Tony Xu on Hiring and Operating Cadence

A. Building your framework and hiring muscle

Build a criteria for what works at your company. Start by writing down your values & culture

• Culture is “80% who you are and 20% who you aspire to be”
• Write down an honest reflection of your company’s current values, not just its aspirations
• It took until year three for Tony to write these down - they were key to figuring out his hiring framework
• This is important because Tony has learned that successful companies operate differently. What works for Tony, might not work for your company. Figure out these values and hire for success on your team

Tony on the importance of writing down your values

“I would start by saying that to get to any criteria, it really stems from the core belief I have that culture is 80% who you are and 20% who you aspire to be. So you need to put on paper, first and foremost, what you think is a real reflection of what your company is, not just a complete aspiration of what you want it to be. A lot of these six things have to do with who we genuinely are. Otherwise, what I’m about to say may not be universally applicable, as I’m not sure there’s a one-size-fits-all approach that works for every company.”

Hiring takes time. Shorten your cycle time by spending time with executives told to be great

• Tony’s tolerance for pain is high - he spoke to 100 candidates during his CFO search. This was fairly consistent for any new role Tony hired for early on - he would brute force the hiring process.
• In hindsight, the process took longer because, for unfamiliar roles (e.g., head of sales, CFO, head of HR), Tony spent time learning from executives considered great in those positions. It took him a while to build his view of the role, what works at DoorDash, and to make a decision
• To shorten your cycle time, schedule regular meetings with executives and make it a consistent routine
  • For 3 years, Tony met 2 executives a week. He met about ~356 executives over this period
  • He asked his board to introduce him to 1 executive per week regardless of role or seniority
  • He would also regularly cold email executives - asking them for time or to shadow them
• While this didn’t lead to executives, it helped Tony build his benchmark for what great looks like

Best practices for building your executive benchmarking data set

• Find executives at companies you believe are representative of the way your team operates
• There is no set agenda - the goal is to learn how they operate within the day-to-day
• Ask leaders if you can shadow them - spend an afternoon with them to see what they do
• Ask leaders how they score themselves - this is a great way to understand what matters
  • ~10 years ago, Tony met with David Eberson (Facebook’s early CFO)
  • David explained that he scores FB’s performance by X and investors score FB by Y
  • This helped Tony understand truly what mattered for the CFO role

Identify the companies and vintages that match your company’s style

• Tony learned that there is no single definition of great for executives
• Companies have different styles and methodologies to get to the same answer
• Identify the companies that align to your culture and the way your company operates
• It’s important to recognize vintages as well - companies change during different cycles
• Companies that match to DoorDash: Meta (2009 to 2015) and Amazon (2004 to 2012)
• The only way to figure this out is to regularly meet with executives that are told to be great

Why invest? A great leadership hire can return 10,000x on your time investment

• A successful hire can yield returns up to 10,000x more than your initial time investment
• A bad hire isn't as catastrophic as 10,000x in losses, but Tony estimates it can result in a negative 100x impact. Therefore, the upside-downside asymmetry is substantial and it's worth getting right

Tony on the asymmetric upside of hiring great executives

“You think about the returns on time, and if you get the right executive, it could be something like 10,000x. I can't think of a better capital allocation decision on time than that. However, if you get it wrong on an executive, it goes the other way. It might be negative 100x, not just negative 1x. Hopefully, it's not negative 10,000x for anyone on the call, but it's definitely very negative. The upside-downside asymmetry is significant.”

Don’t forget about internal promotions. 60% of VP+ hires at DoorDash are from internal growth

• 60% of hires above VP are internal promotions at DoorDash today. Up to 300 people, this was 90%
• Every company is a special snowflake in the way they do things - internals have context and know how
• With externals, you run the risk of organ rejection. They can’t operate within your company’s interface
• It’s easy to value a lot the shiny object outside of the company versus evaluating the talent within
• Tony’s framework for <100 FTEs: If he thinks they can scale beyond 1-2 years, he’ll give them a chance

How to identify internal talent for promotions? 

• If people are drawn to a person regardless of title - that’s a sign that people respect their opinion/work

B. Tony’s Hiring Framework

1. Operate at the lowest level of detail

• Great executives can operate at all levels of detail; they can both lead and do great IC work
• Being in the details is a big part of good judgment - you can’t make good decisions without details
• If you have <100 employees, this matters a lot as you need leaders to be 80% player and 20% coach
  • At this stage, you need a leader who builds, obsesses over details, and corrects as needed
• Tony never compromises on details - this is the greatest predictor of success for employees/execs 

How to identify this attribute?

• Spend a lot of time and listen. Tony spends 15-20 hours with each executive that gets hired
  • Tony often asks candidates about projects they worked on to understand their decision making
  • Most candidates will simplify their experiences into 1 or 2 assumptions that mattered
  • The best candidates can clearly explain the final answer, identify key challenges, and recall important experiments. They provide relevant details and connect the dots between the experiments and the final answer
• Conduct blind references with direct reports or skip levels instead of peers or managers
  • Verify claims, ask about decision making, and whether they felt the leader was in the weeds

Example: DoorDash’s CFO

• When Tony met Prabir (DoorDash’s 2nd CFO), Prabir showed up to the meeting with an excel model
• The model outlined Prabir’s view on DoorDash’s next 10 years and modeled out the key drivers
• Tony and Prabir spent 6+ hours together going through the model cell by cell and row by row

2. Bias for Action

• How do you keep a company fast - you have to hire people that have a bias for action
• If you don't hire people who naturally move quickly, no process you set up will get them to move fast
• Most candidates are good at “textbook math”, but few are resourceful and have a bias for action

How to identify this attribute?

Executives

• For execs - no way to truly ask. Best way is to deeply prosecute their role in past projects
• Problem is executives talk well - you will be lied to. In general, 80% is a lie and 20% is truth
• Like with detail orientation, conduct blind references with direct reports or skip levels
  • Executives can manage upward, but it's impossible to lie to your direct reports
  • On this topic, ask them whether the candidate made decisions quickly
• Assignments can also be a useful indicator and also ensure the person is willing to do the work

ICs

• Tony used to give candidates $20 and ask them to acquire 100 customers in 24 hours
• Most people quit the interview - but the best candidates get straight to work on the problem
• Tony found this very predictive for ICs that were launching markets, for salespeople, and for product

All Personas

• Look for slope by assessing how much someone accomplished per unit of time at each role
• If you look at companies, there is usually a predictable path it takes to get promoted
• If someone has risen through the ranks very quickly at a company, this is an indicator of slope

Tony on slope

“Let's say you want to become a VP at Amazon, and you join right out of college. Roughly, it takes about 10 to 12 years. So, if someone gets promoted to VP in six to eight years after college, something unusual happened. Now, it doesn't necessarily mean that person is the best thing since sliced bread—that could be true—but if you believe Amazon has a high talent bar, which is true for some parts of the organization, and someone beat the average by a significant number of standard deviations, that's something worth paying attention to.”

Example: DoorDash’s COO and first GC

DoorDash's COO

• Chris Payne had run big parts of Amazon and was a very senior executive
• In their first meeting, they had a long discussion about DoorDash’s logistics algorithm
• That weekend, Chris went home and spent a few hours conducting deliveries on DoorDash
• He wrote Tony a 2,000 word manifesto and said these are the 6-7 things I would ship on day 1
• The fact that he did the work and offered to make changes highlighted a bias for action

DoorDash's GC

• The candidate showed up to the meeting and told Tony the contractor agreement had 17 flaws
• Tony and the candidate spent the interview fixing 14 of the 17 flaws in the contractor agreement
• The fact that he did the work before the meeting and made changes live highlights bias for action

3. Ability to Recruit

• The best executives tend to have a right and left hand they can bring over to their new role
• They also have good judgment of people and a high talent bar - they can identify great talent
• Note - sometimes great “doers” with bias for action are lacking in this attribute. Watch for this

How to identify this attribute?

• Ask the potential executive about the best teams they’ve built and where all the people are now
• If they’ve built great teams, their reports would have scaled and gone on to do impressive things

4. Followership

• The best executives are able to bring other people with them

How to identify this attribute?

• Before interviewing, look to see if there are groups of people that stick together accomplish great things
• In the interview - ask candidates for three people that would join the company after they joined
  • Call these people and ask whether they’d follow your candidate

Tony on followership

"We look for people who have strong followership. So I'll ask them literally to give me a list of like three to five names of people that would join our company after you join. And I will call them individually and ask that question just to call BS [and see] if that is true or false. It's what we call followership."

5. Ability to hold two opposing ideas at once

• Executives will likely come in with a playbook that worked at their old company
• The playbook that worked at their old company might not be right for their new company
• The best executives understand this fact and think about problems from first principles
• Tony thinks this is hard to identify in interviews, but is a great predictor of success on the job
• Tony uses this as a test to see if someone is scaling: if they are rigid in their positions, defensive, and unable to accept counterpoints, they likely aren’t scaling and need to be replaced

How to identify this attribute?

• Ask candidates about career mistakes and things that they got right / caused their success
• Usually with hard projects, it goes either way. The best candidates know how to shift odds in their favor
• You want to learn whether candidates understand the world moves probabilistically and not in absolutes
• In these discussions, candidates tend to talk about conflicting ideas and how they evaluated them
• Lastly, the really good people are effective at learning all the playbooks, not just what worked for them

Tony on holding two opposing ideas at once

"The ability to hold two opposing ideas at the same time...A lot of times this is to test for the like willingness to give up some of their priors. They're gonna come from some awesome company, right? And that's why you recruited them, but they're gonna think that's how they should just do things for your company. And that might be true. There likely may be some like, transpose function that works. But candidly, I think all of our best people realize things from first principles: which things worked and which ones didn't work."

6. 1% better each day

• You can always improve. By getting 1% better each day, the compound effect will lead to substantial progress over time. The best executives constantly try to improve their performance or their functions

How to identify this attribute?

• Ask candidates to explain one specific area in their personal life they are working to improve. The best people are usually committed to something else
• The best people, in Tony’s experience, are exceptional at something else in their life
• As a data point, 30%-40% of Nobel Prize winners are semi-professional or better in a field unrelated to the one in which they won their prize. This phenomenon is studied in the book Range by David Epstein

Tony on continuous improvement

“Thirty to forty percent of Nobel Prize winners are apparently semi-professional to professional in a discipline completely unrelated to the one in which they won the Nobel Prize. They might be violinists, ballerinas, or whatever. I tend to ask questions about what they do in their personal life or what projects they are taking on to improve themselves.”

C. Reasons for Hiring Mistakes

• Insufficient time investment and giving in to desperation. Spending adequate time with candidates is crucial. If you get the hire wrong, they will build a team and will take months to unwind. If you get the hire right, the person can pay off orders of magnitude more than the time you put into the investment and recruiting process. Do not give in to the pressure to make the hire before you feel 100% comfortable.

Tony on giving in to desperation

"Causes for when I’ve screwed up—I gave into my desperation. That’s probably the single biggest factor. For both hires (early CFO and Head of HR), that was definitely true. Contextually, I probably spend around 15 to 20 hours with the candidate myself, and collectively, the candidate will spend about 30 to 40 hours with our company before we make an offer.”

• Giving in to pressure to make a hire you don't actually need. Hiring an executive you don't need is another common failure case. It usually comes from someone you trust recommending a hire or you feeling the need to address a function without truly understanding what you are looking for. Tony's advice is to assess if you have product-market fit in that function. If you do, it's likely a good time to hire as you'll need someone to hire and manage people. If you don't, his gut instinct is you are probably better off waiting .

Tony on when to hire an executive

"It [referring to a mis-hire] was just one of these things where I felt like I don't know like I got pressured into without me feeling like you know, I knew what I was looking for. And so my biggest advice at your stages, like, you know, what are you looking for?

"I think it's a little bit case by case but I think it has something to do with this notion of like, do you already have the product market fit within that function? And if you do, then yeah, I think you're going to need recruiting and all that good stuff. More rigor, probably in that function. If you don't, then my gut is you're probably better off not making the hire. The reward risk ratio is not there."

D. Moving fast at scale

1. Hiring people that want to go fast

• You can implement any principle you want, but it won’t work if you don’t hire people that move fast
• The 80% of the 80/20 rule for moving fast is hiring people who move quickly and get stuff done

Tony on hiring for speed

“You can have the best tires and the best vehicle, but if you have a slow driver, you're not going to win”

2. Identify inputs and build instrumentation for measurement

• To move fast, identify 1) inputs that lead to your target outcomes and 2) frequency of measurement
• You may not know exactly what to measure initially—start with your best guess and refine over time
• For example, DoorDash now predicts the growth of its restaurant business with 99.99% accuracy, but this was achieved through an iterative process over more than a decade
• Once you know what to track, build the instrumentation to measure your selected inputs

3. Pick the right cadence for scorekeeping

• You need to keep score - pick the right cadence depending on the maturity of the initiative 
• For example, Tony might look at newer initiatives daily and he might look at restaurants monthly

Example: Metrics at DoorDash

Goal
Grow restaurant delivery business by $XM of GMV

Input
Grow restaurant supply by Y. This is the scorekeeping metric, not GMV

Measurement/Tracking Frequency
Monthly. Restaurants have 15-30d sales cycles

DoorDash’s Operating Cadence

• Rough math - there are 7 to 8 initiatives DoorDash needs to get right each year to hit their goals
• On top of this - there is a portfolio of 20-30 bets DoorDash makes to ensure success long-term
• The latter is a longer list because you can’t guarantee the success of each new initiative
• Tony tracks each of these initiatives at varying frequencies - depending on maturity

At 50-100 People

• DoorDash was single product (Restaurants) and single geography (USA)
• Tony drove cadence via a Weekly Business Review that he led each week
• They reviewed topline metrics, input metrics, and week/week variance
• They would decide: 1) where to double down, 2) what to stop, and 3) what to watch
• The only problem they were solving at the time: Survival. Found intensity right at weekly pace

At 10,000+ People

• DoorDash has 5 business lines and operates across multiple geographies
• Despite the dimensions expanding, the process for maintaining cadence is still similar
• The main change for Tony is he reviews initiatives at different frequency levels depending on maturity
  • The restaurant business is predicted with 99% accuracy, so Tony reviews this monthly
  • For new bets (seed stage/pre-PMF bets), he might spend 15 minutes a day with each team
• He also adjusts his judgment/patience levels depending on the project or business line
  • A miss in restaurants is treated differently to a miss in one of the pre PMF bets they make
  • Pre PMF bets with 20% success rate require more patience than a mature unit like restaurants
  • Treating everything the same is sure fire way to fail at laying foundations for new bets at scale
• At the company level, Prabir (Tony’s President) leads a Weekly Business Review
  • For the 7-8 top initiatives, they have monthly reviews as these are more mature business lines
  • For the 20-30 new bets, Tony and Prabir review data more frequently (daily, hourly, weekly)

Why do companies slow down? It’s likely you have no decision maker or no deadline

• In Tony’s experience, companies slow down because decisions have no clear owner or deadline
• There are three paths you can take for decision making - no right answer. Choose what is right for you
  • Option 1: Make the decision yourself
  • Option 2: I want the decision made by X time. Y person will decide
  • Option 3: I want the decision made by X time. Y person will decide. This is the criteria
• Tony prefers number 3 because it sets the “rules” and lets him set the standard for the decision
  • Every decision has tradeoffs - by defining the criteria, you are setting the acceptable tradeoff
• Often times, teams at DoorDash move fast because he gives them 24H or 20 minutes to decide
  • Tony does this because he wants teams to be in the habit of moving fast and failing quickly
  • 99% of decisions are reversible - so it’s ok to make a change and to roll it back

Tony’s strategy for encouraging new people to move fast at DoorDash?

• With new people, Tony will encourage shipping imperfect experiments to get them in the habit of failing
• This is initially difficult - new joiners often don’t want to push changes that they might have to rollback
• Below is an example of how Tony encourages shipping fast and “failing” at DoorDash

Context: Tony is working with a new employee at DoorDash.

Employee: "What do you mean? Don’t I need to write a document for this?"

Tony: "No, just ship it. You have 30 minutes, or you’re never going to ship anything—your choice."

[Later...]

Employee: "Oh my God, we got it wrong!"

Tony: "Okay, and what happened?"

Employee: "Well, we degraded conversion by whatever."

Tony: "Cool. What are you going to do now?"

Employee: "I’m going to roll it back and do the next one."

Tony: "Awesome. Great job! Congrats on shipping quickly—you just shipped three things today versus writing a document to ship one."

Constantly evaluate how you spend your time. Create compounding opportunities

• Working long hours doesn’t mean achieving progress.
• Every Sunday, Tony reviews how he spent the last week and identifies 3-4 to dos for the next week
• He also maintains a longer to-do list for projects and ideas that have a longer time scale
• This is an important accountability hour - it forces him to identify whether he is making progress 
• Anything you do compounds
  • He spends 15 minutes/day on support, but this becomes days over time
  • He spends 2 hours a week meeting executives, but this becomes months over time