Build your own custom predictions using Einstein Prediction Builder, Pratyush Kumar
Report
Share
•0 likes•8 views
Build your own custom predictions using Einstein Prediction Builder, Pratyush Kumar
•0 likes•8 views
Report
Share
Download to read offline
If you’re a trailblazer working with an organisation tired of the guesswork? Join this technical walkthrough to make your Salesforce users more predictive and proactive using Einstein Prediction Builder. You’ll learn what’s under the hood so that you can create custom AI models on any Salesforce object to predict business outcomes, and can create your own predictions for your customer’s business to power a workflow and make users more efficient and smarter, all using just point and click.
Recommended
More Related Content
More from CzechDreamin
More from CzechDreamin(20)
Recently uploaded
Recently uploaded(20)
Build your own custom predictions using Einstein Prediction Builder, Pratyush Kumar
- 1. Build Your Own Custom Predictions Using Einstein Prediction Builder by Pratyush Kumar
- 2. #CD22 Salesforce Data Architect @PMI ● Certified System and Application Architect, 14x certified ● 10+ years within Salesforce eco-space ● Based out of London ● Aspiring ultra-marathon runner ● Huge cricket fan and sports enthusiast Twitter: @Im_Pratyush LinkedIn: https://www.linkedin.com/in/pratyush-kumar-salesforce/ Trailblazer: https://trailblazer.me/id/salesforcepratyush Who am I?
- 3. #CD22 Forward Looking Statement Statement under the Private Securities Litigation Reform Act of 1995 This presentation may contain forward-looking statements about the company's financial and operating results, which may include expected GAAP and non-GAAP financial and other operating and non-operating results, including revenue, net income, diluted earnings per share, operating cash flow growth, operating margin improvement, expected revenue growth, expected current remaining performance obligation growth, expected tax rates, stock-based compensation expenses, amortization of purchased intangibles, shares outstanding, market growth, environmental, social and governance goals and expected capital allocation, including mergers and acquisitions, capital expenditures and other investments. The achievement or success of the matters covered by such forward-looking statements involves risks, uncertainties and assumptions. If any such risks or uncertainties materialize or if any of the assumptions prove incorrect, the company’s results could differ materially from the results expressed or implied by the forward-looking statements it makes. The risks and uncertainties referred to above include -- but are not limited to -- risks associated with the effect of general economic and market conditions; the impact of geopolitical events, natural disasters and actual or threatened public health emergencies, such as the ongoing Coronavirus pandemic; the impact of foreign currency exchange rate and interest rate fluctuations on our results; our business strategy and our plan to build our business, including our strategy to be the leading provider of enterprise cloud computing applications and platforms; the pace of change and innovation in enterprise cloud computing services; the seasonal nature of our sales cycles; the competitive nature of the market in which we participate; our international expansion strategy; the demands on our personnel and infrastructure resulting from significant growth in our customer base and operations, including as a result of acquisitions; our service performance and security, including the resources and costs required to avoid unanticipated downtime and prevent, detect and remediate potential security breaches; the expenses associated with our data centers and third-party infrastructure providers; additional data center capacity; real estate and office facilities space; our operating results and cash flows; new services and product features, including any efforts to expand our services beyond the CRM market; our strategy of acquiring or making investments in complementary businesses, joint ventures, services, technologies and intellectual property rights; the performance and fair value of our investments in complementary businesses through our strategic investment portfolio; our ability to realize the benefits from strategic partnerships, joint ventures and investments; the impact of future gains or losses from our strategic investment portfolio, including gains or losses from overall market conditions that may affect the publicly traded companies within our strategic investment portfolio; our ability to execute our business plans; our ability to successfully integrate acquired businesses and technologies; our ability to continue to grow unearned revenue and remaining performance obligation; our ability to protect our intellectual property rights; our ability to develop our brands; our reliance on third-party hardware, software and platform providers; our dependency on the development and maintenance of the infrastructure of the Internet; the effect of evolving domestic and foreign government regulations, including those related to the provision of services on the Internet, those related to accessing the Internet, and those addressing data privacy, cross-border data transfers and import and export controls; the valuation of our deferred tax assets and the release of related valuation allowances; the potential availability of additional tax assets in the future; the impact of new accounting pronouncements and tax laws; uncertainties affecting our ability to estimate our tax rate; uncertainties regarding our tax obligations in connection with potential jurisdictional transfers of intellectual property, including the tax rate, the timing of the transfer and the value of such transferred intellectual property; the impact of expensing stock options and other equity awards; the sufficiency of our capital resources; factors related to our outstanding debt, revolving credit facility and loan associated with 50 Fremont; compliance with our debt covenants and lease obligations; current and potential litigation involving us; and the impact of climate change. Further information on these and other factors that could affect the company’s financial results is included in the reports on Forms 10-K, 10-Q and 8-K and in other filings it makes with the Securities and Exchange Commission from time to time. These documents are available on the SEC Filings section of the Investor Information section of the company’s website at www.salesforce.com/investor. Salesforce.com, inc. assumes no obligation and does not intend to update these forward-looking statements, except as required by law.
- 4. #CD22 Introduction How to build your own custom predictions? Demo Consideration & Best Practices Resources Presentation Outline
- 5. #CD22 What is Salesforce Einstein? Einstein is an application that uses AI, which makes Salesforce smarter. It is built on the robust and flexible Salesforce security architecture Discovers hidden insights and patterns your data Predicts outcomes for every function & industry Recommends the best action, offer or engagement Automates business processes and workflows Introduction DESCRIPTIVE What happened and why? PREDICTIVE What’s going to happen? PRESCRIPTIVE What should I do about it?
- 6. #CD22 Allows to make custom predictions related to any custom or standard objects, helping the user to work smarter by focussing time on the right tasks ● Point, Click and Predict - No code required ● No need for any tedious task of ETLing the data ● Predictions can be made based on the data given and continue to score as your dataset changes ● Learns from examples in the past to make predictions ● Using projections it helps you work smarter by focusing your time on the right tasks ● Predicts the answers to binary questions, i.e., Yes/No and numerical questions Focus on Specific, High Priority, Business Outcomes What is Einstein Prediction Builder?
- 7. #CD22 Prediction Data Terminologies Dataset - the set of records on selected objects Segment - Create a subset of your dataset to focus your prediction on Example Set - Einstein identifies patterns in data from these records to create predictions for other records Prediction Set - The set of records that Einstein predicts values for How to build your own Custom Predictions?
- 8. #CD22 Steps to configure How to build your own Custom Predictions? Choose Object Choose what to Predict Choose what information to use Save Prediction Plan Preview Build
- 9. #CD22 Plan Your Predictions ● Define a use case ○ What is the problem my company is facing? ○ What type of prediction will help? ○ Can this prediction be phrased as a yes/no or numeric question? ○ How will I use this Prediction? ○ How do I measure if this Prediction is successful? ● Choose Object ● Field and type of field you want to predict values for: ○ Binary Predictions ■ Leverage historic information to answer a Yes/No question ○ Numeric Predictions (beta) ■ Leverage historic information to predict a number Data Considerations If you can track it, you can predict it. ● Choose a repeatable workflow that you can report on to build your prediction ● Take the time to identify the data that will enable the best quality signal for your prediction. ● More data connected to Salesforce means more options with predictions, and better quality data means better- quality predictions.
- 10. #CD22 Predicting the likelihood of closing a deal? Demo
- 11. #CD22 Considerations Considerations & Best Practices ● Works with all custom objects, and supports limited standard objects. ● Your dataset needs to have enough records in order to build a successful prediction. ● Einstein Prediction Builder can make predictions for the Checkbox, Specially constructed formula fields and Numeric. ● Predictions created in production aren’t copied to sandbox orgs. ● Predictions can’t be edited when in Enabled or Pending status. ● Predictions are packageable only in managed packages. ● Be aware of the Licenses and their Limits ○ Try v/s Paid Einstein version ○ Einstein Predictions license v/s Einstein Analytics Plus license ● Be aware of what’s GA vs Pilot/Beta features.
- 12. #CD22 Best Practices Considerations & Best Practices ● Some key points: ○ It is all about Data ○ Pick the right tool for the Job ○ Start with the End in mind ○ Embed Predictions and Actionable Insights within Salesforce ○ Machine Learning projects are different ● Use “Data Checker” while selecting an object to predict ● Improve your Prediction’s Quality (Yes/No Predictions) ○ Collect More Data ○ Try different data segmentation ○ Example Set needs to have right records ○ Review the fields which are excluded and included in your prediction ○ Avoid Hindsight Bias Descriptive Diagnostic Predictive Prescriptive What Happened? Why It Happened? What will Happen? What to Do about it?
- 13. #CD22 1. Get closer to your customers with Salesforce Einstein. 2. Get Smart with Salesforce Einstein - Trailhead Modules 3. Training | Article View 4. Einstein Prediction Builder and Discovery Best Practices 5. The Salesforce Einstein Team – Medium 6. Tableau CRM Developer Org 7. Einstein Prediction Builder * For your own practice - sign up for Tableau CRM Developer Edition org developer.salesforce.com/promotions/orgs/analytics-de Resources
- 14. #CD22 Do you have any questions? We hope you learned something new.
- 15. Thank you! #CD22 Pratyush Kumar salesforcepratyush @Im_Pratyush
Editor's Notes
- Ahoj :) Thanks to the organisers - Czech dreamin, and their sponsors for making it possible to present in person. Thank you all who all are here, and it’s a pleasure to be here.
- Intro -
- Because forward looking statements are inherently subject to risks and uncertainty, reminders that you should make any purchasing decisions or investment decisions based on products that are currently commercially available.
- CRM has become massively data-led in the last few years. Data-based insights are a critical component of strategic decision-making, but transforming data into actionable insights can be challenging. This brings the role of predictive analytics to the forefront. What is Salesforce Einstein? Salesforce Einstein gives sales, service marketing a complete and up-to-date view of customers and sales prospects, driven by AI. Built on the Robust & flexible Salesforce security architecture. ------------- With Einstein Platform, customers can build AI-powered assistants and apps by leveraging: Discovers using Einstein Discovery - for e.g. to understand what trends are driving your business or what marketing campaigns are performing Predicts using Einstein Prediction Builder - for e.g. to predict the number of opportunities the company is going to close in next quarter Recommends using Einstein Next Best Action - for e,g, the actions you should take to prevent your customers from leaving your company And, lastly Automation, when customers think about AI, they think about Automation.. Having said that you can’t automate everything.. There are restriction to that but Think about a repetitive path which is what you can automate using AI and Einstein. We build AI-powered assistants and apps by leveraging Einstein Bots, Einstein Voice, Einstein Vision and Einstein Language for Automation ---------------------------------- Provides a Full spectrum of AI with descriptive, Predictive and Prescriptive capabilities Here we’ll be exploring how Einstein Prediction Builder operates.
- Einstein Prediction Builder: allows admins to create custom predictions on standard or custom objects in Salesforce with clicks, not code, using a simple wizard, which we’ll see in a minute. Predictions can be made based on the data given and continue to score as your dataset changes. It learns from examples in the past to make predictions. There is no need for any tedious task of ETLing the data. It is just a few clicks process that allows you to make custom predictions related to any custom or standard objects. Using projections it helps you work smarter by focusing your time on the right tasks. It predicts the answers to binary questions, i.e., Yes/No and numerical questions. Examples: Maximise Win Rate Maximise CSAT Score MAximise Cross Sell Einstein Forecasting Lead and Opportunity Scoring Territory Scoring
- When using Einstein Prediction Builder, it’s helpful to understand some terms we use for sets of data. One of the first steps in building a prediction is to select the object that contains the field you’re predicting. The set of records on that object is your overall dataset. For example, you want to predict your prospects’ likelihood to attend an event. So your prediction is based on the Lead object. You have 600 leads. That’s your dataset. https://help.salesforce.com/articleView?id=sf.custom_ai_prediction_builder_filters.htm&type=5 After selecting an object, you have the option to focus your prediction on a segment of your data. You use condition logic to filter your dataset, which creates a subset of that data. For example, you’re interested in only your leads from non-banking industries. Your segment has 501 leads. Then you tell Einstein which records in your segment (or in your whole dataset, if you’re not focusing on a segment) to use as examples for building a prediction. This is your example set. Einstein identifies patterns in data from these records to create predictions for other records. Your example set must include leads who have and haven’t attended a previous event. So you set up your conditions to include only leads that are being worked on. Using working leads rather than new leads makes your example set more likely to include a variety of values for prior event attendance. The set of records that Einstein predicts values for is your prediction set. Your prediction set consists of the records in your segment (or dataset if you don’t have a segment) that Einstein provides prediction results (scores) for. The default selection is to score records that aren’t in the example set, which in this case means new leads. ------------ As you build a prediction, use Data Checker to make sure you have enough records in your segment, example set, and prediction set to build a prediction.
- Before you start building a prediction, take some time to think about what you want to predict and how it translates to Salesforce objects and fields. In some cases, you might need to create a custom field to get your data in a format that Einstein can predict. When planning your prediction, you first decide what field you want to predict. How you set up your prediction depends on the data type of the field you’re predicting. Decide what field and type of field you want to predict values for. You can predict numeric fields (beta) You can predict custom formula fields and checkboxes Get familiar with datasets, segments, example sets, and prediction sets. Build the prediction. Einstein Prediction Builder guides you through all the steps. No coding needed! Take time to review your prediction, check the score, and revisit it if required.
- Before you start building a prediction, take some time to think about what you want to predict and how it translates to Salesforce objects and fields. In some cases, you might need to create a custom field to get your data in a format that Einstein can predict. Make sure you follow: Focus on Specific, High Priority, Business Outcomes https://www.salesforce.com/content/dam/web/en_us/www/documents/e-books/analytics/sfdc-predictions.pdf You can predict numeric fields (beta) like revenue, year-over-year growth, cost, tax, and salary. You can predict custom formula fields and checkboxes, such as the likelihood that a deal closes, or the likelihood that year-over-year growth is greater than last year. Choose a repeatable workflow that you can report on to build your prediction. Then take the time to identify the data that will enable the best quality signal for your prediction. More data connected to Salesforce means more options with predictions, and better quality data means better-quality predictions Few Examples of what to expect in terms of predictions: Whether the next invoice will get paid or not? Number of hours to close high priority cases? The likelihood of closing a deal?
- Use Case and Demo: Predict the likelihood of a project missing its delivery date so that you can better allocate resources and reduce project costs. A very basic example of predicting the likelihood of getting an invoice paid. Or winning a deal that you’re trying to close? Predicting the likelihood of closing a deal? Or in other words: Predicting the number of opportunities the company is going to close? Predicting a true/false field is a binary classification problem. For this type of prediction, Einstein tests these model types: Random Forest Logistic Regression Predicting a number field (beta) is a regression problem. For this type of prediction, Einstein tests these model types: Random Forest Linear Regression
- https://help.salesforce.com/articleView?id=sf.custom_ai_prediction_builder_considerations.htm&type=5 Licenses and Limits: The Try Einstein version of Einstein Prediction Builder lets you build up to 10 predictions and enable up to one of them at a time. You also have access to most features. The paid version of Einstein Prediction Builder is available with the Einstein Predictions and Einstein Analytics Plus licenses. The Einstein Predictions license lets you build up to 20 predictions and enable up to 10 of them at a time. The Einstein Analytics Plus license lets you build up to 45 predictions and enable up to 35 of them at a time. It also gives you access to display the Einstein Predictions component on records, which shows top predictors at the record level. Pilot/Beta As a beta feature, numeric field prediction is a preview and isn’t part of the “Services” under your master subscription agreement with Salesforce Scoring Frequency is a pilot program Use this feature at your sole discretion, and make your purchase decisions only on the basis of generally available products and features. Salesforce doesn’t guarantee general availability of this feature within any particular time frame or at all, and we can discontinue it at any time. This feature is for evaluation purposes only, not for production use. It’s offered as is and isn’t supported, and Salesforce has no liability for any harm or damage arising out of or in connection with it. All restrictions, Salesforce reservation of rights, obligations concerning the Services, and terms for related Non-Salesforce Applications and Content apply equally to your use of this feature. If you’ve joined the Scoring Frequency pilot feature, you can decide how often records are scored for your prediction. Talk to your Salesforce Account Executive if you’d like to be nominated to the pilot. Select Hourly to have Einstein check records hourly, and score any records that have changed. Select When Records Change to have Einstein continuously check records and provide new scores whenever they change. For e.g. - “Einstein Account Scoring” is part of a pilot program - Predict the likelihood of an account being receptive to your brand, messages, and products, so that you can identify and prioritize the right accounts and sell more.
- https://help.salesforce.com/articleView?id=sf.custom_ai_prediction_builder_considerations.htm&type=5 First, notice whether the Prediction Quality title says “(Estimated)” or not. If so, it means that either you haven’t enabled your prediction yet, or it’s enabled but not yet running on live data. Either way, the quality score is estimated, based on a subset of your example data. It’s typical for the quality to go down after the prediction starts running on live data. For example, if your prediction quality is Great but estimated, check the scorecard again in about a month to see how it’s performing with live data. Collect more data. Certainly not all quality problems are due to the amount of data, but in some cases you can improve the quality simply by waiting until your dataset is more full. Try changing the way your data is segmented. For example, make sure your segment includes only records that are relevant to your prediction. If you have two distinct datasets that would each benefit from having its own prediction, create a separate prediction for each segment. Make sure your example set contains the right records for both yes and no examples. Look for possible sources of bias in your data that might influence your prediction. Make sure you’re not excluding any fields that could provide additional, useful data. If your prediction uses formulas, check to make sure they’re correct. If your prediction quality is Too High when estimated (before it’s running on live data), it’s likely too good to be true. Review the fields included in your prediction. Do any of them look wrong, or contain data that couldn’t have been known until after the event being predicted? If so, exclude those fields. This type of problem is called hindsight bias. Also look over the Top Predictors on the scorecard. Do they make sense? If any of them look wrong, consider excluding those fields. Predicting a true/false field is a binary classification problem. For this type of prediction, Einstein tests these model types: Random Forest Logistic Regression Predicting a number field (beta) is a regression problem. For this type of prediction, Einstein tests these model types: Random Forest Linear Regression
- https://www.salesforce.com/content/dam/web/en_us/www/documents/e-books/analytics/sfdc-predictions.pdf With Salesforce Einstein, you can rapidly progress from prototype to pilot to scale. Start building now. Track ROI and performance along the way with A/B testing through pilot rollouts and see the value yourself.
- Last but not the least, thanks to the organisers, and everyone attending this session. And, thanks again for listening to me, hope you all learned something new in this session.