“I will not emphasize on baseline theory of product management as the world is having many anecdotes of great products and the vision behind…. I am more concerned about the people who are visionary but not claiming the pie…”
Evidence: Creating a data science product end to end. In this project, I was playing the role of technical program manager for the project where we were solving the problem for SKU recommendations. In a way, it can be said that we were building a data product having my role in finding the gaps between the old solution, the data team, and the analytics team. It was the revision of the product vision where my role was to set up the way ahead for the product. Now let me deep dive and characterize this project as to how it was a project for product development capturing all the aspects of a product development paradigm.
How it landed into product management fundamentals:
The kickstart- Secondary research was conducted on how retail organizations operate, what an SKU is, and where it fits in the present solution. The information obtained regarding how they sell, service, and communicate their products and services. The Value Management Process was used to examine their business. Find out what their compelling value is and check if there are any "strange" hidden characteristics in their solutions that are reinforcing it!
Diagram by author.
Oh, I am taking you away from AI/ML context, talking product is a bigger game. Let’s resume on the product topic (I feel it’s a product made using advanced analytics) which I am talking about.
So basically, here we are talking about forecasting and predicting demand for the cigarette levels for the US market. The ML solution!! There had been already a third-party tool in place in the ML pipeline for doing the needful of training the model and subsequent prediction. The ask of the project was to replace the current concept of a third-party solution for forecasting and prediction (assumed to be a black box, customer has no idea what’s going on???) with the newly developed pipeline for forecasting using statistical and machine learning models.
If I will go with the above image flow for the definition of the product, I can say the project is showcasing a product getting developed from scratch where we were solving a problem(for forecasting and prediction of product demand in the store)--> imparting solution to remove complexity(the black box – third party tool)-->using cloud solution which will make it a scalable product-->customer favored features like the factors that affect inventory--> Accuracy, insights generation, business output.
Wow, I am justifying myself with the above flow…. Hmmm
Now let me dig into the technical aspect of the solution that we are trying and achieving. When we talk product management there are so many aspects and features to talk about like • Understanding the customer. • Segment the market. Deciding who is not our customer. (For the current project no need to bother, we are in the customer zone with a problem in hand.) • Designing Value proposition
Making the right technology and prioritization decision.
At the start I talked about the value management process, another wider topic, I will have a separate thought sharing on this in future through my lens. But for now, I will talk relevant here about the product development I had experienced while doing the forecasting project.
Diagram by author.
So understanding Value is more about • who is the customer? For a current product, it’s a retail partner and if we see it on the element of persona, primarily it will be data scientist, software developer, business analyst, TPM, and technical managers and likewise, there would be more.
• What is the underlying problem? Redeveloping a machine learning pipeline by replacing a third-party tool that is impenetrable, it’s something there in the solution pipeline but nobody knows what’s happening behind the curtain of analytics solutions.
• How can the problem have been solved? Okay here comes our reflection, as a team we would build a pipeline using moderns days data services. We would be exploring the current stack the customer has, what the disruptive solution pack can be recommended, and the development of the solution would be done from scratch. There would be a standalone ML pipeline getting developed in a manner on which human control would be an utmost priority.
Diagram by author.
• What does the customer want? Customer wants control over the ML solution, they want to have transparency on the solution and competitive advantage of doing future enhancement of their own by reducing dependency on the third party.
Let’s talk about creating value. Its mainly about the product type like if it’s a mobile app, a SaaS product, or any SDLC tagged application getting developed. It also includes the insight of product features going to be developed\implemented, assessing the team skill, and the vital technology stack needed.
And here my role was very much prominent as I was working closely with the project manager, engineering manager, senior data scientist, data architect and engineer, and most important group member business analyst (the problem chronicler). This stage involves prototyping of product (the finest stage\segment in the product journey, will have an embodiable solution with an example.) which helps in leading to MVP(Minimal Viable Product) now its in product ambiance everywhere we hear. So here it will start with a ML solution\product with a very thin pipeline which should train the data and predict the expected target label and have requisite accuracy and insight for forecasting the SKU.
Coming next is capturing value. For any product capturing values talk about the business model. Here in that context, this product will be rated from the customer’s lens like after having the newly developed solution how the SKU recommendation is making an impact on the monetization, the business revenue. It’s something out of my boundary as I am talking on the technical side of product management. But an area I will be interested in future basing the kind of product and business model I will be part of.
Communicating value is all about marketing your product. Here for the ML solution which we created, we being on the technical side of the entire solution tried to communicate the solution advantages and made a mark that compelled clients to trust the data product and got into an agreement that what value customers were looking for had been delivered and this has reached to the fulfillment of all the involved personas who were looking for a solution considering their prerequisite. This marketing is done by providing the elements like a datasheet, white papers, product demonstration (multiple iterations for training and prediction pipeline), solution briefs, getting customer testimonials.
Delivering Value, it’s the stage where the final product is available to get launched in the market. Mostly it’s about sales, delivery channels, training of product, perquisites like infrastructure. If I talk about the ML solution\product fortunately it has been developed on the customer’s infrastructure stack, so we were having an agreement before itself for delivering the product. Here it was more about giving understanding to the team for initiating the model training and prediction pipeline, how to capture the result like insight and accuracy, how to utilize the output in visualizing on BI tools.
**Okay, it seems this story telling will have a series to follow. I will stop here and will work on the next chapter for bringing more proof that will help me to demonstrate that being a delivery manager in data science and analytics I actually have a garb of technical product management….. Also, this product development story didn’t reach its conclusion… next will be more on unveiling the journey encompassing data science theory, ML conceptualization like model selection, EDA, data pipeline ETL , autoML procedures, and MLops….
This is just a beginning of Technical\Data product management tale. lots to share with on-the-job learning, capstones, and assignments. I will keep on ringing the bell…..**
I DON’T KNOW WHO YOU ARE, OR WHAT YOU WANT… BUT WE’LL HAVE A WORKING PROTOTYPE NEXT SPRINT!!