Data Demystified: Product (Not)
What is a data product? There are no shortage of answers coming from the data management community, and most of them are shite.
When you consider attributes of successful products in general, you'll quickly see that what passes for product in the data space rarely conforms.
Let's start with an example of a product that I've attached myself to recently - Gemini on Android.
In preparation for writing this article, I vocalized the following question: "What are the common characteristics of a product?".
A simple question that produced a concise list of physical and non-physical characteristics that made perfect sense and easily to any would-be data product.
However, the listed attributes didn't satisfy.
Fortunately, Gemini also provided follow up suggestions to the original question, such as:
- "What other product characteristics contribute to its success?"
- "Give examples of strong and weak value propositions."
- "How do intangible and tangible product attributes differ?"
Finally, Gemini provided several links to related content, the top choice being
The 5 Qualities That Make a Product Great (And How to Apply Them)
This is a community post from May 25, 2020 by Jorge Rodriguez-Ramos on www.mindtheproduct.com, and it's a great summarization that makes perfect sense as criteria for judging if a data deliverable is a worthy of the product label.
That is, if you truly want to transform data into valuable products. Otherwise, as you were.
Jorge presents these 5 qualities as a pyramid where each succeeding quality builds on those preceding. They are: Usefulness, Explainability, Simplicity, Scalability, Habit Formation.
I'm not going to dig any deeper into Jorge's work. It stands on its own and needs no help from me.
Rather, I'll map my newfound fondness for Gemini to these qualities - as well as apply them to data team efforts I've seen that were called products.
Usefulness. Over the past two months, Gemini has been crucial in helping me navigate not only my previous paid gig, but also in research for my writing. Compared to Copilot which gave me wrong answer after wrong answer about Snowflake syntax or FiveTran quirks, Gemini solved problems in minutes that might have taken me hours or days of searching/researching using other avenues - including the vendor's product documentation.
Now consider the dashboard built over six months for an audience of one, only to be told that it's too confusing to use. Or worse, completely ignored with zero clicks within 30 days of deployment.
Explainability. Well, I touch the Gemini icon on my Google search bar, then touch the microphone icon and ask a question as concisely as I know how. Then a summary answer, plus follow up questions as hyperlinks and top-ranked online content related to my question. No one had to explain shit.
Contrast with PMO SDLC templated documents given to users from an IT-led data team that are perfect for proving that you understood requirements, data sources and anything else needed to cover exposed tushies. Everything, that is, except for how the heck the product user is supposed to get the requested value from what's been deployed.
Simplicity. Point a finger, touch an icon, and ask a question in your own words. Then read. Yes, maybe you have to refine the question a time or two, but that's life with AI. Get used to it.
As opposed to: submit a request, wait hours or days for that request to be acknowledged (hopefully), attend meeting after meeting to explain what you want, why you want it, and if you have "business justification" (or the political juice) for the request to proceed. Then wait 3-6 months for someone to bring you something that most likely won't be very close to what you asked for, or poorly tested and filled with errors you identify in 15 minutes or less.
Scalability. Is your product built to cost less than the financial benefit received? Will it remain so if demand for increases exponentially? I think the jury is still out on all the AI platforms, but I have to assume that I'm one of millions of people asking Gemini questions and it responds in seconds with useful information. Hopefully, Google isn't losing money.
Compare that to the PowerBI dashboard slapped together with 17 different table definitions from seven separate sources (including uncontrolled Excel), 50 slicers, and 15 tabs - using direct querying that takes a minimum of FIVE MINUTES to pull up a product catalog. Did it pull more quickly after the first request of the day? How much compute cost per user request?
No one knows because no one measured.
Habit formation. Has this product built enough of a foundation to make it seem indispensable? Or at least significantly inconvenient to live without?
I ask Gemini questions all hours of the day as a vehicle for my writing and a facilitator for my learning. I would be greatly inconvenienced if it went away, at least for a time.
I've been a part of delivering some data products that met this criteria. However, even those products were plagued by mistrust when others could come up with different results devoid of context. Moreover, I seldom knew how important some of those products were except when they were broken or called into question by another group using different definitions.
Like much of what I've called out in this series on demystifying data, product discipline is largely lacking when it comes to many organizations' attempts to realize true value from their data investments.
And much of that failure has to do with being shackled to an IT- or PMO-centric mindset.
Imagine what could happen if we started evaluating data products the way we ALL evaluate the products and brands we choose from in the consumer marketplace?