Moving data teams from cost center to profit driver

Data teams are having their Jevons Paradox moment.

The historical price of computer memory and storage has plummeted. In 1990, a terabyte of storage would cost you $10 million. Today, that same terabyte costs a mere $50. Similarly, cloud providers such as AWS continue to slash prices on services; AWS has famously lowered prices 107 times since its inception.

That doesn’t mean, however, that data teams can stop worrying about costs. Because, as costs have gone down, demand for data has gone up. That has data teams consuming more resources than ever.

That’s the Jevons Paradox: Increased efficiency creates increased demand. It’s like adding new lanes to a highway. An added lane will alleviate traffic for a while. Eventually, more people start driving because they realize they can get around faster. Pretty soon, the road’s backed up again.

To address this challenge, more data teams are transitioning from cost centers into strategic drivers of value. We talked with some experts—Ben Kramer, senior director of data analytics at Bilt Rewards, and Colin Lennon, an analytics engineer at ClickUp—about how they’re navigating this transition through deliberate strategies and cultural shifts.

The mindset shift: From reactive to proactive

Moving from a cost center to a profit driver means going from a reactive to a proactive mindset when it comes to data. That means going out and seeking out value, rather than waiting for stakeholders to approach with questions. At ClickUp, a project management platform, the analytics team implemented this mindset shift, positioning themselves as value creators rather than service providers.

Similarly, at neighborhood loyalty program Bilt, data engineering leaders looked for opportunities beyond reporting and analytics, extending their reach into operational use cases. That’s moved the data team from the background to being a product displayed directly to end customers.

This transition from a background support function to a front-and-center role in product delivery fundamentally changes how data teams are perceived. With greater visibility comes increased responsibility and more stringent Service Level Agreements (SLAs). The upside? Additional resources and recognition of the data team's critical function within the organization.

Quality and standardization as cornerstones

When data becomes customer-facing, quality standards naturally increase. That requires doubling down on data quality practices and instituting gates and processes to ensure all new data projects meet a high bar for release.

For Bilt, the shift to profit driver led them to lean more heavily into good analytics development lifecycle practices. These include:

• Increased code reviews
• Using tools such as the dbt Semantic Layer to create singular definitions of metrics used across teams
• An increased focus on data documentation, so that engineers, analysts, and business decision-makers better understand the data they’re using

Engineers at Bilt feel more bought into creating high-quality data because they know they’re more directly embedded in the business. “The engineers understand the importance of the data that they create so that it can be displayed to the customer directly,” Kramer said.

ClickUp found that aligning with the business on standardized key metrics was critical for success. By establishing common definitions for both themselves and downstream functions, they created efficiencies that ultimately reduced costs while improving value delivery.

Measuring ROI: The perennial challenge

For many data teams, demonstrating return on investment remains challenging, particularly for platform work. Changing this requires a shift in thinking—both on the part of data teams as well as on the part of the business.

Bilt found that their operational integration - embedding data teams deeply within product teams—provided natural Return on Investment (ROI) justification. When data directly supports customer-facing features, the business value becomes self-evident. In this situation, it’s less likely that every new query will come under stringent quarterly cost scrutiny.

ClickUp takes a direct approach to ROI measurement. When they take in new work—whether work that comes to them reactively or that they find proactively—they associate it with Annual Recurring Revenue (ARR) or OKRs (Objectives and Key Results). That enables them to show how their work ties back directly to the business.

This focus on measurable outcomes helps data leaders make strategic decisions about where to invest their limited resources. If a project can't be tied to business growth or operational improvement, it may be deprioritized in favor of initiatives with more straightforward value propositions.

Balancing cost control with innovation

One concern with over-focusing on cost control is that it might stifle innovative thinking. Maintaining innovation is critical, especially as organizations grapple with how best to incorporate AI into their businesses. The key is to keep innovation at the forefront, but always have costs in the back of your mind.

Tying back experiments to business OKRs alleviates much of this concern. “Especially with AI,” Lennon said, “there needs to be a little bit of room to kind of experiment and say hey, I'm willing to take this bet and this is going to provide value for us.” If the bet doesn’t pan out, he said, it can be refined and tried again.

ClickUp isn’t shy about experimenting with different AI-based approaches—whether that’s leveraging Snowflake’s Cortex technology or feeding product review data to a Large Language Model (LLM) to extract sentiment. The data team is primarily using AI now to facilitate existing workflows built upon dbt models to make them more efficient, with a focus on growth marketing.

Bilt follows much the same tack, experimenting with building a natural language interface for their business users. Instead of users asking data questions via Slack, which requires an engineer to stop their work, go into the data warehouse, make a query, etc., they can self-service answers using a natural language UI that connects Claude to Google BigQuery.

That experimentation, however, can lead to budget pressure. “Self-service is awesome until you look at your BigQuery bill or Snowflake bill at the end of the quarter,” said Kramer. Bilt handles this by building out alerts for high query or per-user costs. It’s also committed to obsoleting high-cost tests or dbt model builds.

Creating a culture of cost consciousness

Part of reducing costs while innovating is fostering a cost-consciousness culture. As the keepers of the data keys, data teams are uniquely positioned to drive this shift.

Bilt’s data team partners with finance to review contracts, analyze tool usage, and identify opportunities for efficiency. For example, the data team led an initiative to reduce overhead within both their eventing tools and the BigQuery engine. While very much a behind-the-scenes workflow, it’s an important one that supports the company’s growth-oriented teams, such as platform or engineering.

At Clickup, the team asks their stakeholders to help them identify how to associate a given initiative with business goals or OKRs. That drives judgments about whether the team can cut costs.

ClickUp’s Lennon also said that concerns about costs tend to ebb and flow and that the data team remains flexible in response to that. Before the COVID-19 pandemic, for example, the focus was on shipping fast and making products “good enough.” During the pandemic, the focus shifted back to focusing on costs. With things settling back down now, Lennon says the team is finding “a happy medium” that balances rapid delivery with cost consciousness.

How a $2 test led to thousands in savings

Another area of focus for both teams is hidden costs. Tool sprawl and unused data models represent significant hidden expenses for many organizations. Small inefficiencies, left unmonitored and unaddressed, can compound into significant expenses over time.

ClickUp experienced unexpected cost spikes that necessitated an investigation into historical usage patterns. By implementing cost monitoring dashboards, they gained visibility into spending trends and could identify anomalies for further investigation.

Bilt implemented a two-layer approach to cost visibility: internal dashboards that track hourly costs by user and automated alerting for high-cost operations. This visibility enabled them to identify a downstream test costing approximately $2 each time it ran (hourly). On further inspection, they realized the test had been rendered redundant by upstream testing. Removing this single test translated to thousands of dollars in monthly savings.

Balancing access with control

As Bilt realized, self-service is great until the organization finds itself saddled with a big bill. As self-service analytics capabilities expand, especially with AI-driven natural language processing, organizations need to balance democratized access with cost and security implications.

Bilt has felt the effects of the Jevons Paradox on its own data team. A year ago, the team had two people. The team is now six people. And those six are just as busy, if not more so, than the two were last year. As the excitement around data and ease of access grows, so does the demand.

Using Claude and BigQuery, Bilt can do far more with its six people than it could otherwise. However, that requires constant monitoring to ensure costs don’t spiral out of control.

ClickUp said it had to shift its mindset around new data projects. Previously, the goal was to build things that were useful to the business. Now, as their work has gone global, the team has to consider other factors, including:

• How sensitive is the data?
• How valuable is the use case we’re supporting?

Within this framework, says ClickUp’s Lennon, the company takes a stance of granting access to data, on the assumption that greater access will drive faster project completion and get the business to its OKRs.

Looking to the future

Both Bilt and ClickUp see themselves facing challenges with scaling. As Bilt grows as a company quarter over quarter, it’s challenging the data team, both in terms of servicing the business as well as retaining innovation without increasing costs.

Bilt’s focus for the near term is two-fold:

• Deeper product analysis to better understand customers
• Leverage AI in both internal and external use cases—e.g., enabling customers to interact directly with an ecosystem-connected LLM

Both of these initiatives, while valuable, could lead to cost spikes that threaten business ROI. That makes monitoring and reining in costs an ongoing priority.

Meanwhile, ClickUp is asking similar questions of its burgeoning AI initiatives: Is this given project helping us be more effective? Or is it just incurring costs? The company is looking at how it can build AI functionality into its existing infrastructure, as well as leveraging AI tools to boost productivity within existing workflows.

There may be no escape from the Jevons Paradox. However, as Bilt and ClickUp have shown, fostering a culture of cost consciousness and making smart, strategic bets on technologies like AI can enable data teams to do more with less. By working closely with business stakeholders and tying data projects back to business objectives, data teams can shift from being perceived as a cost center to being regarded across the organization as a key driver of value.