Use purchase history and service data to automatically suggest relevant add-ons and upgrades at the right moment.
Most ecommerce brands leave 15–30% of LTV on the table because their 'you might also like' recommendations are generic — or because cross-sell is only attempted at checkout, long after the customer has decided what they want. AI-powered upsell and cross-sell recommendations for DTC brands use actual purchase-sequence data (what do people who buy X buy next, and when), inventory velocity, and margin per SKU to surface recommendations at three high-conversion moments: post-PDP, checkout, and post-purchase (Day 14 email). Unlike generic 'frequently bought together,' this factors in the customer's specific purchase history, product compatibility, and current inventory availability. It also avoids the 'recommend the $8 item to the $200 customer' trap.
A premium skincare brand in SF moves post-purchase upsells from 'frequently bought together' to sequence-based recommendations ('customers who bought [cleanser] added [serum] within 3 weeks'), lifting AOV on order-2 from $68 to $94. A pet-food brand in Austin uses consumption-rate data to recommend dental chews to customers 45 days into their kibble subscription — attach rate 34%. A home-goods brand in NYC recommends complementary SKUs at checkout based on cart composition (buy bedding set → recommend matching throw pillows), lifting cart value 11%.
List every service or product you offer. For each, define: what's the natural next step? What complementary service would benefit this customer? Example: teeth cleaning -> whitening, Botox -> filler, HVAC repair -> maintenance plan.
Create automation rules: when a customer completes [Service A], wait [X days], then trigger [Recommendation B]. Use tags and custom fields to track what's been offered and declined.
Automated SMS or email with a specific, relevant suggestion: 'Hi [Name], since you just had [Service A], many of our patients also love [Service B] — here's why: [1-line benefit]. Want to learn more?'
Interested replies get booked immediately or routed to a team member. Declined? Tag the contact to avoid re-suggesting the same service. No response? One gentle follow-up in 7 days, then stop.
Monitor acceptance rates by service pair. Double down on high-converting recommendations and retire ones that get ignored or generate negative feedback.
Tuned for Ecommerce Support Teams. Use as-is or adapt to your voice.
For each customer, analyze past 90 days of co-purchase data across the entire customer base: For SKU [X], compute: - Attach rate within 0-30 days post-purchase: top 5 SKUs and their attach % - Attach rate within 30-90 days: top 5 SKUs and their attach % - Customer-specific filter: exclude SKUs this customer already owns Output at three touchpoints: 1. POST-PURCHASE THANK YOU PAGE: 'Customers who bought [product just purchased] often add [top attach SKU within 30 days] for [specific use case / benefit]. Add to your order before it ships — 10% off, one-click add.' 2. ABANDONED CART (if in cart, didn't check out): 'Saw you had [cart item] — our top customers pair it with [complement] for [specific reason]. Want us to drop it in?' 3. DAY 14 POST-DELIVERY EMAIL: 'You bought [Product X] 2 weeks ago. The most common 'what's next' for customers like you is [Product Y] — here's why: [specific benefit tied to the sequence]. 15% off if you reorder [Product X] and add [Product Y] in the same order.' Guardrails: - Never recommend something >50% cheaper than avg cart value (don't anchor down) - Never recommend out-of-stock SKUs - Never recommend SKUs the customer already owns within 90 days - Skip if customer is in an active support ticket (don't sell while complaining)
When cart is built, analyze composition and trigger cross-sell at checkout:
RULES:
1. If cart contains a 'category primary' SKU (e.g., bedding set), recommend category complement (throw pillows, sheets in matching color)
2. If cart contains a 'consumable' (supplement, skincare, coffee), recommend size upgrade ('save 22% per oz with the 16oz vs. 8oz you have now')
3. If cart contains a 'starter' SKU, recommend the 'pro' or 'advanced' version with educational framing ('since you're starting [routine], most customers graduate to [advanced] within 6 weeks — skip the middle step for 15% off the bundle')
4. If cart is $X below free-shipping threshold, surface a specific high-margin complement that closes the gap ('add [item at exactly the gap amount] for free shipping — saves you $X')
5. If customer is VIP (>$500 LTV), skip low-ticket recommendations entirely; only show high-ticket complements
GUARDRAIL: Never add >2 cross-sells at checkout. Never delay load speed. Never pop modal that blocks checkout — recommendations must be in-line.For consumable products (supplements, skincare, coffee, food), compute expected depletion date: EXPECTED_DEPLETION = purchase_date + SKU_lasts_days (from metadata, adjusted by customer's historical reorder cadence) TRIGGER email 10 days BEFORE expected depletion: Subject: [Product] running low? 'Hi [First Name], Based on when you ordered and how long [Product] usually lasts, you're probably about 10 days from running out. Three options: 1. **One-time reorder** — same product, ships tomorrow: [one-click reorder link] 2. **Subscribe and save 15%** — auto-ships every [X] days, cancel anytime: [subscribe link] 3. **Try the upgrade** — [Advanced SKU] is the next level for customers who've been using [current product] consistently. 20% off first order: [upgrade link] If I got the timing wrong or you've switched products, reply and I'll adjust. [Brand]' This single email drives 22–35% of repurchase revenue on consumable brands when timed correctly.
Hi [First Name]! Hope you're loving the results from your [recent service]. Many of our clients pair it with [complementary service] for even better results. Interested? I can check availability for you — just reply YES.
A quick suggestion based on your recent visit, [First Name]
Hi [First Name], Since you recently had [Service A] with us, we wanted to let you know about [Service B] — it's a natural complement that [specific benefit]. Right now we have availability [this week/next week] and your first session is [offer if applicable]. Want to book? Reply to this email or call us at [Phone]. Best, [Business Name]
Get one new AI workflow per week, tuned for Ecommerce Support Teams teams. Real templates, real ROI.
Avoid aggressive upselling if your core service satisfaction isn't high — fix the fundamentals first. Also don't automate upsells for sensitive services (e.g., medical procedures) where the recommendation should come from a licensed provider during a consultation.
One practical AI tip per week. No fluff.
Get the full guide with niche templates and workflow imports.