On the eve of Google Cloud Next ’19, we hosted Anders Hagman from Spotify, who runs the Technical Procurement team for the tech unicorn in Stockholm. Spotify is reported have committed to use a fully cloud-native infrastructure to drive their wildly popular streaming music service. Anders’ team is responsible for FinOps in the organization, driving best practices and managing the relationships with Spotify’s cloud providers.
Anders led a lively discussion in the group about how to relate cloud spend to metrics that matter to the business. Many organizations with large cloud spend struggle to give engineers meaningful metrics about spending that put it in terms they can understand. Imagine your organization is going to spend $100m/yr in cloud. At that scale, individuals struggle to understand the size of the numbers involved. Is a $70k a month optimization a little or a lot? How much effort is it worth to achieve? Ho do you keep a culture of cost accountability when the numbers are so large? Engineers, who already are not used to thinking cost as a metric that they need to track, have to begin considering spending and relate its impact to the business.
Anders gave an example of how they correlated the amount of cost savings a team could make to cases of beer. If the team took a certain set of actions, they could save the equivalent of 43 thousand beers. The challenge, however, for the Stockholm-based tech unicorn is taking this type of cost efficient action while also maintaining growth and velocity. When you’re in a highly competitive space, the goal is often less about cost savings than about driving more features and additional growth faster.
Though Spotify represents a special type of company focused heavily on growth, the theme is the same as we see in most cloud-first companies. FinOps is not about saving money. It’s about making money. Spending more in cloud can allow you to drive more revenue, it can be a sign of growth in the customer base or it can enable you to more quickly shut down a data center. It’s about aligning the entire organization to empower engineering teams to deliver better features and enabling a cross-functional conversation about where to invest and when. Sometimes you’ll decide to tighten the belt, sometimes you’ll decide to invest more. In either case, an intentional choice is made based on cross-functional conversation.
As Anders put it, the goal is to “Cut fat, but also to build muscle.” Cutting too deeply doesn’t necessarily help the business grow.
Historically, procurement teams were goaled on reducing rates paid for software. Now, cloud-focused enterprises look at their procurement teams as strategic sourcing dedicated to unblocking engineering team access to technologies they need to deliver better experiences for their users, faster. This aligns well with Spotify’s strategy as laid out in the letter from their CEO in their IPO’s F-1 filing:
“To unlock the potential of human creativity — by giving a million creative artists the opportunity to live off their art and billions of fans the opportunity to enjoy and be inspired by it. To build a better world by unlocking human creativity, we are committed to creating a better experience for users — and to enabling more creators to live off their work. We firmly believe that in the long run, these priorities will provide greater returns to all of our stakeholders.”
This related to a wise maxim shared by Anders for cloud spending: “Spending is fine, but wasting is not.”
The challenge with cloud is telling the difference between the two. Did the money invested have an impact on the business relative to the dollars deployed? In order to determine the impact, the spend needs to be related to a more meaningful metric. It needs to relate to the business in the form of unit economics.
Moving away from simply reporting spend in dollars is important because each team involved with cloud spend has a different set of motivators. Anders brought the concept of the Balanced Scorecard into the conversation. These are a sometimes forgotten general-purpose performance management system from the 90s. Balanced scorecards look at individual teams’ characteristics and motivations in order to drive organizational change.
Some examples: Engineers look to work on meaningful and fun work. Engineering managers look to increase headcount and improve the productivity of those heads. Finance looks at money and how to relate it to forecasts. Sourcing — in their new role in cloud — looks to unblock engineering so they can get what they need to move more quickly.
Leveraging the motivations of the engineering manager, Anders was able to put cloud spend into a metric that mattered for them. Instead of reporting in dollars, which can be ambiguous when you’re spending tens of millions a year in cloud, his team started reporting out cloud spend in terms of the cost per million DAUs (million Daily Active Users). For Spotify, this is a big number. The Swedish-based company recently reported that over 200 million people who use their service each month. Perhaps more impressively, their paid subscriber count topped 87 million customers in January of 2019. The mDAU number helped show the spend in a way that related to the metrics of the business, but it still didn’t hit the motivations of the teams.
The most impactful metric came when Anders’ team began reporting cloud spend in terms of the monthly cost of an engineering squad. In other words, what was the squad-month cost of a team’s cloud spend. By showing engineering managers the impact of their spend in relation to the additional headcount that could be paid for or hired, Anders used a metric that hit directly on the motivations of the engineering team leaders. When evaluating a cost saving action, they were now able to say that that an optimization action would result in a specific amount of additional headcount — or how quickly an action could pay for a team’s salary.
To complete the picture, he began reporting cloud spend in relation to the company’s total revenue. They set a target for a percentage not to exceed. When they realized they were on target to spend more than planned, a campaign to “bend the curve” was put in place. They employed a number of optimization techniques to drive down the growth of cloud spend relative to their revenue growth. This created a cost management program aligned to the top-line goals of the business. The Technical Procurement team looks to FinOps processes to make the bending of the curve possible.
To join the conversation about FinOps, we invite you to join the FinOps Foundation. Each week, we host conversations like this with the world’s thought leaders in this space. To hear a transcript of Anders’ story, join the Foundation now.
Keep breaking down the silos…