Salesforce for Retail, Online and Loyalty
Retail customers shop in stores, online and on apps, but most retailers see them as separate people. Here is how Salesforce brings those views together.

Retailers know a great deal about transactions and surprisingly little about customers. The point-of-sale (POS) system records a basket, the ecommerce platform records an order, and the loyalty programme records points, but the same person often appears as three records with different identifiers. Without a reliable link between them, personalisation stays shallow and store teams cannot recognise their best customers.
Retail engagement challenges
- Anonymous store transactions. Many in-store sales are not tied to a customer at all. Capturing an identifier at the till (loyalty number, phone, e-receipt email) is the foundation for everything else.
- Duplicate and conflicting profiles. A customer who shops online with one email and in store with a phone number looks like two people until you resolve them.
- Promotions without context. Sending discounts to people who would have bought anyway erodes margin.
- Inventory-blind messaging. Recommending products that are out of stock locally frustrates customers.
- Store and head office misalignment. Clienteling tools, service desks and marketing often work from different customer lists.
Retail use cases mapped to Salesforce
Unified customer profile
Data Cloud (now branded Data 360) ingests POS, ecommerce and loyalty data and applies identity resolution so one person has one profile.
→Loyalty tier messaging
Journeys triggered by tier changes, points expiry and anniversaries, with offers that respect tier rules.
→Segmentation and personalisation
Segments built on recency, frequency, category affinity and preferred store, used in email, SMS and paid media.
→Abandoned cart and browse recovery
For retailers with online channels, timely reminders with suppression when the customer buys in store.
→Lapsed customer win-back
Customers with no purchase across any channel for a defined period enter a win-back programme with a holdout group.
→Getting identity right
Identity resolution only works as well as the identifiers you capture. Before configuring match rules, look at what each channel collects and how clean it is: loyalty numbers are usually reliable, phone numbers need normalising, and email addresses captured at a till are often mistyped. Decide which identifiers can be used for exact matching and which only support fuzzy matching, and test rules against a sample before applying them to the full base. Over-matching (merging two different people) is usually more harmful than under-matching.
Using store and inventory context
Messages become more useful when they reflect the customer's preferred store and what is actually available there. That means bringing store attributes and, where practical, stock signals into the data that drives segments and content. Real-time stock is not always necessary; a daily availability flag by store and category often covers promotional needs.
Data, consent and compliance considerations
- PCI DSS. Payment card data should stay with your payment provider and POS. Marketing and CRM systems should never receive full card numbers.
- Consent at the till. Capturing an email for an e-receipt is not the same as consent to marketing. Ask separately and record the answer with a timestamp and source.
- Privacy law. India's DPDP Act and, for UK or EU customers, GDPR apply to customer data. Offer clear ways to withdraw consent and access preferences.
- Loyalty terms. Make sure how you use purchase data for personalisation matches what your loyalty terms say.
Your legal team should confirm specific obligations; this is general guidance.
Integration landscape
POS and store systems, ecommerce platform, loyalty engine (Salesforce offers Loyalty Management, or you may run another), order management, product catalogue and pricing, customer service tools, and data warehouse. POS data is often loaded in daily batches; online events and loyalty tier changes benefit from streaming or API-based ingestion.
How GeekMidst would start
- 01
Inventory identifiers by channel
Measure capture rates and quality for loyalty number, email and phone in stores and online.
- 02
Agree a profile strategy
Match rules, source priority for conflicting fields and how unified profiles will be used.
- 03
Ingest two sources first
Typically ecommerce and loyalty, then POS once the model is proven.
- 04
Launch one cross-channel journey
For example, loyalty tier messaging with suppression of recent store purchasers.
- 05
Measure against a holdout
Compare behaviour with a control group so the programme's contribution is clear.
Workflow diagram
Explore how the pieces fit together. Select a step to read what happens there.
Select any step in the diagram, or use Step through, to see what happens at that point and why.
Text version: every step in order
- Views products VisitorKnown visitor, no cart
Browse journeys only work for identified visitors, for example people who arrived from an email click or are signed in, so views can be tied to a contact.
Next: Views captured
- Views captured WebsiteWeb tracking with contact key
Tracking on the site records product, category and time against the contact key. Without the key, views stay anonymous and cannot trigger an email.
Next: Audience query
- Audience query Automation StudioScheduled SQL activity
An Automation Studio SQL query picks recent views, removes anyone with a cart or order, applies a frequency cap and writes the journey entry data extension.
Next: Bought since viewing?
- Bought since viewing? Journey Builder
Orders can land between the query and the send, so the journey checks again at entry. Purchasers should never receive a browse reminder.
Next: Yes: Order already placed; No: Browse reminder
- Order already placed WebsiteExits without email
Visitors who bought are removed from the journey. Order data synced from the commerce platform is what makes this check possible.
End of the flow.
- Browse reminder VisitorViewed items plus similar
The email shows the viewed products with current price and stock, plus related items. Content falls back to bestsellers when product data is missing.
Next: Wait 2 days
- Wait 2 days Journey BuilderSend windows apply
Gives the visitor time to return on their own before any follow-up. Send-time windows keep messages within reasonable hours.
Next: Clicked or ordered?
- Clicked or ordered? Journey Builder
An engagement or decision split checks for a click or an order. Clicks and orders are better signals than opens, which mail privacy features inflate.
Next: Yes: Returned to the site; No: Follow-up picks
- Returned to the site WebsiteEngaged, exits
Engaged visitors leave this journey. Their next steps are handled by cart or post-purchase journeys instead, so messages do not overlap.
End of the flow.
- Follow-up picks VisitorCategory or bestsellers
One last, lighter message with category picks or bestsellers. Sending more than this risks unsubscribes from people who were only browsing.
Next: Exits journey
- Exits journey Journey BuilderRe-entry rules
The visitor leaves the journey. Re-entry settings and the frequency cap in the query stop the same person being reminded too often.
End of the flow.
Questions we hear
How do we link in-store purchases to customers?
Do we need Data Cloud, or is Marketing Cloud enough?
Can store staff see customer preferences?
Connect your store, online and loyalty data
Tell us which channels and systems you run today. We will outline how to build a unified customer view in stages.




