Seeing these "Copilot in Fabric" figures in a recent Microsoft blog from Ignite made me spit out my coffee. Completely absurd to just toss out numbers like this and back it up with nothing more than hot air... #PowerBI #MicrosoftFabric #Copilot - ThreadSky

data-goblins.com • 103 days ago

Seeing these "Copilot in Fabric" figures in a recent Microsoft blog from Ignite made me spit out my coffee.

Completely absurd to just toss out numbers like this and back it up with nothing more than hot air...

#PowerBI #MicrosoftFabric #Copilot

Comments

mikeainoz.bsky.social•102 days ago

Kurt, I would imagine an AI wrote that review, so it was sucking its own exhaust🤣

lonskov.bsky.social•103 days ago

I fail to see the issue; I thought that in Fabric, one Fabricates 🫣

(Think I need to start thinking of different jokes; I am getting quite one dimensional)

wynhopkins.bsky.social•102 days ago

36% more accurately? What does that even mean?

wynhopkins.bsky.social•102 days ago

Does it mean 64% were less accurate, and by that we really mean wrong?

sqlgene.com•102 days ago

I could imagine a situation where they moved higher up on Blooms Taxonomy.

I used to go to a marketing college class and the students would would analyze my GA. There's insights weren't *wrong*, just shallow and descriptive.

merrykerry.bsky.social•102 days ago

You uncovered an insight or you didn't. It doesn't make sense.

sqlgene.com•102 days ago

I acknowledge I'm being extremely and unreasonably charitable here:
https://bsky.app/profile/sqlgene.com/post/3lbhoohcd3k2p

But when students would tell me: "You get a lot of blog views from India, so maybe your should cross-translate your blog into an Indian language", I would consider that an inaccurate insight.

merrykerry.bsky.social•102 days ago

Just to understand you a little better...
That is a matter of interpretation? So rather than "uncovering imsights more accurately", "interpreting insights more accurately"?

sqlgene.com•102 days ago

Yes, I think that's a good phrasing of what I'm getting at. And unless they had a clear and well-defined scoring rubric it's still BS / up to interpretation. In practice, I suspect the source study is rather shallow and the numbers are being abused in the blog post. 1/2

data-goblins.com•102 days ago

I would wager that the author of this blog and even the person who orchestrated the "study" couldn't even say (or don't know)

koenv.bsky.social•102 days ago

The author didn't know that 87% of bugs are found only after 1.2 weeks ;)

r1k91.bsky.social•102 days ago

We could print random data in a card without paying for an F64

kendralittle.bsky.social•102 days ago

This reads to me like a Product Manager using telemetry data and an in app survey, but working hard to make it sound sciency

The same size is so small that it implies either nobody is using that or they very selectively close the dataset tho 🤡

evogelpohl.bsky.social•98 days ago

86.78% of all statistics are made up on the spot.

data-goblins.com•103 days ago

1) "Four common metrics associated with the consumption experience in Power BI" -- yet the author extrapolates this to the entirety of "Copilot in Fabric" in the body text? And just consumption experiences, not developer experiences? Uh, what?

data-goblins.com•103 days ago

2) n=209 user studies -- so 209 users? Or user _studies_? Of who? Customers? MVPs? What was their prior experience with Power BI? This is not only a tiny sample size but also meaningless without the context of _who_ the test was done on?

data-goblins.com•103 days ago

3) _What was actually tested?_ These figures about "faster data analysis tasks" and "accurate insights" are completely meaningless fluff. If someone is to take this even remotely seriously as evidence of Copilot being useful then this context is essential.

data-goblins.com•103 days ago

4) Qualitative sentiment about "likely to adopt Copilot in Fabric" is equally meaningless. Are you simply asking people "would you use this"? Is "maybe" a yes? How is that adoption, which is whether someone is actually effectively _using_ the tool to do more and improve existing tasks?

data-goblins.com•103 days ago

Honestly, this is a good example of how the environment around Copilot and other generative AI tools is dripping with dishonest, disingenuous greed. Just be honest - this shit is being publicly beta-tested; there's some value in certain circumstances, but stop selling the damn snake oil.

purplefrogalex.bsky.social•103 days ago

Gotta feed those marketing goals somehow eh?!

data-goblins.com•103 days ago

Pump those numbers up

sqlclause.no•103 days ago

Scott Adams is a muppet, but he has had some pretty good Dlberrt strips

data-goblins.com•103 days ago

Classic. Some others that a colleague recently sent me:

sqlclause.no•97 days ago

Yesterday this Dilbert story was re-enacted…

Comments

Posting Rules

Reply