Ali Ghodsi on the 5 Ingredients a CEO Needs

A. Databricks

• Databricks was founded in 2013 and reached $1.6M ARR by EOY 2015. Ali became CEO in Januay 2016. 
• Despite having great products, growth was flat for 2 quarters - Ali was tasked with scaling revenue
• He shifted the company from a pure product-led / zero touch motion to a sales-led go-to-market (GTM) motion
• Fast forward to 2024, Databricks has 8,000 employees and $2.5B in Annualized Revenue Run-Rate

Ali on PLG versus Sales at Databricks

We launched an effort called zero touch, which meant we are going to touch our customers zero times. We instructed all the sales teams to stop talking to their customers and focus everything on just making the product seamless. We thought then everything would just grow by itself. Revenue was kind of increasing [initially], and then it got flat for two quarters. I think we all then realized, oh shit, that's not gonna work.”

B. Ali’s 5 Leadership Attributes

• Ali looks for five ingredients in every leader he hires: 1) vision, 2) strategy, execution, 4) hiring, and 5) culture
• He views execution and team as the most important factors followed by strategy, vision, and culture
• A high-performing team executing quickly and effectively can compensate for mistakes in strategy
• You should rate yourself as a CEO and your direct reports on each of these five leadership attributes

1. Vision

• A leader must communicate a compelling vision that motivates employees, especially as the company scales. This vision helps people connect their daily tasks to a larger purpose, inspiring them to work their “asses off”
• At 10-20 employees, this is less important as the founders know everyone and the objectives are clear. As you grow to 200 employees, this is no longer true. Great leaders regularly and clearly communicate to all employees to make sure they understand the mission and are motivated to work hard.
• Ali looks at Salesforce and Marc Benioff as a successful example of a leader that has mastered this ingredient. Salesforce employees are generally extremely excited by the mission and willing to work extremely hard despite selling sales productivity software

Ali on why vision matters

“In a small team of 10-20 people, you don’t need to articulate the vision much—they already get it. But when you have 200 or 2000 employees, especially those in departments like marketing, they may not intuitively understand what the company does. You need to constantly communicate a clear and inspiring vision. If you can do this well, you can energize people to go far beyond what they thought possible.”

• If you start noticing teams that say things like “I’m burned out”, “this is hard”, “customers are calling to complain” – it is likely that the manager running the team does not have this ingredient. Ali ran into this issue in his support team. By changing his leader, the support team went from the lowest culture survey tools the highest scores within 12 months

Example: Vision and Customer Support at Databricks

• The customer support team was facing constant pressure from customers who were frustrated with the early-stage nature of the product. Many on the team were burned out by the repetitive nature of the work
• As a result, the support team at Databricks had the lowest culture scores at the company

Challenges:

• Constant complaints: Customers were unhappy with issues, and support was taking the brunt of the pain
• Burnout: The work was emotionally draining, and the team felt isolated from the company’s mission

Ali on his support team challenges

“Customers are calling us. They're saying Databricks doesn't work. They are complaining. They are yelling at us on the phone. The product is just broken and I’m burnt out. I need to leave.”

Solution

• Ali hired Hatim Shafique to lead the support team at Databricks. He immediately reframed the role of support at Databricks. He told the team the platform is amazing, but the product is early. It has issues. They (the support team) were the frontline defenders of the product’s success
• He also created a clear roadmap for scaling support, so the team felt confident they weren’t going to be running on a never ending wheel of customer pain as the product and the support function grew over time

Ali on his support leader’s vision

Databricks is an amazing platform. But right now, it has issues. So you can't be successful. We are the frontline soldiers to make this platform work. We are the most important department in the whole company. We make the customers win. Without us this platform is not really usable. It will be one day, because we believe in this company, but right now it's all, on our shoulders”

2. Strategy

• A leader must craft a winning strategy. They must study their market, identify competitors' weaknesses, and formulate a plan to win. Ali recommends reading Good Strategy, Bad Strategy as a starting point
• A key part of Ali’s framework on strategy is studying competition. To make good decisions, you must understand your competitors' product, their strategy, and their weaknesses. Ideally, your strategy takes advantage of your competitors' vulnerabilities. 
• Ali takes inspiration from generals/global conflicts. In conflict, countries/militaries study their enemies deeply when devising their strategies

Ali on strategy

“For each of you on this call, there are probably five other companies claiming to do the same thing you are, each with their own angle. Some of your competitors are much bigger than you, in a completely different weight class. Others might have a special advantage or secret sauce. So how do you plan to beat them? What are the one or two things—maybe just one thing—that will give you an unfair advantage over the next 12 to 18 months, where you can use your competitor's weaknesses to your benefit?"

Example: Devising a winning strategy versus AWS

• Databricks realized they couldn’t compete with Hyperscalers on price when offering open source Spark
• Given they were the creators of Spark, they knew all the committers of the project and understood the APIs
• They developed a strategy focused on speed by carving out specific parts of the Spark engine and replacing them with much faster components. This new engine, called Photon, was not open source. It allowed them to maintain full API compatibility with Apache Spark while offering significantly better performance
• Their goal was to get customers in a proof-of-concept (POC). Once in POC, customers would see the product is 10x+ faster and would commit to using it more versus standard Spark via a Hyperscaler.

Example: Devising a winning strategy versus Snowflake

• When they launched their Data Warehouse, Snowflake had the best Data Warehouse on the market
• The strategy was to disrupt Snowflake’s model. They told customers that instead of sending data to Snowflake or Databricks, store it on your own lakes and let Databricks process it for you. This lets you keep control of data and costs without losing functionality. 

‍

Ali on positioning versus Snowflake

“Instead of building a data warehouse, we recommend that you don't give your data to Snowflake—or even to us. You can’t trust any vendor with your data because they’ll lock it in and drive up costs. The same would happen if you gave your data to us. Our recommendation is to own your data, store it in your own lakes, and let us process it on your infrastructure. This strategy is disruptive because it gives you control over your data and reduces costs, without relying on any one vendor.”

Best practices for studying competitors

• Ali strongly disagreed with the common business advice that says, “Ignore your competitors, focus only on your product.” He believes this approach can be dangerously naive, as it blinds companies to what’s happening in the market and allows competitors to pull ahead without being noticed.

Ali on competition

“The idea that you should ignore your competition is nonsense. Only losers believe that. You have to know your enemy if you want to defeat them.”

1. Build a competitive intelligence team

• Ali stressed the importance of having a dedicated Competitive Intelligence Team that actively tracks what’s happening in the market. This team is responsible for gathering intelligence on competitors, analyzing their product updates, market moves, and even their internal challenges.
• The team should not be siloed in marketing or sales. It should be integrated with product management to ensure that insights gained from competitors directly influence product development and strategy.
• Ideal archetypes for this role are Product Managers who come from a Sales Engineering background. This means they are product oriented, technical, and capable of working across many problems and customers.
• This is one of the hardest teams to build, alongside Sales Enablement and Human Resources

2. Gather feedback from customers

• Customers, especially those who use multiple platforms, can provide insights into competitive products
• Product managers at Databricks are trained not only to collect feedback on the company’s product but also to ask targeted questions about how customers view competitors. This data is incorporated into competitive intelligence

Ali on listening to customers

“Listen to your customers. Have your product managers gather feedback on what’s wrong with your product and what’s working. But don’t stop there—have your product management team also gather competitive intelligence by asking customers what they think about the competition and how competitors are performing.”

3. Try to use your competitors’ products

• Ali advocates using competitor products to gain firsthand experience of their strengths and weaknesses. This was previously impossible for Databricks, but Snowflake opened up their platform after Databricks pushed them to do it to enable benchmarking

3. Hiring

• Building a world-class team is fundamental to scaling a business. This involves not only hiring the right people but also managing out under performers. The best way to scale yourself as a CEO, is to hire exceptional leaders

Ali on the importance of hiring

“Building a stellar team that works well together should probably take up half your time, or at least a third of it. This session is about scaling yourself, and the way you do that is by finding other people you can trust and who are awesome.”

‍

Interviews are not predictive. Backchannels, sample work, and evidence of great are.

• Ali thinks interviews are risky. Some people are good at interviewing, but are bad employees
• Instead, Ali built Databricks’ hiring around: 1) backdoor backchannels, 2) work assignments, 3) identifying candidates that have seen “great” and 4) ignoring job hoppers
• Ali admits his system errs on the side of false negatives (missing out on good candidates), but he is OK with this tradeoff. He would rather maintain his hiring bar than let in a larger number of false positives (under performers)

Backdoor backchannels

• Backdoor backchannels are much more predictive than an interview. You learn how someone actually works
  • They do backchannels for all functions and candidates
  • This is easy to scale in sales where networks are small
  • In engineering, this is harder but still possible. 

Ali on backdoor references

“One of our top engineers was weird in the interview—asking odd questions and focusing on work-life balance. I rejected him. But when we did a backdoor reference, we found out that this guy was a perfectionist who worked 24/7 and was being told at his research lab to ‘take it easy.’ That convinced us to hire him.”

• Ali described an instance where an engineer interviewed poorly, but the backdoor reference revealed that he was a relentless worker with exceptional skills. They hired him and he is now a top 3 employee at Databricks

Work assignments

• Another tactic Ali uses is to have candidates perform actual work tasks during the hiring process, rather than relying solely on interview questions. This could include coding assignments or even working sessions that simulate the job they’ll be doing if hired
• When Ali was hiring his CMO, he had him work on improving part of the website with him. This helped him see how the candidate approached real-world challenges, not just how they answered interview questions

Candidates that have seen “great”

• Databricks optimizes for finding candidates who have either seen great things (worked at top companies or attended top schools) or done great things (achieved significant results in their previous roles).

Ali on looking for great

“We look for have they seen great or done great before? Did they go to a great school? Did they work at a great company, or did they accomplish something great in their lives?”

Avoiding job hoppers

• Ali believes in avoiding job hoppers—candidates who switch jobs frequently hinting at lack of commitment

Managing out low performers is key to building a great team

• Most teams do this poorly. First step is to measure regrettable and non-regrettable attrition
• At Databricks, non-regrettable attrition is someone who left who previously had a performance improvement plan
• They aim for 7% non-regrettable and then managers are evaluated on if they are good performance managers
• Including it in performance reviews aligns incentives so managers throughout the company care about this
• A company could aim for 15%-20% attrition, but this would veer towards a Netflix-like cutthroat culture

How does Ali spend his time?

• Ali is a trust but verify type personality. His main focus areas are GTM and Product. A B2B software company's main goals are to build products and sell products. His remaining time is problem oriented – he will focus on areas that are burning or require his attention. 60% of his time is in Product/Engineering, 30% of his time is in GTM, and 10% is in other areas

4. Execution

• Ali compares great execution to personal fitness, like losing weight. Everyone knows what they need to do—eat healthy, exercise, and follow a routine. However, the challenge lies in consistently executing these actions.
• Great execution requires the four qualities below:
  • Structure: Creating a disciplined, structured environment where everyone follows a process
  • Consistency: Ensuring that the process is adhered to regularly and without fail
  • Accountability: Establishing clear systems of inspection, checks, and balances to maintain accountability
  • Regular Cadence: Keeping the process on a predictable schedule with ongoing reviews
• His advice: Pick an established process and follow it. Ali used the example of scrum in engineering
  • You break tasks down into two week sprints, assign tasks, and put them in Jira
  • You put everything on the scrum board to give visibility and a timeline for completion
  • After its finished, you do a post-mortem and implement any changes into the next sprint

Example: How Databricks’ ships fast

• Databricks starts by looking at the total team and estimating the number of person weeks of capacity
• After this, they break tasks down into three buckets: D0 tasks, D1 Tasks, and D2 tasks

D0

• These tasks are the highest priority and must be completed by the end of the quarter or sprint
• Databricks allocates 50% of a team’s capacity to D0 tasks. If a team doesn’t hit, D0 something is broken

D1

• These tasks are the next level of priority. The team focuses on D1 tasks after finishing D0 tasks
•  D1 tasks fill up the remaining 50% of the team’s capacity. Ideally, teams also finish all their D1 tasks

D2

• D2 tasks are tackled only after all D0 and D1 tasks are completed. While D2 tasks don’t get allocated a specific percentage of team capacity upfront, they are used to push high-performing teams to maximize their output

Why does this system work?

• The framework discourages teams from under-promising to make themselves look better later. Teams are always pushed to work at maximum efficiency, even after the core tasks are completed (D0 and D1). Databricks can also use this to benchmark performance across teams and engineers

‍

Ali on why the D0, D1, D2 framework works

“If you ask engineering teams, ‘How much can you do in a quarter?’ they’ll sandbag. They’ll say, ‘This is going to be really, really hard,’ because we all want to buy ourselves a little room, you know? So, they buy themselves time. But the problem with buying yourself time—saying, ‘Oh, this thing is going to take a month’—is that once you say it will take a month, it will at least take a month. It's like if we say a meeting is two hours, you’ll fill up the two hours. If it's half an hour, you’ll fill half an hour. So, we don’t want to get into a situation where teams are sandbagging, maybe finishing early, and then slacking off. Instead, we make everybody work their asses off the whole time, all the time, because you’re never done.”

Regular Inspection is key to great execution

• Great execution requires measurement of key metrics, regular inspection, and some process for course correction
• Whatever “execution” process you design, you need to make sure there are regular mechanisms for inspection
• Additionally, as a CEO, you need to create a culture where leaders operate at any level. It needs to be normal for the CEO to speak to a junior person without their manager prepping them. At Databricks, all managers are expected to do skip levels to go and see what is happening

Ali on inspection

“As a CEO, you better be able to operate at any level. People should be comfortable with the fact that I might go talk to a junior person, and they shouldn't feel uncomfortable. I should be able to speak to that person without their manager, their manager's manager, and so on, being in the room or having a prep meeting to prepare them. That’s why one of our cultural principles is truth-seeking. You should be honest about what’s going on at Databricks. All managers, not just me, should do skip-levels, even multiple skip-levels, to see what’s really happening. Inspection also means that if you're in go-to-market, you should be able to inspect what’s happening with deals. Join customer calls, join forecast calls, and verify. You can’t have a culture where you just take people at their word—you have to go verify.”

5. Culture

• Culture is a softer but essential component. It needs to be consistent and reflect the personality of the CEO. For example, the culture of “inspection” instilled throughout the company is from Ali’s trust but verify personality. This is the opposite of Netflix, which gives employees rope but fires them on the first signal of underperformance