Ashley Fidler (@aefidler) header
Ashley Fidler (@aefidler) profile picture

Enterprise products & AI platforms. I just lurk now.

Boulder, CO
Joined July 2008
Want a product job where you talk to and think about users every day? Where amazing user experience is your top concern? If so, DM me. We're hiring :) JD: purepm.co/join/roles/sr-prod…
2
In your product resume, don't forget to mention what your products do. You built a product strategy at Company X? Great! What the heck does that company do? I'm noticing that at least half the resumes coming though miss this key info.
5
Back in the US building a new product and a new product-focused company. Having a blast so far. Looking forward to sharing more as we move forward.
1
8
As my time at @Eigen_Tech officially comes to a close, I am so grateful to my amazing team. Not only are they consumate professionals who do their work with incredible care, they also planned such a touching send-off - including a poem! I will miss you all❤️
2
9
Time for an end-of-year social media break. Wishing you all a restful and enjoyable holiday. See you in the New Year!
6
One more ML Q&A before the holidays: How does prod mgt differ between traditional SaaS and ML/AI products? 1. Platform + portfolio, rather than product 2. Services and margins 3. Non-determinism and testing 4. The R&D process 5. Technical + market feasibility in parallel
1
2
14
4/ Ongoing R&D is critical for ML products - people expect them to get better over time. Build flexibly from day one so that new science can be easily tested in the product. Also hire a science PM to facilitate the R&D into production process.
1
1
5/ AI products often entail simultaneously addressing technical feasibility and market fit. This is hard. Make sure you have a bread-and-butter problem that drives revenue with well-understood tech so that you can make these investments. Otherwise be prepared for pain.
Enterprise Product Q&A: What does "good" look like for a machine learning model? Thinking about this question often brings me back to natural language user research I was involved in at Microsoft....
1
1
2
We did several small studies to figure out how people engage with voice systems. When accuracy was about 75%, people were annoyed. The results seemed like crap. At 80-85%, people would use it, but didn't really find it trustworthy. Above 90% started to feel natural. 2/3
1
Of course, "good" depends a lot on the problem and the business need. But watching people get frustrated with systems at the 75% accuracy mark gave me a visceral reminder of what ML accuracy means to product quality. 3/3
Enterprise Product Q&A 3: How do you avoid building what the customer "wants" and then it’s still the wrong thing? (I'm assuming a large-ish feature for an enterprise customer that has been determined worth doing)
1
Keep in touch with the customer during the build process, but watch out for scope creep. Plan a second milestone, if it makes sense, but don't just keep cramming things in. Instead, get them excited about the benefits and prep them for a smooth roll-out. 8/9
1
1
In short: making sure you build the right thing for a customer is all about understanding the problem, setting clear expectations, and managing them carefully. This requires persistence. But also a business-focus and a sales mentality. 9/9