Hi, I'm Katie! One of the main things I use this account for is writing threads about data science management. This Thread of Threads™ is a collection of some of my favorites.

Big things are happening in the world of Gemini for Looker

Don't lose heart, and don't sell yourself short!.

Jevons paradox is becoming a little bit of a cliche to describe the second order effects of genAI, but IMO it's at play here. Everyone works with data now, and folks who know what they're doing are more valuable because of it.

There's a lot you can read into this, but since I know this sort of trend can spook people, it's worth pointing out that if someone is positioning their services or product against you, that means what you do is probably in pretty high demand.

I have been seeing an increasing number of posts from consultants and fractional data leaders recently suggesting that companies don't need to hire a full time leader, and that their services should be retained instead.

Session link:

If you're interested in some real talk about AI and data work, you should check out the "Lessons From the Data Trenches" panel at Snowflake BUILD on 11/14. We'll be talking about how it actually impacts the ways we're building and operating.

It's 2024. LinkedIn overflows with thought leaders proclaiming AI will get better & cheaper if we just wait 6 months. Data quality is still a mess, despite insistence that AI will finally justify investments in it. Sparkle are stamped on every UI, ruining a perfectly good emoji.

If you see the introduction of data slowing down conversations more often than pushing them forward, it's time to start teasing opinions about the data actually means. People have more of them than they realize.

The cost of using data is lower than ever, but low cost is not the same thing as value. As a data professional, your responsibility is to make sure your company is using data productively.

I'm open to the former, and I'm a definitely believer in the latter two. They represent clear opinions about what otherwise-abstract data represents, and they provide structure that can help folks build the requisite skills to translate their thoughts into something quantitative.

IMO, the practical consequence of this is that the industry is now searching for the correct information compression techniques and technologies. Some of the trendiest topics right now--generative BI, metrics trees, weekly business reviews--seem to fit into that category.

If you think about data work as fundamentally being about shaping and curating the information environment within a company, at least as it pertains to business metrics, it seems like we may be approaching the reaching a broad audience phase.

Any form of information dissemination that reaches a broad enough audience will eventually need a technique or technology usefully compress that information--newspapers have headlines, social media sites have algorithmic feeds, and so forth.

That designer *did* actually have a lot of thoughts about how to improve the onboarding flow, but struggled to translate them into something quantifiable or measurable. They got there with help, but if no one had pushed back on their request, they'd have drowned themself in info.

But if you want to assume your colleagues are at least somewhat competent, there's a better way to think about interactions like this. Thinking quantitatively is a skill that needs to be developed and practiced.

If you want to be especially harsh, you could say that it's "a convenient lie--an elaborate cover-up for poor prioritization or inadequate resourcing"

There are many negative ways to frame this interaction: analysis paralysis, data-driven procrastination, fear of being wrong.

The compromise we reached was to come up with some specific questions to help them better understand the things they were worried about. It required back and forth to formulate those questions, but it was quicker and easier to address than listless fishing in a datalake.

But the designer insisted we needed to see the whole funnel before we could start, even if it meant it would delay the project kickoff and that we'd have less time to actually work toward our goals. It didn't matter if the data we *did* have wasn't relevant, we just needed it!.