Winter ‘23 Release is here!
CRM Analytics Winter ‘23 release is here! I chose my top 10 features which I believe will make CRM Analytics users’ day to day way easier and more productive! From new UI features in Data Prep, to new embeddable dashboards in the Customer Insights App, to Einstein Discovery improvements! If you are not sure if your org has been updated to the new release leverage status.salesforce.com, enter your instance which can be found under Company Information in Salesforce and check for maintenance.
Now let’s talk about some awesome features!
When you are building a recipe two of the tricky pieces are: 1. Identifying the Grain and 2. Knowing the Relationships between the objects you are joining and selecting the appropriate keys. To address these I used recommended 2 tricks:
Checking Schema Builder in Salesforce to visualize your data model and the relationships between your objects
Using Dataset Builder within Dataflows to identify related objects and fields
Now, similar to dataset builder- in the new Recipe interface, you can see now all the related objects per the Salesforce data model and columns within Data Prep and click on Join..
The join keys will be preconfigured AND you will be able to see all the possible combinations for the Join keys in case you wish to replace the selected join key.
I have included a short video below describing the new functionality!
2. Self-Join Nodes in Data Prep
This feature is also extremely helpful if you have an object which has a foreign key relationship to itself. For example, the account object has a parent relationship using the Parent Account id. Previously we had other workarounds like adding the Account multiple times to the recipe.
Now we can simply join the account on itself using the Parent Account id on the left side and the Account id on the right side.
What this will allow us to do is augment fields from our single Join 0 lookup instead of having multiple nodes.
3. Copy paste nodes as well as keyboard shortcuts
Another great data prep feature is you can now replicate node configurations with the Data Prep new copy, cut, and paste functionality in recipes. You can copy and paste within a single recipe or across multiple ones. Previously, if you wanted to reuse a formula or join key definitions, you needed to create a node and enter the information again, which made it extremely time consuming.
Additionally there are some new data prep keyboard shortcuts you’ll find very handy!
4. Enhancement to the Customer Insights App
I’ve written other posts about the Customer Insights App! If you have Customer Insights Analytics 1.0 you should be able to upgrade now to 1.5! Now here are some of the reasons I absolutely love this app and recommend it to a lot of people:
1) It creates recipes instead of dataflows which are WAY easier to edit and update. You can also clone a recipe by downloading JSON and re-uploading the JSON and making whichever changes you want.
2) It leverages components (which are essentially mini dashboards within a dashboard) which means they are easily embeddable.
3) All the visualizations were thought of carefully with the idea of embedding them into Salesforce records.
Now, in order to get advantage of the new features which I’ll describe below, you need to Open the app and Upgrade to the new version. Once you select to upgrade, you will get a glimpse of the release notes for Customer Insights 1.5 and move forward with the upgrade!
So what are the updates?
1) You can now customize security by controlling row level security for your data as well as selecting custom amounts from the updated Customer Insights configuration wizard.. When it comes to security you have 3 options: 1) allow users to see all data, 2) their own data, or 3) the data of users with the same role and their subordinates. Previously, security predicates were applied automatically, now you have a saying in the security similar to other templated apps!
2) Try the updated Customer Insights Case embedded dashboard, which makes it easy to get details about your caseload whenever you need it. The nice thing about this embedded dashboards is they are leveraging components which makes it super easy to embed in a Case record page in Salesforce.
Now let’s talk about some improvements for Einstein Discovery…
Top 5 Features for Einstein Discovery
6. Improvements to Einstein Discovery: Impact alerts on live data!
You can understand if predictive models are introducing bias, by
Monitoring your live data for any disparate impact issues that may be introduced over time
Get notified about disparate impact issues in live metrics in the Model Manager in Einstein Discovery
Take action via a direct link back to the story editor to review alerts and data issues
7. Bring your Scikit-learn Model to Salesforce
If you are building custom data science models in Python you can now bring them to Einstein. You can now upload and deploy Scikit-learn models in addition to TensorFlow models.
The following files should be bundled in a zip:
saved_model.pkl
data_processor.py
validation.csv
8. Importance vs correlation & Ethical AI alerts
You will notice now, once you run your story you’ll see Ethical AI Alerts: which basically mean if 2 variables are sensitive fields and if there’s any potential to be bias.
Feature importance rather than correlation- Feature important or Variable importance whatever you want to call it, tells you the specific impact each of these specific variables has in the model (we still allow you to see correlation if you edit the story)
9. Live Predictions with Snowflake
When you build a model, you can now tell it if you have resident data in Snowflake!
Live connections to snowflake → deploy mode > Link to supplemental dataset connected to an External ID
Variables can stay in Snowflake and be included in your model
More resources: CRM Analytics Direct Data for Snowflake Tips and Considerations
10. High volume predictions
Now we can process high volume predictions and see the story for those predictions. With this release, you can now process 5 million predictions per 25 hour period (vs ½ a million before!). More efficient and faster data writeback avoids row locks. Additionally, you have 30 days of prediction history accessible via API
NOTE: Default on for new orgs, opt in for existing orgs via Analytics Settings (screenshot below)
