Run the Numbers – How Superhuman Structures Its Analytics Team

Host: CJ Gustafson
Guest: Chris Byington, Head of Data at Superhuman (formerly Grammarly)
Date: March 5, 2026

Episode Overview

In this episode, CJ Gustafson talks with Chris Byington about how Superhuman structures its analytics team to drive business growth and data-driven decision making. They discuss the practical details of embedded data science, how analytics should partner with the business, the pitfalls of certain metrics (like ship goals), and strategies for setting impactful goals. They also explore team structure, the debate of build versus buy for data tools, fostering self-service, and proving the ROI of an analytics function.

Key Topics and Insights

1. Analytics Team Structure and Org Positioning

• Superhuman’s analytics team operates as a centralized function responsible for core analytics, BI reporting, product data science, company goal-setting (OKRs), and FP&A forecasting ([04:58]).

• The mission: "Use data and facts to improve business outcomes for customers and for the business." (Chris Byington, [04:58])

• By housing analytics, goal-setting, and forecasting together, there's tighter alignment, fewer handoffs, and better strategic results ([06:10]).

  “So they're perfectly on the same page. So that the strategy and the insights… perfectly kind of ladder up into your FP&A ARR model, which I think has helped us grow a lot faster and make sure everybody in the company is doing things that directly drive business growth.”
  — Chris Byington ([05:28])

• Placement of Analytics in the Org:

  • Three common homes: Finance, Operations, or Engineering ([19:10]).
  • Where it works depends on executive sponsorship, product needs, and company maturity.
  • Analytics in Finance gives more control and authority; in Operations can be more neutral; in Engineering, better for data-powered products ([21:28], [22:59]).

2. Building an Analytics Mindset & Early Wins

• Start with problems, not solutions: Don’t jump to tools or dashboards. Understand core business needs and processes ([08:12]).

• Growth Model Alignment: Make sure everyone in the company knows how the business works (acquisition, conversion, monetization, etc.). Simple shared vocabulary boosts impact ([09:21]).

• CJ’s story: Companies often build “interesting” dashboards disconnected from business levers, leading to wasted analytic effort ([16:41]).

  “Few problems survive their thorough articulation… the solution’s easy. A lot of people just need a pivot table.”
  — Chris Byington ([08:12])

• Pre-Negotiate Actions: Before doing analysis, get stakeholders to commit to changing behavior based on what the data says ([11:22]).

• “Being data informed…is being willing to be proven wrong.” (Chris Byington, [12:32])

3. Hub-and-Spoke Model and Embedded Analytics

• Analytics team operates as a hub-and-spoke:

  • Horizontal: Data platform (data warehouse), BI and data enablement (self-serve, support)
  • Vertical: Embedded data scientists in key business domains (marketing, product, sales), spending 80% with their stakeholder group ([24:09], [29:48]).

  “You spend 80% of your energy with that stakeholder group. You’re much more member of their team than you are of the central team.”
  — Chris Byington ([24:58])

• Not just a ticket queue—success depends on deep partnership and context, not transactional requests ([25:34]).

4. Self-Service Analytics: Potential and Pitfalls

• Self-service is a must, but tooling alone isn’t magic.

  • Rough goal: 70% of questions answerable without data team, measure team-confidence with surveys ([29:59]).

• Tools must be point-and-click, with standardized metric definitions—so non-technical staff don’t have to learn SQL ([31:46]).

  “Data teams who invest a lot of time in training stakeholders on writing SQL: I think that’s a losing game.”
  — Chris Byington ([31:53])

• Define key business metrics unambiguously; otherwise, “self-serve” produces diverging numbers and friction between execs ([32:26]).

5. Buy vs. Build Analytics Tools

• For critical platforms (data warehouse, BI tools), be slow to choose and map requirements first, as mistakes are hard to unwind ([33:44]).

• Choose tools that reinforce your business model and workflow (e.g., inline spreadsheets for OKR tracking in Omni) ([34:19]).

• At Superhuman: Complete ban on using Excel/Google Sheets for official reporting ([35:35]).

  “You want to reduce the amount of energy that’s spent debating what’s true; maximize the amount of energy debating what to do about it.”
  — Chris Byington ([34:33])

6. Metrics, OKRs, and Forecasting: Owning the Loop

• Having analytics own metrics, OKRs, and forecasting closes feedback loops and connects IC work directly to company value ([35:54]).

• Hero metric: A single, top-level metric—N.B. for teams, it was NRR (Net Revenue Retention), enabling strong focus and prioritization ([38:27]).

  “Our hero metric was NRR… It’s like a coin operated machine. When we add a quarter, they grow over on their own… you can imagine like the, the candy mountain charts of cohorted ARR.”
  — Chris Byington ([38:27])

• Ship goals (e.g., "Did we launch X?") are inferior; real goals should be outcomes (retention, expansion, CSAT), not just shipping output ([41:03]).

7. Quantitative Goal-Setting & Measurement

• Pre-define the action to take based on metric outcomes—important for product launches ([42:09]).

• Set threshold values for green/yellow/red status on metrics upfront ([42:49]).

  “If you can’t say that, then you might as well not do it.”
  — Chris Byington ([42:51])

• Bring analytics in at the start of projects, not after the fact, to define and instrument measurement ([44:14]).

8. Prioritizing and Saying No

• Use company goal framework to auto-reject projects not aligned with goals ([46:40]).

• Invest in self-service so that most requests don’t need a data team, freeing time for high-impact, company-level work ([46:40]).

• Data/finance teams should embrace being “kingmakers”—enabling others to succeed ([47:38]).

  “Our job is to kind of be king makers and help everybody else do a better job.”
  — Chris Byington ([47:56])

• Foster a “yes, and…” service mentality; redirect requests with training or alternative solutions ([48:29]).

• Use dedicated office hours as a “pressure valve” for handling and triaging requests ([50:17]).

9. Proving the Value and Impact of Analytics

• Gauge value by comparing a world with vs. without a data team—do people notice if the function disappears? ([51:05]).

• Metrics include self-service survey scores, evidence of changing company-level strategic decisions, and “pull” from stakeholders (like execs not wanting to lose data partners) ([51:05], [52:56]).

  “If you can’t answer that, like, hell yeah, they’re way different. Then like, I think you have a serious problem.”
  — Chris Byington ([51:18])

Notable Quotes & Memorable Moments

| Timestamp | Speaker | Quote / Moment | |------------|--------------|----------------------------------------------------------------------------------------------------------------------------------| | 04:58 | Chris | "Use data and facts to improve business outcomes for customers and for the business. Super simple." | | 08:12 | Chris | “Do not provide any solutions that do not have problems… Few problems survive their thorough articulation… solution’s easy.” | | 12:32 | Chris | "Being data informed… is being willing to be proven wrong." | | 16:41 | CJ | Story of building hurricane dashboards—“interesting but not really impactful” | | 24:58 | Chris | “You spend 80% of your energy with that stakeholder group. You’re much more member of their team than you are of the central…” | | 29:59 | Chris | “About 70% of questions… can be answered without the intervention of the data team… That’s a good benchmark.” | | 31:53 | Chris | “Data teams who invest a lot of time in training stakeholders on writing SQL: I think that’s a losing game.” | | 34:33 | Chris | “Reduce the amount of energy spent debating what’s true; maximize the amount debating what to do about it.” | | 38:27 | Chris | “Our hero metric was NRR… It’s like a coin operated machine. When we add a quarter, they grow over on their own.” | | 42:51 | Chris | “If you can’t say [if a metric is green/yellow/red], then you might as well not do it.” | | 47:56 | Chris | “Our job is to kind of be king makers and help everybody else do a better job.” | | 51:18 | Chris | “If you can’t answer that, like, hell yeah, they’re way different [with a data team], then you have a serious problem.” |

Technical Stack (55:26)

• Data Warehouse: Google BigQuery
• ETL: Fivetran
• Transformation: dbt
• BI / Visualization: Omni

Lightning Round Highlights

• Biggest mistake: “Done is better than perfect—missed an immovable deadline trying to make a report too nice.” ([53:34])
• Career advice: “Stop trying to get promoted. Focus on things that contribute to the company, and promotion is the exhaust, not the goal.” ([56:03])
• Best advice for his younger self: “Don’t take work so seriously…there’s more to life.” ([54:53])

Related Timestamps for Key Segments

• Team Structure: [04:58] to [06:10]; [19:10] to [24:09]
• Growth Model & Early Analytics: [07:53] to [12:32]
• Self-Service, Tooling, & Standardization: [29:57] to [35:35]
• Metrics/OKRs Ownership: [35:54] to [38:27]
• Quantitative Goal Setting: [42:09] to [46:29]
• Prioritization & Saying No: [46:29] to [48:29]
• Proving Impact: [51:05] to [52:56]
• Technical Stack: [55:26]
• Lightning Round: [53:34] onwards

Final Thoughts

Chris Byington offers a pragmatic, clear-eyed approach to building analytics as a business partner, not just a reporting function. Key takeaways include the importance of shared language and growth models, embedding analysts deeply in business functions, the limits of tool-driven self-service, and aligning analytics to the company’s highest-value outcomes. Both the tactical (tool choices, team structure) and strategic (measuring impact, saying no, focusing on business impact) are covered with practical, candid advice.