E-commerce
June 26, 2026
A generic Shopify chatbot responds like an FAQ. A chatbot connected to customer data can recognize the context: recent order, VIP status, regular size, language, purchase history, subscription, or already declared preference.
The promise is not just to say "hello first name" everywhere. True personalization consists of saving the customer from repeating themselves, offering a more accurate response, and passing on the right context when a human needs to take over.
This article #20 is more precise than general content. If API is unavailable: degraded mode (#279). on e-commerce personalization: it explains which Shopify data to use, when to use it, how to stay useful, and how to test without becoming intrusive.
Summary
What customization is truly useful?
Useful personalization changes the response. If the customer data does not modify the advice, the support, or the routing, it only serves to decorate the conversation.
Useful personalization
The bot recognizes an order, proposes the correct status, recommends a compatible variant, adapts the handoff, or avoids asking a question that has already been answered.
Useless personalization
The bot repeats the first name, mentions the history for no reason, or displays internal information that makes the customer uncomfortable.
Simple rule
Before using customer data, write the following sentence: "thanks to this data, the bot can...". If you do not finish the sentence clearly, do not use it yet.
Shopify presents unified customer profiles as a foundation for recognizing a customer's history, preferences, and interactions within a single profile (Shopify, unified customer profiles).
Convert over 2,000 customers on average per month with Qstomy.
The world’s 1st Shopify AI dedicated to customer conversion
Empowering 200+ e-commerce merchants
Which Shopify data should you read first?
Orders: number, items, payment status, fulfillment, tracking
History: products purchased, frequency, last order
Tags: VIP, wholesale, subscriber, frequent return, B2B
Segments: loyal customers, inactive, high average cart, country
RFM: Shopify segment (Champions, At Risk…) for internal bot routing, see RFM chatbot (#276)
Metafields: usual size, preference, language, skin profile
Consent: email opt-in, SMS, marketing, country and language
Start with fields that are actually filled and reliable. A planned but empty metafield does not personalize anything. A poorly maintained tag creates bad responses.
Quick audit
Take 20 random customers and check how many have a usable history, consistent tags,
How to personalize post-purchase support?
Post-purchase is the easiest case to make profitable, as Shopify data immediately changes the answer.
Identify the customer or order
Read the fulfillment status and the tracking link
Adapt the response to the country, carrier, and lead time
Escalate if the package is blocked, damaged, or involves a sensitive refund
Concrete script
« I see that your order #1042 was shipped yesterday with Colissimo. Tracking indicates physical delivery is scheduled for tomorrow. If the status does not change within 48 hours, I can forward the file to the team. »
Gorgias explains that its AI Agent can leverage Shopify, orders, customer profiles, products, and fulfillment to personalize responses (Gorgias AI Agent).
How to customize pre-purchasing without making a mistake?
In pre-purchase, history helps, but it should never override the current intent.
Repurchase: suggest the same product or a compatible refill
Accessory: suggest an add-on to a previously purchased product
Size: use the usual size with caution
Upgrade: recommend a premium version if usage justifies it
Gift: ignore history if the customer is buying for someone else
Shopify describes the AI personal shopper as an assistant capable of using product information, behavioral data, and user queries to provide contextualized advice (Shopify AI personal shopper).
Priority to Current Intent
If a loyal customer is buying a gift, their personal history may mislead the chatbot. Ask "is this for you or a gift?" first before recommending based on their past purchases.
Tags, segments, and metafields: when to use what?
Tags
Use them for simple rules: VIP, wholesale, subscriber, pro customer, frequent returner. They are quick to read, but minimally structured.
Segments
Use them for dynamic groups: high spenders, inactive customers, customers who purchased a specific collection. Shopify Segmentation allows you to create marketing audiences that update in real-time with customer data (Shopify Segmentation).
Metafields
Use them for precise data: shoe size, preferred language, skin type, birthday, usual size, B2B status. This is the best format when the bot needs to answer with exact accuracy.
What rules should be written for the bot?
Personalization must be driven by rules that are readable by the support team.
VIP: priority handoff if there is a dispute or high cart value
Wholesale: do not apply standard D2C scripts
Frequent returns: check size or compatibility before making a recommendation
Subscriber: offer postponement, frequency change, or subscription management
Country: adapt delivery, returns, and taxes if rules change
Gorgias Guidance allows you to write conditional instructions of the IF/THEN type using Shopify variables, for example, if tags include loyalty-gold, then remind them of the associated benefit (Gorgias Guidance).
Recommended format
Write rules in simple language: "if wholesale customer, do not offer D2C promo code; transfer to professional contact if the request is about volume, quotes, or special lead times."
How do you identify the customer without friction?
Personalization requires knowing who you are talking to, but identification must not block the conversation.
Anonymous visitor: only use page context and session cart
Log-in customer: use profile, orders, and authorized preferences
Non-logged-in post-purchase: ask for email or order number
Sensitive action: require stronger verification
Security rule
Never display an address, refund status, or order details without sufficient verification. It is better to have a slightly less personalized response than a data leak.
Example journey
For a WISMO question, ask for an email or order number, display only the useful status, then offer escalation if the package is blocked. Do not read the entire customer history for a simple question.
Which response scripts should be used?
Returning customer
“You ordered this model in size M last time. For this new cut, I also recommend M, unless you prefer a looser fit.”
VIP customer
“I am forwarding your request to the team along with your order history, so you don't have to repeat everything.”
Subscriber
“Your next delivery is scheduled for next week. I can help you postpone it or change the frequency.”
Anonymous visitor
“I can guide you based on the product you are currently viewing. To check an order, I will need your email address or order number.”
What GDPR and trust boundaries should be respected?
Purpose: use the data to help, not to surprise
Minimization: read only what is needed to respond
Transparency: explain why a data point is requested
Masking: never expose sensitive internal tags
Consent: do not turn support into unsolicited marketing
A bot can say "I am retrieving your order". It must not say "you are classified as a high-risk customer". Internal data must guide the service, not be displayed raw.
How to test before generalizing?
Test by scenario, not by global promise.
Logged-in customer WISMO for 14 days
Simple return with verified order
Repurchase or compatible accessory for recurring customers
VIP routing to human with summary
Anonymous customer without data leakage
Measure resolution rate, escalation, CSAT, recontact, assisted conversion, and personalization errors. To be linked with e-commerce chatbot KPIs and chatbot zero-party data.
Validation threshold
Do not generalize if the bot only personalizes a part of the conversations but increases recontacts. Correct personalization reduces repetition and improves clarity.
How does Qstomy customize with Shopify?
Qstomy connects the chatbot to the Shopify context to personalize where it creates real value: post-purchase support, product advice, repurchase, compatibility, and handoff.
DTC fashion scenario
Shopify store, 35,000 customers. Qstomy reads recent orders, usual sizes, and VIP status. Out of 1,500 conversations/month, 420 concern an order, 180 a size question, and 75 a recurring customer. Pilot hypothesis: 58% of WISMO resolved without a human, 90 assisted cart additions based on size or repurchase recommendations, and priority handoff for 35 VIP customers.
The goal is not to over-personalize. It is to reduce repetition, improve advice, and pass the right context to the human.
In this scenario, personalization starts with post-purchase because the risk is low and the value is immediate. Pre-purchase recommendations come next, once the size and history rules are validated.
See Shopify integration, AI customer support, AI sales assistant and request a demo.
Which playbooks should be applied this week?
Playbook 1: fields inventory
List orders, tags, segments, and metafields that are actually filled. Remove from the project any field that is too empty or poorly maintained.
Playbook 2: custom WISMO
Launch identified order tracking: status, carrier, delay, and escalation in case of blockage.
Playbook 3: VIP rules
Define precisely what the VIP tag changes: priority, tone, handoff, but not an invented commercial promise.
Playbook 4: security testing
Test bad email, non-existent order, anonymous customer, sensitive tag, and refund request before going live.
Useful linking
Support: automated post-purchase support
Handoff: human chatbot handoff
Enzo
June 26, 2026
