21:null 22:["$","title",null,{"children":"3 tips for data scientists to move from model to production - fast | Observable | Observable"}] 27:T49e,Track your data sources, decisions, and outcomes in the same place. You may have experimented with a new data aggregation method or created complex datasets, but without the ability to see where the data comes from and the context (your prose about what decisions you made during the process) you may find yourself answering the same questions multiple times. You can do data-tracking a few ways: **Living documents.** You could create a living document that reflects how you approached preparing your training data. This way, it’s easier for stakeholders to grasp your methods. Take a moment to get ahead of the questions and caveats that are important to your team and document them alongside your analysis. With easy-to-read and customized tooltips, make your project or model constraints clear and upfront. **Purpose-built tool.** Or, you could choose a tool that allows you to track the source of your data, as well as how it's been changed. An Observable notebook allows all team members to collaborate in one place that shows, for example, the secure use of an API. This tells everyone exactly where the data is coming from, and when it was added to the notebook. ---28:[{"id":"83594e4e104874b539995556ed10c8c7a6a95225e890e6bcc259d69641cc85616ec6b385a02f320f617d835b8bf0d00695e42c4cd2d5456d331893a4daada2b4","href":"https://static.observableusercontent.com/files/83594e4e104874b539995556ed10c8c7a6a95225e890e6bcc259d69641cc85616ec6b385a02f320f617d835b8bf0d00695e42c4cd2d5456d331893a4daada2b4","name":"DS Blog hero.png","type":"image/png","lastModified":1663935523548.77,"size":187015},{"id":"9390ddd87ed64bee02ce235bce1bb277efe4c28c02adaf6980f401b4084317e6c849e6b4934e1f5a2ce7152064e038be8b10020b9b3e8e0aaa45ed88092ed440","href":"https://static.observableusercontent.com/files/9390ddd87ed64bee02ce235bce1bb277efe4c28c02adaf6980f401b4084317e6c849e6b4934e1f5a2ce7152064e038be8b10020b9b3e8e0aaa45ed88092ed440","name":"Minimap Next v4-1.mp4","type":"video/mp4","lastModified":1663935523548.77,"size":8548278},{"id":"e63ae9d1a7f5535f720bac718bd729a9b463897202ff0893ab51791bc18158fba787ace1d7f15133bfddd19f02d3942ee4ec4ad6d72565046940dbbd189c564c","href":"https://static.observableusercontent.com/files/e63ae9d1a7f5535f720bac718bd729a9b463897202ff0893ab51791bc18158fba787ace1d7f15133bfddd19f02d3942ee4ec4ad6d72565046940dbbd189c564c","name":"Screen Shot 2022-09-22 at 2.57.28 PM.png","type":"image/png","lastModified":1663935523548.77,"size":1088327},{"id":"0410c82c7836b053e074449fe8ffb5dad37f4ec44f18cc1f75ccc4e50093f9168782eb44c73e4340f489e75a401d371ebe6c1fe6901ba61caea3fbd340ce2501","href":"https://static.observableusercontent.com/files/0410c82c7836b053e074449fe8ffb5dad37f4ec44f18cc1f75ccc4e50093f9168782eb44c73e4340f489e75a401d371ebe6c1fe6901ba61caea3fbd340ce2501","name":"Data_Table_Cell_2_Blog-2.mp4","type":"video/mp4","lastModified":1663935523548.77,"size":321682},{"id":"5916323b48bbb52271464c3d476b0a6a137edfb5230814444e54d512177bae97e21739df78d812343aacf006ad0b70d7ec5dfca6b61cca253f567e6d242a4c16","href":"https://static.observableusercontent.com/files/5916323b48bbb52271464c3d476b0a6a137edfb5230814444e54d512177bae97e21739df78d812343aacf006ad0b70d7ec5dfca6b61cca253f567e6d242a4c16","name":"DataWrangler.mp4","type":"video/mp4","lastModified":1663935523548.77,"size":1909530}] 25:{"id":"5a1ca3116aacbb2c","slug":"3-tips-for-data-scientists-to-move-from-model-to-production","title":"3 tips for data scientists to move from model to production - fast","thumbnail":"b4c889cbd847132d52d18e9f649f8cbfd82260565e2ed8f77159b93bd359665f","access_level":"public","create_time":"$D2022-09-23T12:18:43.548Z","update_time":"$D2022-11-23T15:06:52.126Z","view_time":null,"first_public_time":"$D2022-09-23T12:25:32.079Z","has_unpublished_edits":null,"fork":true,"stars":4,"starred":false,"owner":{"type":"team","id":"f35c755083683fe5","login":"observablehq","name":"Observable","avatar_url":"https://avatars-next.observableusercontent.com/424c6c1f7794371c9915a160519bdfb34cf34e50069fcd6def38a1a04d4e9702","tier":"enterprise_2024","bio":"The end-to-end solution for building and hosting better data apps, dashboards, and reports."},"version":920,"publish_version":null,"role":null,"authored":false,"authors":[{"type":"individual","id":"11aec0643a62d225","login":"ayodeleodubela","name":"Ayodele Odubela","avatar_url":"https://avatars-next.observableusercontent.com/7fe22a228e160e1c7c451030cd309d49cf23cdbb8001d4028bb2959969902efb","tier":null,"bio":""}],"members":[{"type":"individual","id":"4530d0072aa49b47","login":"erinn","name":"Erinn Wong","avatar_url":"https://avatars-next.observableusercontent.com/9119ac38e4f1c92d50b28822c47306212f06e085b7d1416e851a1c397194e2d4","tier":null,"role":"editor"},{"type":"individual","id":"074c414ad1d825f5","login":"mbostock","name":"Mike Bostock","avatar_url":"https://avatars-next.observableusercontent.com/7dbba99ced40bd5ab8013c99b1a11da61231bce321ba1148a9d16ad0f121550a","tier":null,"role":"editor"},{"type":"team","id":"f35c755083683fe5","login":"observablehq","name":"Observable","avatar_url":"https://avatars-next.observableusercontent.com/424c6c1f7794371c9915a160519bdfb34cf34e50069fcd6def38a1a04d4e9702","tier":"enterprise_2024","role":"viewer"}],"comments":[],"custom_thumbnail":true,"trashed":false,"lapsed":null,"subscribed":null,"forks":1,"collection":{"id":"01e1972879c83ae3","slug":"observable-blog","title":"Observable Blog","owner":{"id":"f35c755083683fe5","login":"observablehq","name":"Observable"}},"type":"notebook","body":{"titling":"auto","title":"3 tips for data scientists to move from model to production - fast","cells":[{"id":223,"value":"# 3 tips for data scientists to move from model to production - fast\n## _Part I_","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":684,"value":"dsBlogHero = FileAttachment(\"DS Blog hero.png\").image()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":325,"value":"Moving from proposal to project launch can be a slow process in the data science world. Whether it's a directly assigned project or a special interest, getting buy-in from leadership often involves a lot of moving parts. \n\nDefining a project’s scope, goals, and deliverables is vital for success, and Data Scientists face unique challenges in moving from model to production.","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":538,"value":"Launching a data science project may mean:\n- Pushing a model to production, \n- Approving further modeling work, or \n- Guiding your team to use model predictions to inform decision-making. \n","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":867,"value":"**In this two-part series, I identify why the process of moving from model to production is slow, and offer ways to speed it up.** There are two sides to it. Team collaboration, and the work itself. \n\nHere, In Part I, I offer ways to help team members get aligned with a model, and get collaborating. In Part II, I suggest ways to expedite working with the data. \n\n---","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":467,"value":"## Help team members trust the model\n\nProject stakeholders, a.k.a. the people on your team who have unique skill sets and expertise, may not fully understand how you developed a model. Thus, they will want to know details about the relevant data and modeling process. Without a clear understanding, getting stakeholder buy-in **quickly** is out of the question. \nHow can you help them understand the model, and thus become aligned with how the work may be done most effectively?\n","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":896,"value":"### Tip #1: track data sources, decisions, and outcomes","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":723,"value":"

\n \n

This video demonstration of the Minimap shows opening it by clicking the icon in the right sidebar; what you can learn about the different cells.

\n \n

Combine datasets easily with Data Wrangler.

\n\n
","pinned":false,"hidden":false,"mode":"html","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":704,"value":"Building ML models is usually a complex process. Ensure stakeholders understand the path from data to initial model clearly. Visualize the training data path with animated ERDs or use animate how users are segmented with decision trees. \n\nIf you’re asked to show a different time period, chart, or create an entirely different chart on the fly, do so in seconds – with confidence. \n\nYou can do this in Observable by leveraging the strength of Observable Plot plus over 30 building blocks to make infinite charts. Snippets allow you to use minimal code to create a new plot and customize it quickly. \nIn addition, the notebook provides a history of how that data has been changed. Observable’s [Data Wrangler ](https://observablehq.com/@observablehq/data-wrangler)feature provides an [interface](https://observablehq.com/@observablehq/release-notes#wrangler) to guide code composition and also display the results.\n\n---","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":909,"value":"sankey = FileAttachment(\"Screen Shot 2022-09-22 at 2.57.28 PM.png\").image()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":708,"value":"### Tip #3: Create strong visualizations for simulated predictions\nWhen you’ve got a good first version of your model, import model files and simulate predictions in the browser. ML models can be hard to illustrate, but strong visuals help stakeholders understand how a model makes a decision. \n\nDemo what-if scenarios or create a new record and see how predictions will be made based on your new data. Observable allows you to create a visualization of machine learning models that will make it clear to your whole team what types of outcomes they may expect to see. \n\nUsing a tool that allows stakeholders to view model changes on their own, without code (or with minimal code), will make them more apt to give the green-light on a project.\n","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":767,"value":"To get started with a more streamlined data science workflow, and enable your team to get to analysis faster, [try Observable for free](https://observablehq.com/solutions/data-scientists).","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":450,"value":"________________________\n\n","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":509,"value":"banner(\n \"Interested in learning more about collaborating with data?\",\n \"Sign up for free\",\n \"https://observablehq.com/product\"\n)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":506,"value":"import {banner} from \"@observablehq/brand-assets\"","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":676,"value":"import {tools} from \"@observablehq/state-of-dataviz-2021\"","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":669,"value":"import {pieChart, surveyCompareBars, surveyHistograms, longList, bigPercentage, bigPercentageWithDescription} from \"cf42a44ac2f24b7a\"\n","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":786,"value":"data_table_cell_2_blog2 = FileAttachment(\"Data_Table_Cell_2_Blog-2.mp4\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"}],"files":"$W28","stdlib":"1"}} 26:"$25:body"