Does this breakdown of HiSeq labs sound correct?

[potomac]

This is a continuation a previous thread I made: http://seqanswers.com/forums/showthread.php?t=89484

I have been trying to better understand how many Novaseq's Illumina will sell in the coming years. It's a difficult question to answer because a lot of academic labs seem to be sticking with their HiSeq's for the time being.

Illumina has said that at the time of the Novaseq launch in 2017, there were 800 labs with HiSeqs 2000 through 4000 machines.

To further break down that 800 lab figure, I made a 2 x 2 matrix below with the 2 variables I think most influence labs to switch from HiSeq to Novaseq. This is based on a parsing of comments in the other thread.

Factor 1: Having no control vs. partial/full control over library creation

Factor 2: primarily using small "Rapid Run" mode runs with two lanes on a flowcell vs. "High Output" runs with all 8 lanes

I was hoping you guys could tell me if the chart below seems realistic. Does the breakdowns sound like they're in the right neighborhood? (15 / 5% / 20% / 60%)

Is there anything else that doesn't seem to reflect reality based on what you've seen in academic core labs etc?




p.s. As mentioned in the other thread I've been working on a personal project to better understand financial modelling and I thought Illumina was a great example of a company I know a little bit about. Would be happy to share my end result with folks here.

[potomac]

Here's how I arrived at the 10% / 5% / 20% / 65% breakdown. It's really a shot in the dark. I make some fairly arbitrary assumptions in certain places (highlighted in red).

-I started by looking at some official Illumina figures.

-Since 2017, about 40% of all their instrument and consumable shipments went to academic labs. This could tell you that 40% of all HiSeq instruments are in academic labs (320 of 800). In the past though (2011), the academic shipment figure was as high as 80%.

-Then I focused on the top two boxes in the 2 x 2 matrix

-Further assume that 50% of those 320 labs have no control over library creation because samples are mailed in by customers. This gives you a total of 160 of 800 labs. This is a super arbitrary percentage I picked, need help here.
-Further assume that 75% of those 160 labs primarily use "Rapid Run" mode. Again, super arbitrary percentage.
-This gives you the 15% (120 of 800) and and 5% (40 of 800) pictured for the top two boxes of the 2 x 2 chart.

-Then I focused on the bottom two boxes in the 2 x 2 matrix

-I assumed that 75% of the remaining labs use "High Output" mode since they have control over library creation.
-This gives you the 20% (160 of 800) and and 60% (4880 of 800) pictured for the bottom two boxes of the 2 x 2 chart.

[SNPsaurus]

I think most academic cores run the cheaper high output with 8 lanes rather than the fast and expensive rapid run mode and make users wait for a run of the appropriate type to fill up. Academic users are price-conscious more than time-prioritizing. I just looked up "Illumina core" and got these snippets:
1)For HiSeq 4000 runs, a flowcell can be sequenced once there are 8 samples ready.
2)you will be able to see which kinds of samples are already in the queue to be run, and how many samples of each kind are needed to complete a set of 8 that is required to fill and run a flow cell.
3) HS4000 flow cells require 8 lanes to sequence. Submitting 8 samples will decrease the wait time.

There are a few services out there that match facilities waiting for a few lanes to start a run with users needing a few lanes to be sequenced. This wouldn't be possible if only a few places ran high output flow cells with a mixture of users in each.

Quite a few facilities have a HiSeq4000 plus a NextSeq (or a HiSeq2500 kept for Rapid Runs) and a factor driving them to a NovaSeq is being able to have small and large runs on the same platform.

That said, I have no idea how to estimate things any better! Other thoughts--40% of shipments going to academic labs might not translate to 40% of the HiSeq market. Lots of individual academic labs have a MiSeq/MiniSeq. You could look at ABRF meeting attendance (maybe?) for a sense of how many academic cores run a HiSeq. Or google away. That set is maybe the easiest to put a firm number on, since a core facility has to have a public-facing web page. Is there a Tier 1 research university without a HiSeq? I tried a half dozen lower tier universities and they all had one.