Leadership Diary — Focus

A series of blogs on learning and practicing leadership.

Principle 2

• Focus: Fewer Things Done Better.

Corollary

• Done better means get things DONE DONE DONE. Dev complete is not done, alpha release is not done. DONE DONE DONE means your project is delivered in the GA shape and it meets all the success criteria you contracted with your customers.

I was debating if I should continue my series of principles on the “How” part, the culture and value, or pick the next one based on what matters more to me. I decided to prioritize the latter, since I’m really passionate about discussing this topic here, that is Focus. If there is one thing that that transformed the “What” part of my management the most, it is Focus— Fewer Things Done Better.

My story

Ironically, our group lead has been repeatedly and relentlessly telling us to focus — fewer things done better. He has a story: he used to lead a team of 300 engineers but they only track 3 KPIs, why would you guys have 10 OKRs to track for a team of 10? However, I honestly didn’t give a shit about it till recently.

Back to previous story of Candid Feedback. Out of the quarterly review with org leadership. The feedback was brutal and direct: “You have a nice punch list of 10 items done, but what do you really want to do?” “Who is your target customer?”

Initially, I felt innocent. What am I supposed to do? I have all these asks from customers, I just have to get them done, what is wrong? As an infrastructure team, my customers are the AI developers. What is the problem of doing 20 different things for them? After calming down a bit, I sit down and reflected on the past half a year, what are the things our team has really DONE and we are proud of?

• We contributed a patch to TensorFlow community for Avro IO. Sounds exciting, but it seems the performance sucks and people complain about our training I/O performance all the time. It occupies 98% of the training time in some cases.
• We hand held an AI customer perform an experiment and proved that library X could improve our training speed by 2x. Sounds exciting, took us 2 engineers * 1 month. Ok, experiment was successful, however (1) we gained almost no expertise, we handheld that team but we don’t yet have the necessary fundamentals to understand how that library worked. (2) the experiment was not moved on to production, it is still a successful experiment, not something powering up production. (3) we kept being bugged by different people to proceed on that experiment and bring that to production, but we have 10 more things to do and I have no clue how to choose now. It has become a persistent headache at back of my head.
• We did Kubernetes logging for big data jobs. Ok, what does this even mean? We don’t even know how we want to deal with Kubernetes for AI training, what is value of this single component?
• We helped a team improve their training speed by 5x. We learnt almost nothing and we didn’t get time to implement the gaps when we helped them debug, we provided a manual tuning service.

I have a much longer list, but very few things I can sing a song about publicly. I started to realize something might be wrong.. and reflected on my boss’ mantra of “fewer things done better”.

This “fewer things done better” principle is very counter intuitive. You are a new manager, you are passionate about building new things, you have a capable team, you want to change the world with your team. You want to build as many things possible, you want to start 10 different initiatives in the same time. However,

More is better is a horrible myth for any organization, any team, any individual.

More is harder, not better. Smart people tend to show off their skills by doing more, but that is not how you scale an organization, you want simplicity.

Jeff Weiner shared a story of Steve Jobs back when he was at Yahoo:

When Steve Jobs re-joined Apple, Apple was months away from becoming bankrupt, insolvent. He asked his leadership: “If you could only do one thing, what would it be?” After a few round of answers, he said: “Let’s go back to iMac, let’s reinvent, let’s reimagine the iMac”. And then they focused on that one thing, and they got it right. Once they got iMac right, they started working on other things, iPod, iPhone, iPad, etc.

Back to my story, the primary reason why I committed a laundry list of OKRs is: I don’t know really what I need to do, I am strategically lazy. It is hard to distill and articulate one or two strategical objectives you really want to from a galaxy of ideas. Leader’s laziness at strategical level inevitably would push the team to the edge of an abyss.

It is far easier to say Yes to everyone and do a ton of random things than understanding what you need to do, stay focused and say No. The harder part of leadership is to figure out focus on who, and focus on what, how to focus. There is one saying I heard from a talk:

Tactical diligence will never be able to compensate strategical laziness in a sustainable manner.

You might still wonder, besides the dissatisfaction from you boss, what are the other consequences of not staying focused? There are a few repercussions in my observation of my team and sibling teams.

• Your team end up becoming a customer support team. Your team is stretched to support different customers, but your team doesn’t have time to grow necessary expertise for long term success.
• Nothing is DONE DONE. You embarked 20 OKRs in two quarters, all made some progress, few got to a barely “usable” done state. The “almost done” projects would continuously haunt you and you get endless user complaints. You have a permanent headache choosing between “make XX product greater again” vs “another project that hasn’t got to a barely stable state.”
• Team burnt out but received no appreciation. When you commit to too many OKRs, you are absolutely overcommitting your team. Your team has to burn their mid night oil every day to achieve the goals you committed for them. However, in your leadership review: you only hit half the OKRs after each quarter and scored Red. In your customer A’ review: you committed to finish this, but you didn’t keep your promise. In your customer B’s review: you delivered some shitty product. Now folks in your team look at you, thinking: wtf, I’m burning out but we are like shit in front of everybody.

How to focus?

Focus is a new hard lesson learnt for me , so I don’t have tremendously successful stories to share yet. But I have a few ideas learnt from others that I’m experimenting:

• Focus on key customers, use cases. Understand your immediate customer, or core target audience. It should not be everyone, every case. For example, in our case, our prioritized scenario should not be all AI training possibilities, but use cases funneled through our AI production platform. We will support other cases, but they are in the customer support category, not strategic category.
• Align with company priorities. You have to align with company priorities in order to succeed. You must understand how your customers and management define success.
• With clarity in Target Audience and Company Priority in mind, you can brainstorm with your leads to turn your two page long OKR laundry list into 2 sections, (1) KPIs & Strategic Objectives. KPIs for known domains, OKR for unknown/new initiatives. You have 2–3 KPIs each you want to track for at least a year. 1–2 strategic objectives each runs for at least 6 months. (2) Customer Support. For (1), your team and you should all onboard that those goals and performance indicators are something reachable, and transcendent that they want to spend their life working on. For (2), you need to consciously balance the time between (1) and (2).
• Sending monthly team update to stakeholders. Organize the update in 3 sections: 1. Shipped (projects done done done, customer ask or incremental milestones in your OKR/KPI projects) 2. Updates for the KPI/strategic and customer asks. 3. Goal of next month. The purpose of this update is many folds: 1. Keep everyone focused, everyone should strive to get fewer things done better (Shipped). 2. Bring transparency to your customers and management so they could help you correct course. 3. Helps you think, every time you write up the summary you would identify some gaps.

Focus is also applicable to personal life. Every one has a limited span of life, even the smartest people can’t excel in every area in modern world, we should all focus on our personal mission/purpose of existence.

In the next story, we will discuss: Concept + Capacity = Green Light.