How’s everyone enjoying #DOES21?? We’re having the best time connecting with you all!! Here’s what we got going on today at Split today. We’d love to meet you if we haven’t already! 🎮 Stop by Our Booth We’re giving away an Oculus Quest 2 📣 Wednesday 5/19 at 12:50pm-1:20pm BST A Layered Approach to Progressive Delivery with Dave Karow on Track 4 🪄Wednesday 5/19 @ 6:25pm BST/Happy Hour Happy Hour Gone Mental with Magician/Mentalist, Tom Pesce!

That was an awesome talk!

Thanks @abhibansal60

I’ve moved Q&A from the talk to here… I’ll hang out for a bit!

My deck will be up in DOES library soon, but it’s also here: https://speakerdeck.com/davekarow/does-layered-approach-to-pd-2021

how does http://split.io compare to LaunchDarkly? Do you have suggestions on evaluating the offerings?

@logankd Both platforms handle decouple deploy from release pretty well. LaunchDarkly has tinkered with the automated stats analysis side of things but isn’t really all in on that.

I would suggest reaching out to Split and asking for a demo or POC. Our team’s approach to evaluation is deep and disciplined. You’ll like it.

Thanks! I was impressed by the stats analysis of your presentation. my current client is far from it, we're just trying to get feature flags and less branches to take hold (they still do on premise for a lot of reasons). I'd like feature flags to decouple deploy from release to be a first good step.

@logankd There is also a very different approach to PII (personally identifiable information) between the two. https://twitter.com/davekarow/status/1394790632640507904?s=20

https://try.split.io/hubfs/pdfs/security-primer/split-security-primer.pdf

@logankd One reason I wrote the layered approach presentation was because the benefits to dev’s of the automated guardrails and measuring impact are huge. Using stats can freak folks out who think they need multiple versions and will be doing a/b testing of everything on their backlog.

Instead, you are using the stats engine as an automated vigilance so you can push daily/hourly… A/B in that case isn’t two versions of a code change but before and after code change… we just launched a huge feature and will keep it at 50/50 for a few days to compare folks who get it to those who dont.

having a/b testing not being extra work, but comparing before and after was eye opening

a good story in the DevOps handbook was about the Facebook messenger, the deployed it and had it generate traffic in the background, then learned and made it visible later 🙂

Having the guardrails is quality of life from a stress perspective. Having metric impact to see the impact of your work is quality of life from a job satisfaction/making a difference perspective.

I like that, adding this capability would seem overloading, but with automation it seems like it can improve a lot of things

The FB messenger example is indeed the origin of “dark launch” and that they had end-users generate all the traffic was eye-opening to me at the time as a load testing guy.

@logankd Many teams have plenty of telemetry already, so going this route doesn’t mean new instrumentation, just a smarter lens to look at it.

There's still time to join our Happy Hour with Magician Tom Pesce: About to wrap up Day 2 with a bit of magic!🪄 Split is hosting Happy Hour Gone Mental with Mentalist, Tom Pesce, who will blow your mind 🤯 Join us https://split.zoom.us/j/83701220960?pwd=TlBmNlpqaUMrOFFoSlN2TFlkcUxTZz09.

Join in the fun!!!