⏱ Expected: 11:00 (min 0/100)
Talk track: Welcome to Managing Data Science Teams. I'm excited to spend these two days with you. This opening block is about the operating system you'll run as a manager — the repeatable patterns that make your team predictable, trustworthy, and effective. By the end of this session, you'll have two concrete artifacts: a Team Charter and a Stakeholder Map. Let's get started.
⏱ Expected: 11:00 (min 0/100)
Talk track: Here's what we're building toward in this block. Four concrete outcomes — and by the end, you'll have artifacts for each one.
⏱ Expected: 11:00 (min 0/100)
Talk track: Let me introduce myself. I'm Eduardo Ariño de la Rubia, and I'm a Professor of Practice here at CEU. My background is in building and leading analytics and data science teams. This course is two days, and it is intense — the full lifecycle of analytics management. But here's what's different: this is not a lecture course. You're going to build things. By the end of Day 2, you'll have a portfolio of management artifacts that you could actually use on the job Monday morning.
⏱ Expected: 11:05 (min 5/100)
Talk track: So why does this course exist? The analytics industry has a management problem. You're a great analyst, and then someone says "congratulations, you're now managing five people." Nobody teaches you how. The cost of getting it wrong is enormous — talent flight, lost credibility, misaligned priorities, burnout. The good news is that management is a craft. It's learnable. That's what we're here to do.
⏱ Expected: 11:05 (min 5/100)
Talk track: Here's how the course works mechanically. We have three case contexts. Each one represents a different stage of analytics team maturity. DataPulse is a seed-stage startup — you're the first analytics hire, there's no infrastructure, and the CEO wants "data-driven decisions" but doesn't know what that means. MarketBridge is a Series B company — you've been hired to manage a small team that grew organically, and there's technical debt and unclear roles. FinGuard is an enterprise — you're leading a large analytics org with compliance requirements, multiple stakeholder groups, and legacy systems.
⏱ Expected: 11:05 (min 5/100)
Talk track: You're going to pick one of these, and everything you build in this course will be anchored to that context. Your charter, your stakeholder map, your hiring rubric, your roadmap — they all connect. At the end, you'll have a portfolio that tells a coherent story about how you'd manage an analytics team in your chosen context. The portfolio is due one week after Day 2 — polished, refined, and ready for me to review.
⏱ Expected: 11:05 (min 5/100)
Talk track: Okay, decision time. Read through these three contexts. DataPulse is the scrappy startup — if you pick this, you'll be thinking about how to build everything from zero with almost no resources. It's exciting, but it's also terrifying.
⏱ Expected: 11:05 (min 5/100)
Talk track: MarketBridge is the messy middle — the team exists but it's chaotic. If you pick this, you'll be thinking about how to bring order to an existing team. You'll be dealing with conflicting metric definitions, analysts who built their own pipelines, and a team that grew up without structure.
⏱ Expected: 11:05 (min 5/100)
Talk track: FinGuard is the enterprise — if you pick this, you'll be dealing with governance, politics, compliance, and the challenge of leading through layers of hierarchy. Pick one. Write it down. Commit to it. There's no wrong answer, but you do need to commit. Every exercise for the rest of the course is anchored to this choice. Take thirty seconds, make your pick, and let's move on.
⏱ Expected: 11:10 (min 10/100)
Talk track: Let's start with a question — what IS a manager? Andy Grove, former CEO of Intel, gave us the best definition I've ever seen. Read this quote carefully. Notice it says nothing about telling people what to do. Your output is not YOUR work — it's the output of your team and the teams you influence. That reframe is everything. You are not a boss. You are a system builder. Your job is to design the system that produces great work. Who here has had a great manager? What made them great? I bet it wasn't that they were the smartest person in the room.
⏱ Expected: 11:10 (min 10/100)
Talk track: So where does the analytics manager sit in an organization? There are three classic leadership roles that often overlap. The Manager owns people — hiring, retention, growth, performance. The Tech Lead owns technical quality — architecture decisions, code review standards, reliability. The PM owns product outcomes — what we build, for whom, and why. Look at this table carefully. Notice how each role has a different primary concern, a different success metric, a different key ritual, and a different superpower. In a small company, one person might wear all three hats. In a large company, these are distinct roles that need to collaborate. The tensions between these roles are some of the most productive tensions in an organization — when managed well.
⏱ Expected: 11:15 (min 15/100)
Talk track: The analytics manager role is uniquely demanding because you can't fully delegate any of the three circles. No technical depth — your team won't respect your judgment. No product sense — you build things nobody uses. No people skill — your best analysts leave. This is not "choose two." You need all three. That's what makes this role hard, and it's also what makes it rewarding.
⏱ Expected: ~11:15 (min 15/100) | SKIP — advance past
Talk track: Let me make this concrete. Imagine your VP sends a Slack message at 9am: "Why is churn up? I need answers before the board meeting." Watch how the three roles respond differently. The PM gives business context — they know the churn number, they have a hypothesis, and they can frame why it matters. The Tech Lead does the actual analytical work — they build the model, run the analysis, and present the findings with appropriate rigor.
⏱ Expected: ~11:15 (min 15/100) | SKIP — advance past
Talk track: But the Manager? The Manager's job is to make sure the right person is working on it, that they have the time and space to do good work, that someone else's project isn't silently slipping, and that there's a clear deadline and delivery format. That orchestration work is invisible but essential. If nobody does it, the analysis either doesn't happen, happens too slowly, or happens at the cost of three other commitments nobody bothered to renegotiate.
⏱ Expected: 11:18 (min 18/100)
Talk track: Team Topologies is a book by Skelton and Pais that gives us a language for how teams relate to each other. I want you to think about these four types through an analytics lens. Stream-aligned teams are embedded — they sit inside a product squad and serve that squad directly. Think of a "Growth Analytics" team that sits in the Growth pod — they go to the Growth standup, they know the Growth roadmap, they're fully context-loaded. Platform teams build the infrastructure everyone depends on — the data warehouse, the ETL pipelines, the BI tools. Enabling teams exist to make other teams more capable — data literacy programs, self-serve training, office hours. And complicated-subsystem teams handle genuinely hard technical problems like fraud detection models or demand forecasting. Each type has different management challenges. Stream-aligned teams risk losing their analytics identity. Platform teams risk becoming disconnected from business outcomes. Enabling teams struggle to show impact. Complicated-subsystem teams can become ivory towers.
⏱ Expected: 11:18 (min 18/100)
Talk track: Now take this framework and apply it to your case. If you're DataPulse, you don't have the luxury of team topologies — you ARE the team. But it's still useful to think about which hat you're wearing at any given moment. Am I doing platform work right now, or am I doing embedded analytics? That awareness helps you manage your time.
⏱ Expected: 11:18 (min 18/100)
Talk track: If you're MarketBridge, I bet your biggest problem is that you have a bunch of stream-aligned analysts with no platform underneath them. Everyone built their own pipeline, everyone defined metrics differently, and now you're trying to consolidate. Sound familiar? Your first move is probably to carve out someone whose job is owning the platform — even if it's 20% of one person's time to start.
⏱ Expected: 11:18 (min 18/100)
Talk track: And if you're FinGuard, you probably have all four types but they've calcified into silos. The ML team ships models that the business analysts can't interpret. The platform team builds infrastructure the analysts don't use. Your job is to create the interaction modes between these teams — regular syncs, shared documentation, explicit service-level agreements between the platform team and the stream-aligned analysts.
⏱ Expected: 11:21 (min 21/100)
Talk track: Here's the core mental model for this block. Your Manager OS has four layers. Cadences are the clock — how often things happen. Rituals are the actual meetings and practices. Artifacts are the written documents that create shared memory. And decision hygiene is how you make sure decisions are actually good. I like the software analogy: cadences are your cron scheduler, rituals are your running processes, artifacts are your database, and decision hygiene is your error handling. Over this block, we're going to build out each layer for YOUR context.
⏱ Expected: 11:21 (min 21/100)
Talk track: I want to add a layer to the OS model. LLMs are becoming a force multiplier for every layer we just discussed. You can use an LLM to draft your 1:1 agendas from your manager log, draft decision memos in 5 minutes instead of 45, stress-test your own reasoning before sharing it, and translate technical work into business language for stakeholders. This isn't about replacing your judgment — it's about freeing up your time so you can focus on the judgment calls that actually require a human.
⏱ Expected: 11:21 (min 21/100)
Talk track: Let me give you two concrete examples. First, take your last three weekly team updates, paste them into an LLM, and ask for a draft executive summary. You'll get a solid first draft in 30 seconds. Then you spend 5 minutes editing for nuance, accuracy, and judgment. That's 5 minutes instead of 30, and the quality is just as high because you're the editor. Second, before a difficult stakeholder meeting, role-play the conversation with the LLM. Ask it to play the skeptical CFO. Practice your responses. You'll walk in more prepared than you've ever been. And here's a new hidden stakeholder for your map: whoever controls your team's LLM and API budget. As AI-assisted workflows grow, compute costs can balloon. Map that person now.
⏱ Expected: ~11:21 (min 21/100) | SKIP — advance past
Talk track: Let me paint the picture of what happens without a Manager OS. Without cadences, problems fester — a small misunderstanding Monday becomes a conflict by Friday. Without artifacts, decisions get relitigated — you had the metrics conversation three months ago but nobody wrote it down. Without rituals, career conversations just don't happen. And without decision hygiene, whoever is loudest wins. If any of this sounds familiar, good. That's why we're building your OS today.
⏱ Expected: 11:24 (min 24/100)
Talk track: Let's start with cadences. Here are the four I recommend. Daily — for most analytics teams, you can skip the daily standup or make it async. Weekly is where the magic happens: your 1:1s and one team sync. Monthly, you zoom out and look at metrics. Quarterly, you do real planning and calibration. The key insight is that the rhythm itself creates predictability. Your team should never wonder "when will I get to talk to my manager about this?"
⏱ Expected: ~11:24 (min 24/100) | SKIP — advance past
Talk track: Before you start adding meetings, audit what you already have. I've done this exercise with dozens of managers, and every single time, we find at least two or three recurring meetings that have no owner, no agenda, and no clear output. If it fails all three — kill it or convert it to async. You just freed up hours in your calendar.
⏱ Expected: ~11:24 (min 24/100) | SKIP — advance past
Talk track: Now, if you're starting from scratch — maybe you picked DataPulse and there's literally no team yet — start small. Weekly 1:1s and one team sync. That's your entire meeting load. Don't introduce a monthly metrics review until you actually have metrics to review. And here's the most important piece: when you introduce a new cadence, explain why. "I'm adding a weekly 1:1 because I want to make sure you always have dedicated time with me" lands completely differently than "we're adding a new meeting to the calendar." People don't hate meetings. They hate bad meetings. Give them good ones, and they'll protect the cadence themselves.
⏱ Expected: 11:27 (min 27/100)
Talk track: The 1:1 is the single most important ritual in your Manager OS. Let me be blunt: if you cancel 1:1s regularly, you are failing as a manager. Here's a simple structure that works. First ten minutes are theirs — they set the agenda. This is critical. The 1:1 is their meeting, not yours. You're there to listen, unblock, and support. Middle ten minutes are yours — context from leadership, feedback you need to give, specific asks. Last ten minutes are about their career and growth. You don't need to do the career portion every week, but you should do it at least once a month. And the shared doc is non-negotiable. Both of you add items before the meeting so you're not wasting time figuring out what to talk about.
⏱ Expected: 11:27 (min 27/100)
Talk track: Let me name the anti-patterns. The status update trap — that's what Slack is for. The therapy session — help them solve the problem, don't just absorb the emotion. The canceled 1:1 — nothing says "you don't matter" faster. The monologue — if you're talking more than half the time, it's a broadcast, not a conversation. And watch the diagnostic signals: boring 1:1s mean the relationship isn't deep enough. Fire-drill 1:1s mean you have a structural problem upstream. "I don't have anything" means they don't trust the space yet — that's on you to fix.
⏱ Expected: ~11:27 (min 27/100) | SKIP — advance past
Talk track: Let me give you tactical advice for common difficult situations. First, the report who shows up and says "I don't have anything." This is almost never true — they just don't trust the space yet, or they don't know what's appropriate to bring up. Give them prompts. I like to ask "what's the most frustrating thing you've dealt with this week?" That question almost always unlocks something.
⏱ Expected: ~11:27 (min 27/100) | SKIP — advance past
Talk track: Then there's the truly silent report. Some people are just internal processors. That's okay. Try a written format — share questions in advance and let them respond in writing if that's more comfortable. I had a brilliant analyst once who said almost nothing in our 1:1s until I switched to walking meetings. Something about not making eye contact and moving through space opened her up completely. Experiment. There's no single right format.
⏱ Expected: ~11:27 (min 27/100) | SKIP — advance past
Talk track: Second common situation: the report who turns every 1:1 into a complaint session. Start by validating — their frustration is real. But then redirect toward action. "What should we do about it?" And if they're complaining about a coworker, always ask if they've talked to that person directly. Most importantly, pattern-match. If the same themes keep coming up, there might be a real structural problem underneath. Maybe the team process is broken, or the workload distribution is unfair. In that case, the complaining is actually useful signal. Your job is to separate the signal from the noise and fix the root cause.
⏱ Expected: 11:30 (min 30/100)
Talk track: Let's talk about written decision-making. There are three reasons writing beats meetings for important decisions. First, writing forces clarity — you literally cannot hand-wave when you have to put words on a page. Second, writing scales — you can share a memo with fifty people, but you can't have a productive meeting with fifty people. Third, writing creates institutional memory — six months from now, you can look back and understand why a decision was made.
⏱ Expected: 11:30 (min 30/100)
Talk track: Here is the template for a one-page decision memo. Five sections, each with a clear purpose. Context sets the scene, Options forces creative thinking, Recommendation commits to a path, Risks shows you've thought about downsides, and Decision Type tells everyone how fast to move. The key question to always ask is: "Is this a one-way door or a two-way door?" For Type 2 decisions, empower your team to just go. Don't require a memo for every small call. For Type 1 decisions, write a memo — the act of writing forces clear thinking. That single question will save your team from both reckless speed and unnecessary paralysis.
⏱ Expected: ~11:30 (min 30/100) | SKIP — advance past
Talk track: Let me show you what a real decision memo looks like. This is for MarketBridge — should we migrate from Amplitude to an in-house event tracking system? Notice the structure. The context is crisp — one paragraph, everyone understands the situation. Then three options. Not two, because two options creates a false binary. Three forces you to think creatively. Look at the pros and cons — they're specific. Not "it's expensive" but "$120K per year." Not "it takes time" but "3-4 month migration." Specificity is what separates a useful memo from a waste of paper.
⏱ Expected: ~11:30 (min 30/100) | SKIP — advance past
Talk track: The recommendation is Option C — the hybrid approach. And notice the reasoning: it de-risks the migration. That's strategic thinking. The risks section is specific about what could go wrong and how to mitigate it.
⏱ Expected: ~11:30 (min 30/100) | SKIP — advance past
Talk track: And at the bottom, this is classified as Type 2 — reversible. If it doesn't work, we go back. That classification matters because it tells everyone "we don't need to agonize over this for three weeks." Write the memo, circulate it, make the call. This is what decision hygiene looks like in practice. You'll write your own memos like this during the course.
⏱ Expected: 11:33 (min 33/100)
Talk track: The Team Charter is the first artifact in your Manager OS. Think of it as your team's constitution. It answers the fundamental questions: why do we exist, what do we do, what don't we do, and how do we operate? The template has six sections — the last one, Cadences and Rituals, connects directly to what we covered earlier about management rhythms.
⏱ Expected: 11:33 (min 33/100)
Talk track: Let me give you a real example. I once joined a team where every PM in the company felt entitled to ask for any analysis at any time. The team was drowning. We wrote a charter that said: "We serve the Growth and Monetization pods. Other teams submit requests through a quarterly intake process." That one sentence saved hundreds of hours. You're going to write your own charter in about twenty minutes. Let me show you what the template looks like.
⏱ Expected: 11:33 (min 33/100)
Talk track: The stakeholder map is your relationship radar. This 2x2 grid comes from classic project management but it's incredibly useful for analytics managers. High power, high interest — these are your bread and butter stakeholders. You manage them closely. They're the VP of Product who cares deeply about data and has budget authority. High power, low interest — these are dangerous. Think of the CFO who doesn't care about analytics until something goes wrong, and then suddenly they care a lot. You need to keep them satisfied with periodic updates so they don't blindside you. Low power, high interest — these are often your power users, the PMs who love data but don't control your budget. Keep them informed and they'll be your advocates. And low power, low interest — just monitor. Don't waste energy here, but don't lose track of them either, because power and interest levels change.
⏱ Expected: 11:33 (min 33/100)
Talk track: For each stakeholder, capture not just who they are but the exchange. What do they need from you? What do you need from them? And don't forget the hidden stakeholders — IT controls your infrastructure, Legal controls your data access, Finance controls your budget. The engineers who maintain your pipelines can make or break your productivity. Ask yourself: who could block me if I haven't built a relationship with them? Those are the people you need to map.
⏱ Expected: 11:33 (min 33/100)
Talk track: Every time I do this exercise with a new manager, they map the obvious ones — their boss, the PMs, the product team. But they miss IT, Legal, Privacy, Finance. Think about your case context — who could block you that you haven't thought of yet? The DataPulse folks should think about that one backend engineer. The MarketBridge folks should think about data engineering and compliance. The FinGuard folks should think about model risk management and internal audit. These are the relationships that, if you build them early, save you weeks of delay later.
⏱ Expected: ~11:33 (min 33/100) | SKIP — advance past
Talk track: RACI is one of those frameworks that people either love or hate. I'm going to teach you to use it well, which means using it sparingly. R is who does the work. A is who owns the outcome — and this is the critical one. There must be exactly one A per decision. If two people think they're accountable, nobody is. C is who you consult before deciding, and I is who you inform after. Let me emphasize that rule about the A — exactly one person. This is where organizations go wrong. They say "well, both the analytics manager and the PM are accountable for this metric." No. One person is accountable. The other person might be Responsible or Consulted, but there is one throat to choke, as the saying goes. That clarity is uncomfortable, but it's essential.
⏱ Expected: ~11:33 (min 33/100) | SKIP — advance past
Talk track: The biggest mistake I see is organizations that RACI everything. Don't do that. Pick the five to ten decisions that cause the most confusion or conflict, and RACI those. And watch out for the "everyone is consulted" trap. If you consult eight people before every decision, you've created consensus paralysis. Keep the C list short. Most people should be I — informed after the fact.
⏱ Expected: ~11:33 (min 33/100) | SKIP — advance past
Talk track: I've listed some good candidates here — metric definitions, dashboard approvals, stakeholder intake, model sign-off, tooling changes. These are the decisions where ambiguity causes real pain. If nobody knows who owns metric definitions, you'll have three different teams reporting three different churn numbers to the board. That's a RACI-worthy decision. But "what color should the chart be?" That does not need a RACI matrix. If you RACI trivial decisions, people will roll their eyes at the whole framework and ignore it even when it matters. Use it surgically, and it's powerful. Use it for everything, and it's bureaucracy.
⏱ Expected: 11:37 (min 37/100) | Activity brief — Team Charter; activity starts at 11:45
Talk track: Time to put what you've learned into practice. For the next 22 minutes, you'll draft a Team Charter using your case context. Open the template — let's go.
⏱ Expected: 11:37 (min 37/100) | Activity brief — Team Charter; activity starts at 11:45
Talk track: All right, let's put this into practice. Open the Team Charter template — it's in your materials. You have 22 minutes. I want you to work individually, anchored to your case context. If you chose the small startup, your charter should reflect that reality — limited resources, scrappy, moving fast. If you chose the large enterprise, think about governance, compliance, cross-team coordination.
⏱ Expected: 11:37 (min 37/100) | Activity brief — Team Charter; activity starts at 11:45
Talk track: Be specific. Don't write generic platitudes. I'll circulate and answer questions. Go.
⏱ Expected: 12:07 (min 67/100) | Activity brief — Stakeholder Map
Talk track: Second activity. You'll map your stakeholders on the Power/Interest grid. Open the template.
⏱ Expected: 12:07 (min 67/100) | Activity brief — Stakeholder Map
Talk track: Excellent work on the charters. Now flip to the Stakeholder Map template. You have 18 minutes. I want at least six stakeholders, and I want you to really think about the ones you might miss. Use the prompts on the next slide to jog your thinking.
⏱ Expected: 12:09 (min 69/100) | Activity: Map Stakeholders (18 min)
Talk track: Here are some prompts. Who funds your team? That's probably a VP or C-suite — high power. Who consumes your outputs? Product managers, maybe marketing. Who controls your data access? Who could block your work if unhappy — think Legal, Privacy, Compliance. And who do you depend on but rarely talk to? Those invisible dependencies are the ones that will bite you later. Go deep here.
⏱ Expected: 12:27 (min 87/100) | Pair Share (5 min)
Talk track: Okay, time's up on the individual work. Find a partner — and try to pair with someone who chose a different case context than you. You have five minutes each. Share your stakeholder map. Focus on who your "manage closely" stakeholders are, which stakeholder you almost left off, and where your maps differ. The stakeholders we forget are often the ones who cause us the most trouble later.
⏱ Expected: 12:32 (min 92/100) | Debrief + Transition to Block B
Talk track: Let's come back together. I want one volunteer — who found a stakeholder they almost missed? Tell us who it was and why you almost left them off. Great. Notice a pattern here — the stakeholders we overlook tend to be the ones with "quiet power." They don't show up in your daily work, but they control something critical. IT controls your infrastructure. Legal controls your data access. Finance controls your budget. Map them early, build the relationship before you need something. That's proactive management.
⏱ Expected: 12:32 (min 92/100) | Debrief + Transition to Block B
Talk track: Let's take stock of where we are. In 100 minutes, you've built the foundation of your Manager Operating System. You have a charter that defines your team's identity. You have a stakeholder map that shows your relationship landscape. You have a framework for cadences and a toolkit for making good decisions. These are artifacts you will refine and submit. Your charter and stakeholder map are due as drafts by end of today. They don't need to be perfect, but they need to be substantive.
⏱ Expected: 12:32 (min 92/100) | Debrief + Transition to Block B
Talk track: That wraps up Block A. After lunch, we move to hiring and team formation.
⏱ Expected: 12:32 (min 92/100) | Debrief + Transition to Block B
Talk track: We're breaking for lunch. When you come back at 1:30, we'll tackle hiring and team formation. Here's the key question: look at your charter — look at the purpose, the scope, the success metrics. Now ask yourself: who do I need on this team to actually deliver on that charter?
⏱ Expected: 12:32 (min 92/100) | Debrief + Transition to Block B
Talk track: What's the first role you'd hire? Is it a senior analyst who can mentor juniors? A data engineer who can fix the pipeline mess? An analytics engineer who bridges the gap? We'll dig into structured hiring, work-sample design, and interview loops. Enjoy lunch, and I'll see you back here at 1:30.