How To Evaluate Hot Data Trends
How to evaluate hot new data trends, CDPs Composable CDPs are becoming a thing, Why data PMs need to pick up these skills.
This is the Three Data Point Thursday, making your business smarter with data & AI.
Want to share anything with me? Hit me up on Twitter @sbalnojan or Linkedin.
Let’s dive in!
How to evaluate hot new data trends
Here are three lessons we’ve learned the hard way to evaluate data trends: the data mesh, data observability, or what else the data geeks are cooking up today.
👍Rule of thumb: Ignore half of them. For the other half, build quick PoCs; if it doesn’t work for you, it likely isn’t for you.
Here’s why.
Lesson 1: Data always flows.
It flows through data producers, through the engines transforming data, to consumers of data.
As such, no data trend is ever only “technical.” All trends are about 1. technology, 2—people & processes inside your company.
Lesson 2: Different trends affect vastly different numbers of organizational units.
Shifting from “service data teams” to “product data teams” might affect one team inside your company.
Installing analytics engineers will involve multiple departments, and the data mesh hits the whole company.
Lesson 3: Your company processes are a unique snowflake.
As such, for trends that involve processes and people (and that are almost all!), looking to other companies has little to no relevance to your situation.
So what? So, how do you then deal with data trends?
Never ignore a trend because it affects multiple organizational units; all do.
Rate each trend based on the number of affected units + where they sit on the tech vs. processes spectrum.
Then simply try them out if you feel your company is ready!
CDPs Composable CDPs are becoming a thing
Composable Customer Data Platforms (CDPs) are becoming a real thing.
And unlike many other hot data trends, this one has practical value to the business.
So, what is the CDP? It’s the one data tool that the marketing and sales department actually wants. For more details, we suggest you read the data.beats series on it.
So why did we have to add another word in front of it? Why composable?
Because these handy tools are so helpful because they do many things in a row, unlike a, say, dbt.
That makes using many puzzle pieces possible and building your own composable CDP.
So what? This just means everyone can now have a CDP, composable or not.
Why data PMs need to pick up these skills
Take a pond with one lilly inside. The lilly doubles every day. On day 30, the full pond is covered with lilies.
So what? How many days does it take to cover half of the pond with lilies?
The most common answer is 15, yet the correct one is 29.
If you didn’t answer correctly, you should read the text below; it’s the same principle, and your instincts on exponentials are wrong (as are ours).
The world is moving fast. In 5 years, 90% of all data will be real-time available.
That is in the form of "events" (A ordered X), not state (A’s orders are X, Y, Z).
Better adapt now to real-time to leverage this data next year than to be a late mover.
Second, 99% of all data growth happens outside of your company. You do not collect it; you never see this data.
And yet, this data is relevant to your competitors, customers, and markets. Learn to use these sources, buy data, collect/buy alternative data, and get it into your system.
Third, If you think have unstructured data, you probably mean the data you're not using.
Get used to utilizing all your data, not just a tiny bit. Otherwise, that pile of unused data will be vast and unmanageable in the future.
Key lesson: It’s the same as with the lilies: if data is growing exponentially around you, you won’t see it until it is too late until it is day 29.