# Speed Up Data-Driven Value Delina Ivanova, HelloFresh | MLOps Coffee Sessions | Episode 106 | 53:47 Hosted by Vishnu Rachakonda Source: https://www.youtube.com/watch?v=JM4EXiZzvE0 Channel: MLOps Community, now AAIF Live (https://www.youtube.com/@AAIFLive-x1r). Summarised by MLOps Talks. Page: https://mlopstalks.com/talks/speed-up-data-driven-value Published: 2022-07-05 Tags: engineering-culture, governance ## TL;DR - Delina Ivanova says prior experience in finance, consulting, operations, ethics, and strategy can make someone more effective in data work. - A data manager's job is to protect the team's time, build stakeholder trust with quick value, and guide longer-term, scalable work. - HelloFresh needs experiments and better data collection because its business combines customer choice, food production, supply chain, logistics, and perishability. ## Summary Delina Ivanova describes how she moved from finance and consulting into data leadership at HelloFresh. Her earlier work in forecasting, operations, strategy, data governance, and data ethics helps her judge business problems from several angles. She explains that data storytelling means showing the problem, the data collected, the approaches tried, the reason one worked better, and the resulting recommendation in language stakeholders understand. As a manager, she spends time with business departments, prioritizes work, protects her team's capacity, and thinks about how today's solutions will scale as the business changes. She prefers a small, highly productive team over growth for its own sake. HelloFresh gives her team difficult problems because the company combines e-commerce, food production, manufacturing, menu planning, distribution, and time-sensitive delivery. She also explains why demographic data is insufficient for understanding individual food habits, so the company needs experiments and direct data collection rather than broad assumptions. ## Key ideas ### Prior experience can make a data practitioner better at solving business problems [04:49](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=289s) Delina Ivanova began in finance, working on financial modelling, forecasts, and product analytics, then moved into consulting focused on operations and strategy. Those roles taught her how businesses operate, make decisions, and define problems objectively. At a bank, she later worked on data policy, governance, and ethics, including questions about which data should be used and how to avoid discriminatory or biased outputs. She says these experiences help her recognise when a business needs a machine learning model and when a simple report is the better answer. ### Data storytelling should make the reasoning understandable to nontechnical stakeholders [10:16](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=616s) Delina credits The Pyramid Principle with improving how she structures emails, presentations, and analyses. When presenting a data solution, she starts with the problem, explains the data selected, describes the models or approaches tried, says which worked best and why, then gives a recommendation. Stakeholders also need to understand why a scalable data product takes longer than a SQL query returning a few data points. She teaches her teams to start with the main message, then add supporting information and details for people who want them. ### Data teams solve better problems when they understand where the data comes from [13:33](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=813s) When new hires join her team, Delina walks them from the original source of the data through the customer or employee action that creates it and into the data warehouse. A customer recommendation example makes the point: if someone faces 100 recipes and cannot choose, the business problem is to recommend a smaller set they may like. Understanding that journey gives data practitioners a user perspective. It helps them connect warehouse fields and models to the real experience they are trying to improve. ### A data manager buys the team time by delivering quick value first [15:45](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=945s) Delina's team supports marketing, product, operations, procurement, and other HelloFresh departments. She assesses where the team should help, assigns value to proposed work, and weighs revenue generation against cost reduction. Revenue is difficult to attribute fully to a data solution, while cost reduction is usually easier to measure. Her initial approach was to deliver quick value and build trust with stakeholders. Once stakeholders see that the team adds value, Delina can protect time for more complex work whose benefits are less immediate. ### Managers need to plan for scale while their teams focus on the immediate work [17:54](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=1074s) Delina says a technical team will always know more about its tools than its manager. Her role is to buy people time, find the right opportunities, prioritise work, translate requirements, and think about how a solution will change as the business grows. She gives up control over every implementation detail, while still keeping enough oversight to guide the work. She may build a small prototype to test whether an idea is feasible or valuable, then hand it to a team member who can build it properly for production. ### Technical leadership includes documentation and planning for turnover [24:07](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=1447s) HelloFresh changes quickly, so a model, software product, or data solution built today may need to work with different products or business processes later. Delina wants teams to document how and why something was built, record its longer-term considerations, and explain how someone else could scale or change it. She points out that AWS has more than 600 services, so two developers can choose very different implementation paths. Documentation makes it possible for a new person to understand and continue the work when someone leaves. ### A small protected team can produce more value than a larger team doing low-value work [31:31](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=1891s) Delina prefers a small, highly productive team that operates like a focused response group. Larger data teams can attract low-value requests such as reports and repeated SQL queries, while a smaller team can protect its time for higher-impact problems. She also rejects measuring managerial success by the number of people someone manages. In her view, a manager should be judged by the effectiveness of the work delivered and by how well the team's time is allocated. Adding people does not automatically improve productivity. ### HelloFresh's data problems come from combining food, commerce, and logistics [35:32](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=2132s) Delina describes HelloFresh as a food solution company rather than only a meal-kit company. Customers choose meals through an app or website, while HelloFresh plans menus, produces boxes, manufactures some ingredients, manages facilities, and delivers perishable food. The business must show suitable recipes, hold the right raw ingredients, minimise waste, pack boxes, and keep them cold during delivery. North American geography adds complexity because facilities, suppliers, customers, and delivery routes can be far apart. This creates problems in optimisation, production planning, and just-in-time operations. ### Experiments and direct data collection are needed when broad demographic assumptions are unsafe [41:09](https://www.youtube.com/watch?v=JM4EXiZzvE0&t=2469s) Delina explains that demographic data at a postal-code level may describe a population, but it does not reveal how a particular person eats or shops. Assuming that everyone from a region or cultural background behaves the same way would be biased and potentially unethical. Internet companies can often infer preferences from large amounts of engagement data, while HelloFresh may need more traditional surveys and sampling. Experiments and manual data collection help the company learn about people it has not reached and design food products that fit individual lifestyles. ## Notable quotes - Delina Ivanova: "The number one thing I would always say is never to discount your prior experiences because those will help you be better if you are switching careers." (08:27) - Delina Ivanova: "I think as a leader of a data team and as a manager your job is to buy your team time." (17:30) - Delina Ivanova: "Your team will always know more than you." (18:25) - Delina Ivanova: "I would rather have a smaller team that's highly productive than a bigger team that is not as productive." (31:31) - Delina Ivanova: "We have to depend a lot on those traditional data collection methods." (45:05) ## Tools & references mentioned - HelloFresh - The Pyramid Principle - Barbara Minto - McKinsey - Discourses of Epictetus - Epictetus - Uber Eats - Environics - AWS - Ocado Technology - TikTok ## Who should watch - You are moving from individual contributor work into data or engineering management and need a clearer picture of how the role changes. - Your data team receives many business requests and needs a way to protect time for work with higher value. - You work with supply chain, manufacturing, food, or other physical operations where experiments and direct data collection are needed before useful models can be built. ## Related talks - [Collaboration and Strategy](https://mlopstalks.com/talks/collaboration-and-strategy) (Vin Vashishta, V Squared, 51:54) - [Building Better Data Teams](https://mlopstalks.com/talks/building-better-data-teams) (Leanne Fitzpatrick, Financial Times, 1:01:40) - [Durable Data Discovery: Making Exploratory Analysis Stick](https://mlopstalks.com/talks/durable-data-discovery-making-exploratory-analysis-stick) (James Campbell, Superconductive, 58:26) - [Data Engineering + ML + Software Engineering](https://mlopstalks.com/talks/data-engineering-ml-software-engineering) (Satish Chandra Gupta, Slang Labs, 57:05) - [The Rise of Modern Data Management](https://mlopstalks.com/talks/the-rise-of-modern-data-management) (Chad Sanderson, Gable.ai, 57:53)