Action Strategy
An Action Strategy is a visual flowchart, built in the Strategy Builder canvas, that decides which Recommendation records to surface to a user at a given moment.
Definition
An Action Strategy is a visual flowchart, built in the Strategy Builder canvas, that decides which Recommendation records to surface to a user at a given moment. The strategy takes context as input (the current record, the user, business rules, model scores) and runs through a series of nodes that load, filter, branch, enhance, and rank Recommendations before returning the final list to a Next Best Action component on a Lightning page or to a Flow. Filters use the standard formula language. Branches use If conditions. Predictive nodes call Einstein Discovery models with a single click.
Action Strategies sit at the centre of Salesforce Einstein Next Best Action. They are how a static list of possible offers (Recommendations stored in a custom object) becomes a personalised, ranked, contextual prompt on the record page. A bank uses one to suggest Open a Certificate of Deposit or Schedule Retirement Review on the Person Account based on age, balances, and recent activity. A retailer uses one to suggest service upgrades after a support case. The strategy is the brain; the Recommendation records are the menu.
What goes into a working Action Strategy
The node types you actually use
Strategy Builder offers a fixed set of nodes. Load Recommendations pulls candidate records from the Recommendation object. Branch Selector splits the flow on a condition. Filter trims candidates against a criterion. Map Recommendation rewrites field values inline. Enhance Recommendations attaches data from other records. Invocable Action calls a Flow or Apex method. Sort orders the candidates by a numeric field. Limit Reoffers caps how many can come back. Most real strategies use four or five node types; the rest sit there for edge cases.
Loading Recommendations from the standard object
The Recommendation object is the menu of things the strategy can suggest. Each row stores Name, Description, Image URL, Action Reference (a Flow, URL, or quick action to fire on accept), and any custom criteria fields. The Load Recommendations node pulls a subset (usually filtered by category, eligible product, or audience tag) into the strategy. Without seed Recommendations, the strategy returns nothing.
Filtering against record and user context
Filter nodes use the standard formula language to keep or drop candidates based on context. The strategy receives a $Record context (the record the user is viewing) and a $User context (the running user). Filters like $Record.AnnualRevenue > 1000000 and $User.Profile.Name = Retail Banker are typical. The filter language is identical to the formula language used elsewhere in Salesforce, which keeps the learning curve flat for admins.
Branching and conditional logic
Branch Selector nodes split the strategy into parallel paths based on a condition. Different cohorts can run through different filters and different sort orders, then converge again into a single output node. Branches are how the strategy expresses what offers we show to small business clients vs. wealth management clients vs. credit card prospects, without copy-pasting the logic per audience.
Calling predictive models from a strategy
An Enhance Recommendations node can call an Einstein Discovery model to attach a propensity score to every candidate. A subsequent Sort node ranks by that score, putting the most likely-to-convert Recommendation first. This is how the strategy goes from rule-based to predictive without writing code. The strategy stays declarative; the model does the heavy lifting.
Surfacing a strategy through the Next Best Action component
The Einstein Next Best Action Lightning component takes a strategy as a parameter and renders the returned Recommendations as a list of cards. Each card has the Description, an action button, and a reject button. Accepting a recommendation fires the Action Reference (a Flow that opens a new Opportunity, a URL that opens a quote tool, etc.). Rejecting it logs feedback the strategy can use the next time it runs.
Versioning, activation, and audit
Strategies have a Draft, Active, Inactive lifecycle. Only one version of a strategy can be Active at a time. Saving creates a draft. Activating publishes it to runtime. Old versions remain available to roll back. The Strategy Builder canvas always opens the latest draft, not the running version, which can confuse users who edit and assume their change is live. Activation is the final, deliberate step.
Performance, limits, and monitoring
Strategies run synchronously inside the user's record page load. Heavy logic, large candidate sets, and external callouts can blow the page-load budget. Salesforce caps the number of nodes per strategy and total runtime. The Strategy Tester tool runs a strategy against a sample context and shows which nodes filtered out which candidates, which is the main debugging surface. Monitoring usage tab numbers (accepts, rejects, impressions) is how teams prove the strategy is working in production.
How to build an Action Strategy
Strategies live in Setup under Einstein Next Best Action. They are built node by node on a visual canvas; the right node names and the right context filters make or break the result.
- Open Strategy Builder
Setup, then search for Next Best Action. Open the Strategies tab, then click New Strategy. Give it a name and a description tied to the business question (e.g. Retail Banking - Front Page Offers).
- Add a Load Recommendations node
Drag a Load Recommendations node onto the canvas. Filter to the subset of Recommendation records that make sense for this surface, usually by Category or Audience.
- Add Filter nodes for context-specific rules
Drop in Filter nodes that use $Record and $User context expressions to keep only relevant candidates. Test each filter in isolation before chaining.
- Optionally call a predictive model with Enhance Recommendations
Use Enhance Recommendations to attach an Einstein Discovery score, then Sort by that score. This is what turns the rule-based shortlist into a ranked, personalised list.
- Set the output node and activate the strategy
Wire the final node into the strategy output. Save as Draft, run the Strategy Tester against a known record, then Activate. The Einstein Next Best Action Lightning component on the page picks up the active version automatically.
Human-readable strategy name; surfaced in the component picker and Strategy Tester.
The object whose record page the strategy runs against. Sets which $Record context fields are available.
The final node whose Recommendation list is returned to the caller.
Draft, Active, or Inactive. Only Active strategies serve traffic.
- Saving creates a draft. The running version stays the previously activated one until you click Activate. Editors regularly assume their save is live and chase phantom bugs.
- Filters that touch $Record fields must run against records the user has Read access to. Hidden records produce empty contexts and the strategy silently returns no Recommendations.
- Strategies run synchronously on the record page. Heavy callouts or unfiltered Load nodes will visibly slow page load.
- Only one strategy can be Active per name. To run two variants in parallel, create separate strategies, not two versions of the same one.
Trust & references
Cross-checked against the following references.
- Strategy BuilderSalesforce Help
- Einstein Next Best Action OverviewSalesforce Help
Straight from the source - Salesforce's reference material on Action Strategy.
- Design a StrategySalesforce Help
About the Author
Dipojjal Chakrabarti is a B2C Solution Architect with 29 Salesforce certifications and over 13 years in the Salesforce ecosystem. He runs salesforcedictionary.com to help admins, developers, architects, and cert/interview candidates sharpen their fundamentals. More about Dipojjal.
Test your knowledge
Q1. What is the main advantage of using Action Strategy over writing Apex code?
Q2. Which Salesforce tool has Salesforce recommended as the future of automation?
Q3. What type of Salesforce feature is Action Strategy?
Discussion
Loading discussion…