Gahhh… Machine Learning, Big Data, AI, all these buzzwords!!. What does it mean to me as a Product Manager?
OMG, your team is working on Machine Learning. That’s so cool, have you thought of using it for so and so….
The data scientists don’t want to work the product because it’s not an exciting ML challenge? 🙁
Are you also face palming about all these questions? Here is what I’ve learnt over the past 2 months
Lesson 1 — ML does not change your role as a PM. ML is a tool, apply it to a real-life problem
Machine Learning does not change your role as a Product Manager and our role remains to talk to customers and communicating their problems to the technical team and business. Remember, at the end of the day, it’s the experience that will be driving the product success.
I like to think about ML as a tool for solving a problem and it can be applied in many ways. One question I like to ask myself before suggesting ML is
‘Do I want to perform this task over and over again, it absolutely requires no thinking?”
If the answer is no, then this is a great place to start thinking about ML. However, a lot of these problems can be solved with a simple rule-based program. My word of advice would be to work with your data scientists and engineers, as a team you will be able to work out the most feasible option. Don’t build ML just for the sake of it as it is a large commitment and investment from the business. Remember, you can still build great product experience without it.
Lesson 2 — Managing business expectations is crucial
I can’t emphasize how important it is to manage the business expectations when it comes to ML. Since there is so much hype within the industry, it is easy for the business stakeholders to be in the mindset that ML can solve all their problems and they will get a return on their investment.
This is not necessarily the case. Yes, a promising model/solution can be generated pretty quickly, however as a team, you won’t be certain how good it will be until you get customers using it. Along with that, it takes time for the engineers to build a scalable infrastructure, data scientists time to fine-tune the model and PM’s time to get the right experience.
Communicate to the business this is a large investment and it can fail. ML isn’t magic and its a customer delight feature. If you take the time to do it right, it can be game changing!
Lesson 3 — Success is maximized by quick iterations. Feedback loop is key for both your team and customers
ML is such an interesting space to be working in. You often have the dilemma of this is a great experience, but often run into the question of
‘Is this even feasible with ML, if not, how long will the research take’.
‘This is feasible with ML but the team says it’s not an exciting data scientist challenge for them to work on?”
This is when you need to involve the engineers and data scientists to listen in on customer sketching sessions. Let them listen to the customer problems so they will develop empathy and be more engaged to work with your on feasibility and the product.
ML is so personal and relies on both getting the experience and information correctly presented. To avoid the ‘Target’ incident, I recommend testing with customers on a weekly basis, run experiments, test hypothesis so you can learn quickly. Remember, the quicker you learn, the less risk you will be taking when launching to the wider customer base.
Also don’t forget though, building code is expensive, start off with a mock flow, once you are comfortable, define a basic MVP with your team and run beta programs until you are comfortable with the product
These are my thoughts on building ML products after 2 months into the job. I’d love to hear your learnings and experiences 🙂
For now, peace out!