An attempt to obfuscate and confuse?

An attempt to obfuscate and confuse?

A dollar-based net retention rate is normally very clear. You take the spend from ALL your customers from a year ago, and the total spend from THOSE SAME customers this year, and you divide the second by the first to see how much it has grown. That’s very clear, isn’t it?

If a customer dropped out (churned), and is no longer here this year you count its revenue from last year anyway, because you are looking at what last year’s entire cohort did and how its spend grew, or didn’t. That’s standard. No question about it!

Okay, here’s what Fastly did. They have four different measures, and none of them the correct one as far as I can see. They do specify that each of these is excluding their acquisition

Dollar-Based Net Expansion Rate (DBNER) of 143%…
Net Retention Rate (NRR) of 115%…
Last-twelve-month (LTM) Net Retention Rate of 137%…
2020 Annual Revenue Retention Rate of 99%…

Each of them had a long complicated footnote to explain how they got it.

For the 143% value it was “We calculate Dollar-Based Net Expansion Rate by dividing the revenue for a given period from customers who remained customers as of the last day of the given period (the “current” period) by the revenue from the same customers for the same period measured one year prior (the “base” period)… .”

So this was clearly cheating. They didn’t count any customers who dropped out, only the ones who stayed.

For the 115% value it was “Net Retention Rate measures the net change in monthly revenue from existing customers in the last month of the period (the “current" period month) compared to the last month of the same period one year prior (the “prior” period month). The revenue included in the current period month includes revenue from revenue contraction due to billing decreases or customer churn, (ii) revenue expansion due to billing increases, but excludes revenue from new customers…

As far as I can tell this is still cheating. They still only considered customers who are still there.

For the 137% value it was “Our LTM Net Retention Rate, intended to be supplemental to our Net Retention Rate, was 137% for the period ended December 31, 2020. We calculate LTM Net Retention Rate by dividing the total customer revenue for the prior twelve-month period (“prior 12-month period”) ending at the beginning of the last twelve-month period (“LTM period”) minus revenue contraction due to billing decreases or customer churn, plus revenue expansion due to billing increases during the LTM period from the same customers by the total prior 12-month period revenue. We believe the LTM Net Retention Rate is supplemental as it removes some of the volatility that is inherent in a usage-based business model.”

I can’t tell exactly what they did, except that they were embarrassed by the 115% puny value this quarter, so decided to average it over 12 months, to pick up that big 2nd quarter.

For the 99% value it was Annual revenue retention rate is calculated by multiplying the final full month of revenue from a customer that terminated its contract with us (a Churned Customer) by the number of months remaining in the same calendar year (Annual Revenue Churn). The quotient of the Annual Revenue Churn from all of our Churned Customers divided by our annual revenue of the same calendar year is then subtracted from 100% to determine our annual revenue retention rate.

So this is another metric that they just invented to finally mention their churned customers, but it’s cheating again, because if a customer providing 48% of their revenue totally left them with one month to go, they would count that as 4% churn instead of 48% churn.

Other horrible statistics

The previous two quarters they added about 100 customers each quarter (already tiny compared to Cloudflare’s 10,000 new paying customers), but this quarter Fastly only added 37 new customers. Total customer count increased to 2,084 from 2,047. That’s just a 1.8% increase. And enterprise customers ($100,000 or more) increased by all of 11).

While the above were excluding their acquisition their revenue was “including” the acquisition:

“Strong top-line growth of 40% year-over-year with revenue of nearly $83 million”

They are comparing that $83 million with the $59 million they had a year ago, so they are adding the acquisition revenue into this year’s revenue but not into last year’s revenue. That’s outright cheating again, and they don’t even tell you how much of this year’s revenue was contributed by the acquisition.

I am so glad I exited Fastly.



Hey Saul !

Excellent review of Fastly.
Question: Does anyone (even you) ever confront the company(s) when they do this sort of thing?

I think they should be “called to the carpet” when doing this. I don’t do those kind of number analysis so I can’t talk to them intelligently about it.

Rich - sold all Fastly over 6 mo. ago

1 Like

Question: Does anyone (even you) ever confront the company(s) when they do this sort of thing?

Hi Haywool, I didn’t listen to the conference call as I am no longer in them, but muji states that someone (an analyst) had to drag out the amount of revenue ($6 million) in that $83 million that came from the acquisition, and that real revenue growth was 30%, not 40%.


Someone kindly emailed me to clarify all those retention and expansion rates:

It seems that the Dollar-Based Net Expansion Rate was correctly defined, and does exclude churn as defined, and is only confusing in the sense that most companies don’t use it but use the Dollar-Based Net Retention Rate (which they apparently did use correctly although that was hard to figure out from the way they wrote it).

They seem to have decided to add the Expansion Rate in the last two quarters as the normal and standard Retention Rate (of 115% this quarter) was so low as to be embarrassing, so they added the Expansion Rate (excluding from the calculation those companies that churned and dropped out), apparently so that they’d have a larger number to exhibit, and which they placed first, so that it would be the headline figure.

They also added the Last Twelve Months Retention Rate because the last quarter’s Retention was so low that they apparently wanted to soften the shock by diluting it in the entire year retention rate.

So they posted four separate metrics, and I can’t help suspecting that they did that so that the 115% Retention Rate might get lost to people’s attention.

Sorry if I caused any confusion