same way in September. I’m not saying high growth with such an optimisation issue won’t happen, just that it would be much harder while your customers are churning usage simultaneously. There may of course be mitigations against this - namely pricing.

The main reason this new information/theory is so concerning to me, is that it would represent a competitive risk to Fastly. I would much prefer the macro-environment being a cause for the usage drop, as they had suggested was the case in Q2. Because this is impermanent.

As for Cloudfare’s superior subscription model, I would agree that it makes revenue much more predictable. However, would Cloudfare too be affected by competitive pricing from such optimisations? Ie - would they be under pressure to negotiate down their subscriptions.

In summary, I’m not quite sure what to make of it. I don’t want to unduly concern anybody, I am simply sharing my own recent experience in hope it might be at all relevant. I would really appreciate if someone could offer any further insight here.


This new information concerns me. This is only one analyst’s opinion, but I would really like to know if there is anything supporting this (I have not been able to find anything so far). If there is any truth to it, I would find this far from ‘reassuring’(!!) but I agree with your take that this would only portend trouble for Fastly.

I don’t know about about Tik Tok’s data compression but I do know that the industry has been compressing data for at least 35 years. Some 35 years ago my cousin told me about a data compression company he had invested in. It was a poor technology design and I told him so. About a year later we created a data compression algorithm to use in our own Mac product. You would be amazed at the efforts put into improving latency and bandwidth.

BTW, game type apps are the most demanding in terms of bandwidth to improve the game’s performance. But then a funny thing happens, as bandwidth improves they add more features to the games which then increases usage. For other apps data compression is not mission critical.

Denny Schlesinger


It is only a theory. And to assume the other customers were down in sales because they also at the same time came up with data compression technology is extrapolating that theory even further and has no basis in fact.

Fastly attributed the Tik-Tok shortfall to political. Does India ban in June of Tik-Tok count as political?

Fastly said they got a tailwind from Covid. They said Tik-Tok as well as other customers were down. I don’t need to come up with my own theories outside of that.


India ban was geopolitical, which is what FSLY press release says. The two nations have amassed their armies across the border, ready for war but negotiating an end to the current posture.
Things became heated rather quickly in May and in this context India banned 100+ China based apps for security and as retaliation.

It would have been hard for any company to foresee this but what surprises me is FSLY did not broach this - loss of revenue and risk going forward at the Q2 call. Neither did the analysts.
To be fair, maybe TikTok has multiple CDN vendors incl. FSLY.


1 Like

Also makes me think Bytedance is working on some pretty cutting edge tech. Wish I could invest in it.

This analyst’s theory is completely made up. All you need to do is look at Fastlys language

They say the reduction in revenue from main customers was “due to geopolitical “. How would finding better compression fit with this language?

I don’t buy it


This is a bit befuddling to me. I am just thinking out loud. Compression can speed up the transfer by lowering the bytes transmitted and perhaps the charges realized by Fastly. But the CDN edge is going to have to decompress/compress taking more CPU and paying an over head cost on the delivery content speed. Giving that Tik Tok has several CDN vendors, I am suspicious that this data compression is just a red herring.