Models

Welcome to the second edition, and hello to all the new subscribers. In this edition I summarise what I’m hearing out in the market, at events and behind closed doors.

I wanted to riff on a couple of things, The first is models. Certainly in media circles, you can now talk about predictive models, and people know what you mean. Or at least don’t freeze up in conversation. AI has been that blessing. It wasn’t that long ago that people weren’t interested or didn’t know what it might mean.

The New Yorker had a thoughtful piece on, how much of the world is it possible to model? Case in point, it’s a New Yorker topic, models are in culture. But it is a good piece, so do read it.

In other dangerous news, I am now the proud, ‘local hoster’ of an LLM. Took sub 15 minutes to get set up, I used LMStudio but Ollama is also easy. This has basic steps get set up. What this is, is a running your own GPT on your desktop.

But so, why? And I guess why not. You can experiment, play, all in the privacy and safety of your own device. Are they as good as GPT4? No, but there are literally hundreds of models available. And in time, yes they will be. But the real reason, a local sandbox helps remove any restrictions to just play and experiment.

To activate AI, everyone needs their data & analytics, tidy, clean, up to date and in a good place to build models off. Which has then lead to firms investing in their teams, expediting projects. And then that of course has a compound effect, as you need a bigger team, to manage the current team.

Ben Rogojan (aka Seattle Data Guy) had a riff on analytics teams in these projects, ensuring they remain strategically focused. And he touches on the infamous road to nowhere projects, getting all the infrastructure and data stack up and running. But losing sight of the goal. A recent anecdote I had from a friend was how, someone walked into a role where the data was all set up, then they could use GPT to help them derive insights. So, it is easy to poke fun at but it is necessary to get the data foundation. But do people make it too complicated, yes they do.

At the moment data & analytics is soup du jour. And that’s a good thing! What it is bringing back to the top though, is you need to have a well functioning team that works with other departments. Eric Seufert had a good post on X about this, how the best teams, run as a unit with political air cover from their Head of Data.

Right, so even the hot new thing, does fall to the same limitations of bureaucracy. Not surprising!

All this change opens up a new kind of role, well not necessarily new, but a lot of these efforts are ‘internal analytics’ whereby they need more analytics, for the things happening internally. Not that this is new, but now it’s more in demand. Rather than sitting in marketing, looking at outbound efforts in the market, or with an agency, architecting the measurement. Companies are needing help getting a grapple on insights from their own data, to help build that platform.

A weird by product of this ramp in roles though, is a disconnect on job titles, Head of Data, Product Management, Analyst, Data & Analytics. Which, at least in my discussions making it challenging for people hiring and folks applying, to find the right roles. Leading both to focus on skills in descriptions rather than titles.

Notable stories this week

• Fox renews multiyear measurement deal with Nielsen.

• NBCUniversal focuses on their one platform for audience measurement.

• Dentsu releases their attention economy study on gaming.

• Cookies will finally die in 2024, here’s why advertisers should celebrate the upgrade.

• How much of the world is it possible to model? A thoughtful piece in our moment of modeling everything.

Deals/M&A

• Liveramp acquires data collaboration company Habu.

• Houston energy analytics startup raises $41m.

• Consolidated Analytics acquires Real Info.

Data visualization of the week

• The popularity of the distracted boyfriend meme correlates with the number of statisticians in New Jersey.

Smartest commentary

• “As data professionals we like to joke about quick asks and how silly the "can you just pull this real quick" request can be sometimes. But the more serious side of this is that data teams can become help-desks more than a strategic partner. I have seen teams that have good intentions get bogged down by these constant requests and in turn fail to ever deliver strategic value.” -Ben Rogojan.

Datapoints of note

• Demand for data and analytics expertise at three year high.

• 86% of respondents reported a notable increase in the demand for data and analytics expertise.

• Bot clicks and fake traffic to cost advertisers over $71b in 2024.

That’s it,

-Ben

..

See our most recent on solving analytics discrepancies.

The Analytics Note Resources [continously updated] ← please do send any you recommend, books, podcasts, websites, even other newsletters.