v1.0.0 to v1.0.1

87 added, 111 removed. Audit A to A.

---

name: einstein-discovery-setup

description: "Use this skill when an admin needs to create an Einstein Discovery story in CRM Analytics Studio, configure prediction definitions, deploy writeback fields, set up what-if analysis, or manage model refresh and activation. Trigger keywords: Einstein Discovery story, prediction definition, writeback field, CRM Analytics Studio, model refresh, what-if analysis, bulk scoring, prediction field, 1OR prefix. NOT for the Einstein Discovery Flow action or model-health monitoring in Model Manager — use admin/einstein-discovery-deployment. NOT for Einstein Prediction Builder, a separate product needing no CRM Analytics license — use agentforce/einstein-prediction-builder. NOT for programmatic scoring through the Connect REST API — use agentforce/einstein-discovery-development."

category: admin

salesforce-version: "Spring '25+"

well-architected-pillars:

- Security

- Reliability

- Operational Excellence

triggers:

- "How do I create an Einstein Discovery story in CRM Analytics Studio?"

- "Setting up prediction fields for Einstein Discovery on Opportunity records"

- "Enabling Einstein Discovery recommendations on records using the admin wizard"

- "How do I deploy an Einstein Discovery model so scores appear on page layouts?"

- "Einstein Discovery writeback field is not showing up in reports or page layouts"

+ - "move an Einstein Discovery prediction definition from sandbox to production"

+ - "segment one prediction definition into separate models per region"

tags:

- einstein-discovery-setup

- einstein-discovery

- crm-analytics

- prediction-definition

- writeback-field

- story-creation

- model-refresh

- what-if-analysis

inputs:

- "CRM Analytics license confirmed as provisioned in the org"

- "Target Salesforce object and outcome field for the story (e.g., Opportunity.IsClosed)"

- "Whether the use case requires Insights-only or Insights+Predictions (writeback to record fields)"

- "Confirmation that admin has CRM Analytics Admin permission set or equivalent"

- "List of explanatory variables (fields) available on the target object or related objects"

outputs:

- "Guidance for completing the three-step story creation wizard in CRM Analytics Studio"

- "Prediction definition configuration (1OR prefix, enabled status, target object mapping)"

- "Writeback field setup instructions including field-level security assignment steps"

- "Model refresh schedule and manual activation procedure"

- "What-if analysis configuration guidance for surfacing recommendations on record pages"

dependencies: []

- version: 1.0.0

+ version: 1.0.1

author: Pranav Nagrecha

- updated: 2026-04-13

+ updated: 2026-10-03

runtime_orphan: true

runtime_orphan_reason: "No run-time agent covers CRM Analytics / Einstein Discovery. This skill was previously listed in audit-router's Mandatory Reads, but no audit-router classifier routes to it and report_dashboard's own scope excludes CRM Analytics migration, so the citation was decorative rather than load-bearing. Removed 2026-08-14 rather than left as a citation an agent never honoured. Re-wire when a CRM Analytics agent exists."

---

# Einstein Discovery Setup

- Use this skill when an admin needs to configure Einstein Discovery through the CRM Analytics Studio wizard interface — creating stories, deploying prediction definitions, setting up writeback fields on Salesforce objects, enabling what-if analysis for recommendations, and managing model refresh and activation. This skill covers the point-and-click admin path only. For the developer API path (programmatic scoring, bulk predict jobs, REST endpoint integration), use the `einstein-discovery-development` skill instead.

+ Use this skill when an admin configures Einstein Discovery: creating stories, deploying prediction definitions, writing predictions back to Salesforce fields, enabling improvements for what-if analysis, and managing model refresh and activation. It covers the admin path and the metadata that carries a prediction definition between orgs. For programmatic scoring and REST endpoints, use `agentforce/einstein-discovery-development`.

+ Grounding note: the Einstein Discovery admin guide pages are on help.salesforce.com, which did not fetch for this pass, and the Einstein Discovery REST API guide did not fetch either. The grounded source here is the Metadata API Developer Guide's `DiscoveryAIModel`, `DiscoveryGoal`, and `DiscoveryStory` types. Wizard and UI behavior carried from earlier versions of this skill is marked UNVERIFIED where it is stated as fact.

---

## Before Starting

- Gather this context before working on anything in this domain:

- - **CRM Analytics license:** Einstein Discovery requires a CRM Analytics (formerly Tableau CRM) license. It is not available in orgs without this license. Confirm provisioning in Setup > Company Information > Licenses before starting. This is a distinct license from Einstein Prediction Builder, which does not require CRM Analytics.

- - **Admin surface:** Story creation happens in CRM Analytics Studio, not in the main Setup menu. Navigate to App Launcher > Analytics Studio > Create > Story. Admins who look in Setup > Einstein > Einstein Discovery will not find the story wizard there.

- - **Writeback field limits:** Einstein Discovery supports a maximum of three writeback fields per Salesforce entity (object). If the org already has three writeback fields on the target object, a new one cannot be created without removing an existing one.

- - **Model activation is not automatic:** After a model refresh job completes, the new model version is NOT automatically activated. Scoring continues against the prior model version indefinitely until an admin explicitly activates the new version on the prediction definition. There is no error or warning in the UI when this happens.

+ - **CRM Analytics license.** Einstein Discovery requires a CRM Analytics license; Einstein Prediction Builder does not. The Analytics Platform Setup Guide lists Einstein Predictions as a separately purchased capability alongside CRM Analytics Growth and Plus.

+ - **What a prediction definition is.** In metadata it is a `DiscoveryGoal` (Package Manager label "Discovery Prediction"): "a container object in Einstein Discovery that is associated with one or more deployed models". It names the Salesforce object (`subscribedEntity`), the outcome, the deployed models, and the writeback field (`pushbackField`).

+ - **Segmented models.** A prediction definition "can contain up to ten active models"; each deployed model can carry segmentation filters, and "the first model that has filters matching a specific input row will be used to make the prediction".

+ - **Writeback field lifecycle.** "Removing a pushback field from the goal metadata causes the field to be deleted from the Salesforce object as well." Treat the goal file as the owner of that field.

+ - **UNVERIFIED (2026-10-03) claims carried from earlier versions of this skill:** story creation happens only in Analytics Studio (App Launcher, Analytics Studio, Create, Story); at most three writeback fields per object; writeback fields get no field-level security by default; refreshed models are not activated automatically; prediction definition IDs start with `1OR`. These are help-only and should be checked in the org.

---

- ## Core Concepts

- ### Story Creation Wizard — Three Steps

- Einstein Discovery story creation follows a three-step wizard in CRM Analytics Studio:

- 1. **Define the outcome variable:** Choose whether to maximize, minimize, or predict a binary outcome on the target field. Binary outcome (true/false, won/lost, converted/not) uses classification models. Numeric outcomes (revenue, score) use regression models. This step also selects the target Salesforce object.

+ ## Questions to Ask Before Configuring

- 2. **Choose analysis mode:** Select either "Insights only" or "Insights and Predictions."

- - *Insights only* — Einstein analyzes historical data and surfaces correlations and top predictors, but does not deploy a live prediction model that scores records. No writeback field is created.

- - *Insights and Predictions* — Einstein trains a full predictive model (regression or GBM, auto-selected based on outcome type and data profile) and deploys it as a prediction definition that can score new records.

+ Ask these before creating a story. Each one traces to a gotcha in `references/gotchas.md`.

- 3. **Configure explanatory variables:** Choose which fields from the target object (and related objects) Einstein should use as predictors. More relevant variables generally improve model accuracy, but irrelevant or leakage fields (fields that are only populated after the outcome occurs, such as `CloseDate` on a win prediction) must be excluded to avoid artificially inflated accuracy.

+ |---|---|---|---|

+ | "Which outcome are we predicting, and which fields are only known after it happens?" | Leakage fields inflate accuracy and produce useless live scores (gotcha 7) | The outcome field and an excluded-variables list | A model whose scores hold up on open records |

+ | "Do different segments (region, product line) behave differently enough to need their own model?" | One prediction definition holds up to ten active models; the first matching filter wins (gotcha 2) | Segment filters in priority order | Predictions from the right model for each record |

+ | "Which fields are sensitive, and could any act as a proxy for a protected attribute?" | Model fields carry `isSensitive` and `isDisparateImpact` flags (gotcha 6) | A reviewed list of sensitive and disparate-impact fields | A bias review on record before scores reach users |

+ | "Who sees the prediction field, and where?" | The writeback field lives on the record; its visibility follows field-level security (gotcha 8) | The permission sets and layouts that show the score | Scores visible to the right users from day one |

+ | "Will this prediction be moved between orgs with metadata?" | Removing `pushbackField` from the goal file deletes the field (gotcha 1); models are not authored in metadata (gotcha 3) | A retrieve-then-deploy procedure that never drops the field | Promotions that do not destroy the writeback field or its history |

+ | "Who reviews and activates each refreshed model, and how is accuracy judged?" | `terminalStateFilters` define when an outcome is final for accuracy monitoring (gotcha 5) | An owner, a cadence, and a terminal-state definition | Model drift caught by a person, not discovered by users |

- Einstein automatically selects the best-performing model between regression and Gradient Boosting Machine (GBM) based on internal evaluation metrics. Admins do not choose the model algorithm directly.

+ What a proper setup adds over "just deploying the story": the right model scores each segment, sensitive fields are reviewed, the writeback field survives every deployment, and someone owns each refresh.

- ### Prediction Definition and the 1OR Prefix

+ ---

- When a story is deployed with "Insights and Predictions" mode enabled, Salesforce creates a **prediction definition** — a metadata record that links the trained model to the target Salesforce object. Prediction definitions have IDs with the prefix `1OR`.

+ ## Core Concepts

- The prediction definition is the runtime handle for the deployed model. It controls:

- - Which model version is active for scoring

- - Which Salesforce object records are scored

- - Which writeback fields receive prediction scores

- - Whether scoring uses bulk jobs or API calls

+ ### Story, model, and prediction definition

- Prediction definitions can be viewed and managed in CRM Analytics Studio under the story's Deploy tab, and also via the `PredictionDefinition` metadata in Setup.

+ | Concept | Metadata type | Key facts |

+ |---|---|---|

+ | Story | `DiscoveryStory` | The analysis Einstein Discovery builds from a dataset |

+ | Model | `DiscoveryAIModel` (`.model` file plus `-meta.xml`, in the `discovery` folder) | `algorithmType`, `predictionType`, `modelFields`, `status`; "write operations for DiscoveryAIModel objects are generally not supported" |

+ | Prediction definition | `DiscoveryGoal` (`.goal`, in the `discovery` folder) | `subscribedEntity`, `outcome`, `deployedModels`, `pushbackField`, `pushbackType`, `terminalStateFilters`, `active` |

- ### Writeback Fields — System-Managed and Read-Only

+ ### Algorithms and prediction types

- A **writeback field** is a special system-managed custom field that Einstein Discovery creates on the target Salesforce object to store prediction scores and improvement suggestions. Key behaviors:

+ `DiscoveryAlgorithmType` values are `Best` (tournament model), `Glm` (generalized linear model), `Gbm` (gradient boosting machine), `Xgboost`, and `Drf` (random forest). `DiscoveryPredictionType` values are `Regression`, `Classification` (binary), `MulticlassClassification`, and `Unknown`. Earlier versions of this skill said Einstein Discovery chooses only between regression and GBM; the metadata enumeration lists more algorithms. UNVERIFIED (2026-10-03): which algorithms the story wizard offers, and whether the admin chooses or the tournament picks, is help-only.

- - **Read-only:** The writeback field is system-managed. It cannot be edited by users, Flow, triggers, or direct API updates. Any attempt to set its value via DML will be rejected.

- - **Populated only by bulk scoring jobs or explicit API calls:** The field is NOT updated when a record is saved, when a related field changes, or during any normal record lifecycle event. It only updates when a bulk scoring job runs against the prediction definition, or when an explicit scoring API call writes a result.

- - **Maximum three per entity:** Each Salesforce object can have at most three Einstein Discovery writeback fields across all prediction definitions. Attempting to create a fourth will fail.

- - **Field-level security is not pre-assigned:** After the writeback field is created, FLS is not automatically granted to any profile or permission set. Admins must manually assign read access to the field for the profiles or permission sets that need to see it in reports, list views, and page layouts. Until FLS is assigned, the field is invisible to all users including admins.

+ ### Writeback (pushback) fields

- ### Model Refresh and Manual Activation

+ `pushbackField` is the "automated writeback field for predictions. A custom field on the Salesforce object specified in subscribedEntity." `pushbackType` "must be set to AiRecordInsight"; `Direct` is reserved. UNVERIFIED (2026-10-03): earlier versions of this skill said the field is read-only, updates only through bulk scoring or explicit API calls (never on record save), and is limited to three per object. Test record-update behavior in a sandbox before telling users how fresh scores are.

- Model refresh jobs retrain the Einstein Discovery model against the most current data. Refresh jobs can be triggered manually from the story's Model Manager view in CRM Analytics Studio, or configured to run on a schedule.

+ ### Model refresh and activation

- Critical behavior: **a completed refresh job does not automatically activate the new model version.** After a refresh completes, the prediction definition continues scoring against the previous model version. The new model version appears in the Model Manager with a status of "Ready" but is not used until an admin explicitly selects it and clicks "Activate." Failing to activate results in silent scoring drift — the model is stale, but no error or notification is shown.

+ Each deployed model has an `active` flag, and the prediction definition has its own `active` flag. UNVERIFIED (2026-10-03): that a completed refresh leaves the new model inactive until an admin activates it in Model Manager is help-only, but the operating rule is the same either way: someone reviews accuracy and decides which model is active.

---

## Common Patterns

- ### Pattern 1: End-to-End Story Deployment with Writeback

+ ### Pattern 1: End-to-end story deployment with writeback

- **When to use:** An admin needs to create a new Einstein Discovery story on the Opportunity object, deploy it with prediction scoring, and surface the score on the Opportunity record page.

+ **When to use:** Score open Opportunities with a win prediction and show the score on the record page.

**How it works:**

- 1. Navigate to Analytics Studio > Create > Story.

- 2. Step 1: Select Opportunity as the target object, choose the outcome variable (e.g., `IsClosed` binary, or `Amount` numeric).

- 3. Step 2: Select "Insights and Predictions."

- 4. Step 3: Add explanatory variables. Exclude any fields that only populate after close (e.g., `CloseDate`, `ContractId`).

- 5. Let Einstein train the story. After training completes, open the story's Deploy tab.

- 6. Enable "Prediction" and optionally enable "Improvement Suggestions." This creates the prediction definition and writeback fields.

- 7. Navigate to Setup > Object Manager > Opportunity > Fields & Relationships, find the new Einstein writeback field (prefixed `Einstein_`) and assign FLS to relevant profiles.

- 8. Add the writeback field to the Opportunity page layout and optionally embed the Einstein Discovery component for what-if analysis.

- 9. Trigger an initial bulk scoring job from the Deploy tab to populate the writeback field for existing records.

- **Why not skip FLS assignment:** Without FLS, the writeback field is invisible in reports and layouts even for admins. The field technically exists and is being populated, but no user can see the data.

- ### Pattern 2: Model Refresh Activation After Scheduled Retraining

+ 1. Create the story from the Opportunity dataset with the outcome and explanatory variables; exclude fields populated only after close.

+ 2. Review model fields flagged sensitive or disparate impact before deploying.

+ 3. Deploy the model to a prediction definition on `Opportunity`, with a writeback field and improvements if what-if suggestions are wanted.

+ 4. Grant field-level security on the writeback field and add it to the page layout or the record page.

+ 5. Score existing records and spot-check values.

+ 6. Retrieve the `DiscoveryGoal` into source control so the field and model mapping are versioned (see `references/metadata-examples.md`).

- **When to use:** A scheduled model refresh job has completed and the admin needs to ensure new scores reflect the updated model.

+ ### Pattern 2: Segmented prediction definition

- **How it works:**

- 1. After receiving notification (or checking manually) that a refresh job completed, navigate to Analytics Studio > the story > Model Manager.

- 2. Confirm the new model version shows status "Ready." Note the accuracy metrics — compare to the previous active version before activating.

- 3. If the new model performs at least as well as the previous version, click "Activate" on the new version.

- 4. Trigger a new bulk scoring job from the Deploy tab to rescore all records using the activated model.

- 5. Verify the writeback field values have updated on sample records in Salesforce.

+ **When to use:** Win drivers differ by region.

- **Why not skip activation:** If the refresh completed but the new model is not activated, all ongoing scoring — including any bulk jobs triggered after the refresh — continues against the old model version. There is no UI warning, error log, or notification indicating this is happening.

+ **How it works:** Deploy one model per region to the same prediction definition, each with a filter on the region field. Put the most specific filters first, because the first matching model scores the row. Keep a final model with no filters as the catch-all ("no filters indicates that the model matches all input rows"). Stay within ten active models.

---

## Decision Guidance

| Situation | Recommended Approach | Reason |

|---|---|---|

- | Admin needs predictions without CRM Analytics license | Use Einstein Prediction Builder instead | Einstein Discovery requires CRM Analytics; EPB does not |

- | Prediction should update automatically on record save | Use Einstein Prediction Builder | Einstein Discovery writeback fields only update via bulk scoring jobs or API; they are not event-driven |

- | Admin needs point-and-click story creation | Use this skill (admin wizard in CRM Analytics Studio) | No-code three-step wizard is the standard admin path |

- | Developer needs programmatic scoring via REST API | Use `einstein-discovery-development` skill | The admin wizard and the REST API are separate surfaces; this skill covers only the wizard path |

- | Up to three prediction scores needed on one object | Configure up to three writeback fields per entity | System maximum of three writeback fields per object; plan which predictions are most valuable |

- | Model refresh has completed but scores look stale | Manually activate new model version in Model Manager, then run bulk scoring job | New model versions are not auto-activated; scoring runs against old version until explicit activation |

+ | No CRM Analytics license | Einstein Prediction Builder | Einstein Discovery needs CRM Analytics |

+ | Different behavior by segment | Several deployed models with filters in one prediction definition | Up to ten active models; first matching filter wins |

+ | Promote to production | Retrieve `DiscoveryGoal`, deploy it, keep `pushbackField` | Removing `pushbackField` deletes the field |

+ | Need model authored outside Salesforce | Model Manager upload, not metadata | `UserUpload` source type is not supported in the Metadata API |

+ | Fresh scores at a business moment | Developer scoring API (`agentforce/einstein-discovery-development`) | Admin path is not built for event-driven scoring |

+ | Accuracy monitoring | Define `terminalStateFilters` | Accuracy compares predictions with outcomes that have reached a terminal state |

---

## Recommended Workflow

- Step-by-step instructions for an AI agent or practitioner working on this task:

- 1. **Confirm prerequisites:** Verify the org has a CRM Analytics license in Setup > Company Information > Licenses. Confirm the admin has the CRM Analytics Admin permission set. Identify the target Salesforce object and the specific outcome field for the story.

- 2. **Create the story in Analytics Studio:** Navigate to App Launcher > Analytics Studio > Create > Story. Complete the three-step wizard: (a) define the outcome variable and target object, (b) select "Insights and Predictions" if writeback scoring is required, (c) configure explanatory variables and explicitly exclude leakage fields (fields populated only after the outcome).

- 3. **Deploy the prediction definition:** After story training completes, open the Deploy tab. Enable "Prediction" to create the prediction definition (1OR prefix). Optionally enable "Improvement Suggestions" for what-if analysis recommendations. Note the prediction definition ID for operational documentation.

- 4. **Assign field-level security to writeback fields:** Locate the system-generated Einstein writeback field(s) in Setup > Object Manager > [Target Object] > Fields & Relationships. Manually grant read access via FLS to all profiles or permission sets that need to view prediction scores in reports, list views, or page layouts. This step is not automatic and the field is invisible until completed.

- 5. **Add the writeback field to page layouts and run initial scoring:** Add the writeback field to the relevant page layout. From the Deploy tab, trigger an initial bulk scoring job to populate the field for all existing records. Verify the field is populated on a sample of records after the job completes.

- 6. **Configure model refresh schedule:** In the story's Model Manager, set up a refresh schedule appropriate to the data change frequency (e.g., weekly or monthly). Document the schedule and the manual activation step required after each refresh.

- 7. **After each model refresh — activate manually:** When a refresh job completes, review the new model's accuracy metrics in Model Manager, compare against the current active model, and explicitly activate the new model version. Then trigger a new bulk scoring job to update record scores.

+ 1. **Confirm prerequisites.** CRM Analytics license, admin permissions, target object, outcome field, and the excluded-variables list.

+ 2. **Create and review the story.** Build it in Analytics Studio, exclude leakage fields, and review fields flagged sensitive or disparate impact.

+ 3. **Deploy the prediction definition.** Choose segments and filters (up to ten active models), the writeback field, and improvements; define terminal-state filters for accuracy monitoring.

+ 4. **Make the score visible.** Grant field-level security on the writeback field, add it to layouts, and score existing records.

+ 5. **Put it under source control.** Retrieve the `DiscoveryGoal` (and the models it references) with `package.xml`; never deploy a goal file that has lost its `pushbackField`.

+ 6. **Run refresh as an operation.** Name the owner who reviews each refreshed model's metrics and decides which model is active, then rescore.

---

## Review Checklist

- Run through these before marking work in this area complete:

- - [ ] CRM Analytics license confirmed in Setup > Company Information > Licenses

- - [ ] Story created in Analytics Studio using the three-step wizard (not Setup menu)

- - [ ] Leakage fields excluded from explanatory variables in Step 3 of the wizard

- - [ ] "Insights and Predictions" selected (not "Insights only") if writeback scoring is required

- - [ ] Prediction definition created and status is Enabled (1OR prefix ID noted)

- - [ ] Writeback field FLS assigned to all relevant profiles and permission sets

- - [ ] Writeback field added to page layout(s) where scores should appear

- - [ ] Initial bulk scoring job triggered and writeback field values verified on sample records

- - [ ] Model refresh schedule configured and documented

- - [ ] Post-refresh activation step documented in admin runbook (new model versions are not auto-activated)

- - [ ] Writeback field count on target object confirmed to be three or fewer

+ - [ ] CRM Analytics license confirmed

+ - [ ] Leakage fields excluded from explanatory variables

+ - [ ] Sensitive and disparate-impact fields reviewed and recorded

+ - [ ] Segment filters ordered most-specific first, with a catch-all model, at most ten active models

+ - [ ] Writeback field field-level security and layout placement done

+ - [ ] `terminalStateFilters` defined for accuracy monitoring

+ - [ ] `DiscoveryGoal` retrieved into source control with `pushbackField` intact

+ - [ ] Refresh owner and activation procedure documented

---

## Salesforce-Specific Gotchas

- Non-obvious platform behaviors that cause real production problems:

- 1. **Refreshed model is NOT automatically activated — scoring silently runs against stale model** — After a model refresh job completes, the new model version sits in "Ready" status but is not used for any scoring until an admin explicitly activates it. The system provides no error, warning, or alert indicating that the active model is outdated. This is the most common source of silent model drift in production Einstein Discovery deployments.

- 2. **Writeback field is read-only and only updates via bulk scoring or API — never on record save** — The Einstein Discovery writeback field is system-managed. Changing field values on the source record does not trigger re-scoring. A flow or trigger that modifies related fields will not cause the writeback field to update. Admins and users who edit records expecting to see updated predictions will see no change until the next bulk scoring job runs.

- 3. **Maximum three writeback fields per Salesforce entity** — Each Salesforce object can have at most three Einstein Discovery writeback fields across all prediction definitions. If an object already has three and a new deployment attempts to create a fourth, the deployment fails. Audit existing writeback fields before deploying new prediction definitions on high-use objects like Opportunity or Contact.

+ The deep versions, with sources, live in `references/gotchas.md`.

- 4. **Writeback field FLS is not granted automatically — field is invisible until manually set** — Einstein Discovery creates the writeback field and populates it via scoring jobs, but does not assign field-level security to any profile or permission set. The field will not appear in reports, list views, page layouts, or SOQL results for any user until an admin explicitly grants read permission via FLS.

+ | # | Gotcha | One-line consequence |

+ |---|---|---|

+ | 1 | Removing `pushbackField` from goal metadata deletes the field | A careless deploy destroys the score field |

+ | 2 | Up to ten active models; the first matching filter wins | Badly ordered filters send rows to the wrong model |

+ | 3 | `DiscoveryAIModel` write operations are generally not supported | Hand-authored model files do not deploy |

+ | 4 | `pushbackType` must be `AiRecordInsight` | Other values are reserved or unsupported |

+ | 5 | Accuracy needs a terminal-state definition | Monitoring compares against outcomes that are not final |

+ | 6 | Model fields carry sensitive and disparate-impact flags | Bias review skipped if nobody reads them |

+ | 7 | Leakage fields inflate accuracy | Live scores are poor despite great training metrics |

+ | 8 | Writeback visibility and refresh behavior are help-only claims | Test before promising score freshness |

---

## Output Artifacts

| Artifact | Description |

|---|---|

- | Story configuration summary | Outcome variable selection, analysis mode, and explanatory variable list with leakage exclusions noted |

- | Prediction definition record | Created 1OR-prefixed prediction definition with Enabled status on the target object |

- | Writeback field setup confirmation | Einstein writeback field(s) on the target object with FLS assigned to relevant profiles |

- | Page layout update | Writeback field added to the record page layout so users can see prediction scores |

- | Model refresh runbook | Documented schedule and manual activation procedure for post-refresh activation |

+ | Story configuration summary | Outcome, explanatory variables, excluded leakage fields, sensitive-field review |

+ | Prediction definition | `DiscoveryGoal` with models, segment filters, writeback field, terminal-state filters |

+ | Writeback field visibility | Field-level security and layout placement |

+ | Model refresh runbook | Owner, cadence, accuracy review, activation, rescoring |

---

## Related Skills

- - `einstein-discovery-development` — Use when a developer needs to call the Einstein Discovery Connect REST API for programmatic scoring, bulk predict jobs, or model management via API (not the admin wizard)

- - `einstein-prediction-builder` — Use instead when the outcome is binary (yes/no) and the org does not have a CRM Analytics license, or when the point-and-click EPB setup wizard is preferred for simple binary predictions

- - `crm-analytics-vs-tableau-decision` — Use when deciding whether Einstein Discovery (requiring CRM Analytics) is the right tool versus other analytics or ML options in the Salesforce ecosystem

+ - `agentforce/einstein-discovery-development`: Connect REST API scoring, bulk predict jobs, and model management via API

+ - `agentforce/einstein-prediction-builder`: binary predictions without a CRM Analytics license

+ - `architect/crm-analytics-vs-tableau-decision`: whether Einstein Discovery is the right tool

+ - `admin/einstein-discovery-deployment`: the Flow action and Model Manager monitoring