📖 Building Analytics Teams

- Tags:: #📚Books , #✒️SummarizedBooks , [[Data team vision and mission]] - Author:: [[John K. Thompson]] - Liked:: #3/5 - Link:: [Building Analytics Teams | Packt (packtpub.com)](https://www.packtpub.com/product/building-analytics-teams/9781800203167) - Source date:: [[2020-06-30]] - Read date:: [[2022-05-03]] - Cover:: ![[cover_building_analytics_teams.png|100]] ## Why did I want to read it? [[Pelayo Arbues]], Head of Data Science at [[Idealista]] recommended it to me while I was growing the Data team at [[Mercadona Tech]] and complaining a lot. Now entering [[Freepik]] ([[2023-06-12]]) is a good moment to revisit it and summarize. ## What did I get out of it? *Un poco de bajona*. Most of what I thought about working with data is a bit "What you see is all there is ([[WYSIATI]])" ([[Daniel Kahneman]] would agree here probably). There are many obstacles to success with a data team, but probably the worst is some kind of alienation from the majority of the people you will work with, mostly because data is a "false friend": those obstacles are not that well understood but look like they are (compared to software for example). >Generally, **people do not enthusiastically embrace what they do not understand**. You won't find many, if any, of that generation of executives and managers who will freely admit that they do not fully understand basic computing technology (…) I am certain that there is an infinitesimally small percentage of that managerial population who understand basic descriptive statistics, let alone advanced analytics and AI (…) **Given that this is a reality for the next 7 to 10 years, what do we do about it? Patience** and clear communication are recommended for a start. Normally, I would have said concise communication, but in this case, a serious and significant amount of talking, writing, coercing, and convincing will be needed on your part. (p. 208) ^6c4cf4 >People are like analytical models, but sadly, the great majority of people lock their mental models and rarely, selectively, or not at all do they open those mental models to update their views to align with the new reality of the world. (p. 3) ### Foreword >Moreover, the book thankfully skips past the banality of building basic business intelligence (BI) solutions and focuses exclusively on advanced analytics such as data science and artificial intelligence (AI). As those of us who have been in the business long enough have come to realize, **pretty pie charts, beautiful bar charts, and dashing dashboards rarely move the needle on the business.** >…the often-speculative nature of analytics, that is, hypothesis generation and testing and retesting, can render even the most seasoned leader scrambling for ways to justify projects up-front. ### Introduction ### 1. An overview of successful and high-performing analytics teams #### We are different >You may want the organization to begin operating like a top tier firm in relation to advanced analytics and AI, but the organization may be run by fast followers, laggards, or even worse, luddites (p. 24) >Organizations and managers often fail to realize that **managing a group of analytics professionals is more similar to managing a group of creative professionals** than it is to managing a group of programmers (p. 45) #### The original sin: where to put data. Under the COO or the CEO. >…they will put the advanced analytics team under the technology function managed by the Chief Information Officer (CIO), Chief Technology Officer (CTO), or worse, **under the Chief Financial Officer (CFO). Senior management has taken the first step to failing.** (p. 48) >Advanced analytics projects are not the same as information technology projects. Installing an instance of Salesforce is relatively easy and a well-known and generally linear process. Advanced analytics projects are not, for the most part, linear. They are iterative, marked by exhilarating successes, and punctuated by dead ends, missteps, and disproved theories. (p. 49) >Placing the advanced analytics team under the accountability of the CTO is better than under the CIO if the CTO has an innovation charter within their remit. (p. 49) >At least under the CIO there is a distant history of technologists building and deploying solutions. Under the CFO, the entire mindset is process orientation and cost containment. (p. 50) >...the most successful advanced analytics teams are creative groups staffed with talented, motivated, curious people who can convert business discussions with subject matter experts into analytical applications and solutions that can drive operational change on a daily basis. (...) The best organizational home for the advanced analytics team is **reporting directly to either the Chief Operating Officer (COO) or the Chief Executive Officer (CEO)** (p. 51). ### 2. Building an analytics team #### A disruptive force Communicating with the predictable flows of other teams can be a mess: >Any time there is a creative, iterative, innovative, and unpredictable process feeding into a smooth, well planned, and reasonably static process, there are opportunities for miscommunication, misunderstandings, missed deadlines, and unplanned results. (p. 58) >The problems manifest themselves in inappropriate requests and the delegation of tasks that are **not a good use of the analytical team's time and resources**. (p. 59) And it's more than that, it's change management: >you are seeking staff members who are agents of change and can act as guides to the broader organization about how to undertake changes at a scale and scope that the organization can understand and undertake without experiencing traumatic disruption (p. 60) #### Evolved leadership is a requirement for success >I recently was in a meeting with a C-level executive in a life sciences company that exclaimed that he could build and drive analytics and a company transformation with a handful of interns and that investing in analytics was not required. If this position is the same or similar to the position espoused by the executives in the company you work in, either find another job in the firm if you want to remain in the company or find a job in another company if you want to remain working in analytics. (p. 75) #### Data illiteracy at the organization level can make you fail >"In Gartner's third annual chief data officer survey, respondents said that the second most significant roadblock to progress with data and analytics is poor data literacy, rooted in ineffective communication across a wide range of increasingly diverse stakeholders (p. 76) #### Team architecture / structure options Related to [[Data team roles]], he suggests to start with full-stack data people and over time, as you work in larger projects, move towards a hybrid config with more specialized people. ### 3. Managing and growing an analytics team >…one of the most gratifying and frustrating endeavors I have ever been involved in. (p. 94) #### Creating internal demand >You should not underestimate the concern of the people who want to approach you but feel that they cannot convey the problem well enough or that they have not packaged the discussion in the "right" way. (p. 97) >The impressions you make determines how successful the team and you will be because it determines the people who will engage with you and your team and the level of funding, effort, and resources they will dedicate to the projects that you and they jointly undertake. (p. 99) #### The rhythm of work Isn't this anti-agile? Not sure I agree with this, it really sounds as an anti-pattern. It goes against [[📖 Slack. Getting Past Burnout, Busywork, and the Myth of Total Efficiency]] too, with this idea of maximizing hours work. >After I had run multiple analytics teams, I realized that **no advanced analytics and AI project runs smoothly on a forward basis.** They all encounter some type of friction and delays. This is nothing unique to analytics projects, but the types of delays can include delays unique to advanced analytics and AI projects and delays that are common to every project. The solution that I have developed and have successfully employed in multiple roles is that **every member of the analytics team undertakes and manages their personal project portfolio of work.** (p. 105) >The team found that the gating factors were outside their control – functional team members were on vacation, the data was not ready, the business users could not agree on the objectives to be focused on, and more. The analytics team found out that by having a portfolio of work that encompasses five or six projects, they always had work available to them, they did not have lapses where they had nothing to do. (p. 254) #### Simply the best (observers) >The most talented analytics professionals are keen, unique, and in some cases, unusual observers of life (p. 116) #### People do not change in the time required >You have about a year to execute on the first cycle that results in positive and measurable change. (p. 119) #### Resistance to cultural change >People will listen and determine how much change is good for them and that they judge to be good for the organization and they will reject and work against change that exceeds this perceived threshold. (p. 120) ### 4. Leadership for analytics teams >"Show up early, it pays off." ([[Edward R. Tufte]]) As in [[An EM needs to be technical]]: >...an analytics leader needs to have a mastery of analytical techniques. (p. 126) Although: >The point being that I had no detailed understanding of the technical underpinnings or of the math of the problem being discussed, but I could bring the view of an informed novice to the discussion. (p. 147) But also: >...an analytics leader needs to possess business expertise, knowledge, and acumen. (p. 126) And: >...the ability to understand, synthesize, and communicate the value of all these respective streams of work to a wide range of people in the analytics team, across the organization, and outside... (p. 126) #### Traits of successful analytics leaders - Consistency - Passion - Curiosity & Variety - Ownership - Optimism Given that I tend to get angry, this was interesting: >I have concluded that anger is a good thing. Anger is motivating and is a driving force. What is to be avoided is expressing anger in a forceful manner. Anger scares people and anger is off-putting to many. The expression of anger is to be avoided. The feeling of anger and the experience of anger is to be embraced and controlled, channeled, and used to achieve greater goals for you and your team. (p. 132) This also resonated with me: >Due to my nature and the fact that I am curious, love to learn, get bored quickly, and have lots of energy, I was a challenge for all of my managers early in my career. I apologize to all of them for the heartburn I caused them. These traits also combined to have me labeled as a job hopper in the early stages of my professional life. **People that I interviewed with were scornful of my inability to hunker down and stay in a role for 10 years or more.** (p. 137) >There are very few paths that can lead you to have as broad and deep understanding of business, people, and the world as analytics. (p. 138) #### Bad assumptions about people >I have assumed at various times that people: >- Are much more talented and accomplished than they are. >- Have the same level of energy and drive as I do. >- Want to do the best work they can. >- Are all in and not distracted by other developments in their lives. >- Are honest and transparent in their motivations, insecurities, and personal and economic needs Most, if not all, are bad assumptions. (p. 142) >I hope for your probability of success in leading an advanced analytic and AI team that you are not the smartest person in the room. If you are, then you are either an incredible person, or you have hired poorly, or maybe a bit of both. (p. 146) ### 5. Managing executive expectations You need them and you need to make them understand: >You may want the organization to begin operating like a top tier firm in relation to advanced analytics and AI, but the organization may be run by fast followers, laggards, or even worse, luddites (…) More than likely, they will be the gating factor in how quickly the organization changes. (p. 24) >**They do not need to understand you, your team, and the work you do. You need to illustrate to them** in terms that they can understand that you and your team are a good investment that will deliver valuable results and impact at a faster and higher rate than other investments. (p. 164) > While I do not advocate spending too much of your time selling up, you will be required to engage in selective and strategic selling and sales activities to convince the upper echelons (p. 164) ^0d7410 So much so that: >...it is a critical portion of your primary work duties. (p. 166) #### Not many of us out there... >You need to assume that the executives you are engaging with have little to no knowledge of any part of this environment, and at the same time, they have very little interest in the topic (…) if I am confused and you are presenting or representing something that I cannot understand and the choice is between me being unable to understand the topic or you being wrong or incapable, then the second option is an easy choice (p. 169) >The number of people who truly understand the multiple factors that need to come together in a specific way to deliver positive operational improvements from analytics is **a very small number**. (p. 174) #### Defining goals with subject matter experts is HARD >The definition of objectives and goals has always been challenging when collaborating with subject matter experts from the functional areas of any business. (p. 177) #### Executives as elephants and squirrels >\[Executives\] are typically split into two groups on the ability to retain and remember facts over time. Elephants are easy to work with. You come to an agreement and you work toward that agreement. Squirrels are harder. You never know what they remember, how they remember it, or how much they will change the memory over time. Be alert, be humble, and ask lots of questions of the squirrels. (p. 182) #### A sense of urgency As in consistent value delivery, in terms of reputation: >Velocity in delivery is important for you and your team and in the eyes of the executive team. (p. 187) ### 6. Ensuring engagement with business professionals #### Change management is what we are after Superficial vs. true change: >Again, we in analytics are different: we like complexity, we like change, and we do not only like change, we seek to change things in a proactive and engaged manner. Most people do not have that mental orientation or personal constitution; most middle managers seek to actively avoid change. While it may seem, on the surface, that organizations are in a constant state of change, corporate cultures and the majority of managers and executives work proactively against change. It is true that superficial change is nearly constant. (p. 201) So ideally you would have help: >Although we argue that engaging the data science team to manage and drive this change is a waste of time and money, the organization needs to have another group, like a project management or change management group, that is chartered with driving the process and operational changes required to realize the value of the insights and intelligence generated by the advanced analytics team. (p. 202) It all comes from above: >Aim high for analytics champions. Executive sponsorship is vital to this level of organizational change, and the best champion sits in the corner office. (p. 211) >Encourage leaders to model examples. In meetings, for example, leaders should demonstrate the importance of analytics by asking for data points to back up business decisions. (p. 211) >It's every manager's job here to understand the nature of statistical forecasting and Monte Carlo simulation. (p. 212) #### Linear and non-linear thinking >Working with linear thinkers—and you will have to, because they are the majority of people who are successful in business today—is not as hard as it sounds. You simply need to spend much more time setting expectations for and with them. Think about every date that you can possibly define in a timeline of a project and multiply that by a factor. (p. 213) >The linear thinkers will want to know the date and time when the eureka moment will occur. The process doesn't work that way. Some theories do not pan out, others lead to completely different directions, and others are spot on and take the team forward to new levels of understanding and improved performance. (p. 268) #### Not big data but lots of small data >There will be functional managers and executives who will argue for using limited data sources, possibly in large volumes, to analyze the phenomena that you are seeking to understand. (...) you as the analytics leader need to advocate for the use of a wide range of disparate internal and external data to develop solutions that are differentiated from the competition and that provide a lasting basis for competitive advantage. (p. 218) #### Value realization On [[Measuring a data team impact]]. You need visibility, and... >With visibility comes accountability and responsibility (p. 222) Because people may not use the results: >If the project findings supported what the sales and marketing leader believed and had planned to do, the results were re embraced and used in building the case or executing the plan. If the analytical findings indicated that there was a better way to execute the plan, then the results were ignored, and the plan was executed as the sales and marketing leader intended. This happens on a regular basis in numerous organizations. The question is, what do you do? (p. 221) >Visibility to the management of the analytics team and to the functional team is best achieved through routine status reports and personal meetings. Visibility to the company is best achieved through company bulletins, newsletters, company app updates, social media channels, and other communication vehicles (p. 221) >The projects that you and your team are undertaking **should all conclude with a phase where organizational change is undertaken or where analytical applications or models are put into production usage.** (p. 222) In any case, you were hired for this: >Remember, you are the expert and they are the novices. It is your role and responsibility to teach them and communicate with them and convince them that they should come on this journey with you and your team. At first, this may seem to be a waste of your time, but trust me, it is not. You need them and they need you, even if they do not know it yet. Be patient and describe the challenge, solution, and benefit in as many ways as required for them to buy in and join the journey (p. 224) ### 7. Selecting winning projects > Winning projects are those analytics projects that are widely understood (p. 227) Who decides? Well... >At some level, decisions and commitments must be made, and projects of varying types must be compared and chosen between, but these decisions should be made with the most detailed information and the deepest level of understanding of all the possible options. This is a problem for the company, for the executives and managers, and for you. **In many cases, the decision makers are kicking the can down the road**. (p. 230) >…the finance organization or the portfolio management function in the organization does not know how to calculate the relative value of the possible projects in the analytics portfolio very well or with the needed level of accuracy. (p. 234) >It is paradoxical that the analytics team is not very good at defining the details of the project delivery, cost, timelines, and relative value of analytics projects (…) the clarity of the size, scope, and scale of the analytics effort comes from you, the analytics leadership, and the analytics team. (p. 234) So it ends up being you, the analytics leader, but that's not good: >...the head of analytics is making all the decisions, it is not a sustainable or defensible situation. (...) the role of the analytics leader and team is to develop, present, and clarify detailed project plans and proposals that can be easily and fairly compared with competing projects, and it is the duty of the project review board or executives or other managers to provide sufficient review, governance, and guidance to all project managers on why projects are being selected and undertaken and why others are being deferred or rejected. (p. 236) Careful with subject matter experts too! >"They are subject matter experts, and they should know what good looks like!" Well, perhaps not. (p. 252) ### 8. Operationalizing analytics - How to move from projects to production Very important, because many stakeholders are going to try to push work that doesn't have any action behind. In the line of [[A method for measuring analytical work]]: >We are not undertaking data and analytics projects to indulge our intellectual curiosity; we are doing so to drive change and improvement (p. 258) ^4859ae You can also be a victim of your own success: you may change the work of some other people (such as when automating a forecast that was done by hand by some people) and that needs to be planned for. And unrealistic expectations: >This type of discovery-oriented analytical work is typically driven by the art of the possible. You can plan projects, but discoveries rarely show up when planned. Your team may find interesting insights a few days into the work or it may take a year or more to find an insight that is truly game changing for the business. (p. 267) Another typical problem described here has to do with headcount, data debt, and [[Data team roles]]. If we don't manage debt properly... >It does not take long for every data scientist to become a data management operator with no time to undertake new projects. (p. 273) ### 9. Managing the new analytical ecosystem You need to focus on what moves the needle, but many people will distract you: >It sounds a little trite and perhaps a bit contrived when it is written out in plain terms, but it is important to remember that analytics leadership and staff can sacrifice and ignore the needs of other groups in the service of focus and efficiency. This can be interpreted as indifference and arrogance by the other functional managers in a company. (p. 291) ### 10. The future of analytics - What will we see next? To me, the only interesting piece is a reminder that selling some of the company data can be an additional stream of revenue. ### Other notes #### On [[outsourcing]]. Only for the easy things: >you should consider outsourcing certain projects to competent, capable, and proven third parties. The projects to be considered are those that others in your industry have completed and are now considered tables stakes. (p. 9) #### Do you really need a budget? >You and your team can be very effective by using open source software for a substantial portion of your work (p. 216) #### On [[✍️ GPT-3 me va a quitar el trabajo, pero yo tengo que estar entrenando algoritmia de bajo nivel]] The section "Many jobs will never be changed by AI" (p. 24) did not age well. He refers to [[Polanyi's Paradox]]: "we can know more than we can tell, we shouldn't assume that technology can replicate the function of human knowledge itself" as an argument (but you can argue that recent advances allow AI to perform tasks by infering tacit knowledge on its own). Also: >AI will not create deeply personal experiences. (p. 25) I wonder if the extreme customization with [[LLMs]] won't allow you to reach such experiences. #### Others - There is also a whole section on [[neurodiversity]]! (p. 68) - A discussion on a C-level role: Chief Analytics Officer (p. 148)