Upsell Engine

Introduce timely, relevant offers that feel helpful—not pushy.

Outcome at a Glance

Who it’s for: Creators, teams, and stores with 2–10 clear offers. What you’ll achieve: Context-aware product cards, follow-ups, and alerts that surface the right offer at the right moment — automatically. Time to first win: 60–90 minutes (initial setup), then weekly 20-minute tune-ups.


Why it Matters

  • Right time, right product: Offers appear when users show intent — no hunting, no hard sell.

  • Trust-preserving: Frequency limits and smart Alerts keep promotions contextual, not pushy.

  • Compounding results: Weekly keyword and Alert tuning turn small optimizations into steady gains.

The Upsell Engine works because Delphi doesn’t sell — it listens, interprets, and offers exactly what fits.


What You’ll Set Up

Products: We’ll feature 3–5 key items with short descriptions, images, and links that surface naturally during conversations — shown only once per chat.

Tags: Smart tags will track lead states such as Interested, Buyer, and Not Yet Ready to guide follow-ups and segmentation.

Alerts: Intelligent triggers will detect signals of purchase intent or curiosity in real time to identify engaged prospects.

Actions: Automated responses will activate based on alerts — sending follow-ups, showing product cards, or initiating nurture sequences.

Broadcasts: Monthly “What’s New” updates will keep your Interested audience informed and ready to take the next step.


Build → Launch → Operate

Build (45–60 min)

  1. Products:

    • Add 3–5 products manually, via Shopify, or CSV import.

    • Keep each description ≤ 2 sentences: outcome first, details second.

    • Include 3–5 conversational keywords per product (e.g., “pricing,” “framework,” “next step”).

    • Set frequency = once per conversation to avoid oversaturation.

  2. Tags:

    Create tags for each stage of the buyer journey. For example:

    • Interested: For when users clicked the product or asked further questions.

    • Buyer: When a user has purchased.

    • Not Yet Ready: For when users are not interested.

  3. Alerts: Write alerts in plain language to watch for upsell signals. Delphi will interpret them contextually to fire precise actions.

    Examples:

    • “If user asks for deeper training or advanced examples, tag as Upsell Candidate and show ‘Pro Course.’”

    • “When someone mentions wanting templates, guides, or starter kits, tag as Interested and recommend the Starter Pack.”

    • “If user references previous purchase or interest in upgrades, tag as Repeat Buyer and share Premium Tier.”

  4. Actions:

    • Action A: Product RecommendedSend Message (User) — acknowledge, give one benefit, link, invite one question.

    • Action B: Alert Triggered (Interested)Send Message (User) — personalized offer tied to detected interest.

    • Action C: Inactivity 1d (Tagged Interested)Send Message (User) — gentle check-in (Fixed Delay).


Launch (10–15 min)

  1. Enable Actions A–C in your workspace.

  2. Create your monthly Broadcast for the Interested tag: “New resources and updates you asked about.”

  3. Verify that Alerts fire correctly (Testing → Actions Log).

The best upsell moments aren’t scripted — they’re recognized. Alerts find them.


Operate (Weekly, 20 min)

  1. Review Product Mentions → note which items drive clicks by using UTM links.

  2. Refine Alerts: check if your alerts are triggering at the right times, and try updating or creating new ones to improve accuracy.

  3. Adjust Keywords: align with how users naturally ask (avoid internal jargon).

  4. Audit Actions: confirm triggers fire and feel conversational.


Recipes to Follow

Action A — Follow-Up After Recommendation

Trigger: Product Recommended Action: Send Message (User) (Immediate) Instruction: “Acknowledge context; state one benefit; share link; invite a question.”

Example: “That’s a great question about pricing. My Pricing Playbook walks you through it step-by-step — here’s the link. Want a quick 2-minute overview?”


Action B — Alert-Triggered Upsell

Trigger: Alert “If user shows interest in advanced training…” Action: Send Message (User) (Immediate) Instruction: “Recognize their interest; introduce relevant offer; keep tone helpful.”

Example: “Since you mentioned scaling your audience, you might like my Growth Accelerator. It covers exactly that — want me to send details?”


Action C — Gentle Check-In

Trigger: Inactivity 1d Filter: User Tag = Interested Action: Send Message (User) (Fixed Delay) Instruction: “Check if timing was the issue; offer one-sentence summary + link; invite a question.”


Broadcast — Monthly ‘What’s New’

Subject: “New resources you asked for” Body: Hi {name}, here are the resources people asked about most this month: [Product A] — one-liner outcome [Product B] — one-liner outcome [Product C] — one-liner outcome Reply with your goal and I’ll point you to the best next step.


Metrics to Watch

  • Mentions → CTR (per product)

  • Conversions after recommendation

  • Alert → Action conversion rate (triggered upsells)

  • Return rate (7-day) for Interested users

  • Unanswered product questions → revised trend


Troubleshooting

  • Feels spammy: Drop frequency to once per user; refine benefit copy; review alert precision.

  • Low clicks: Lead with outcomes; move link earlier; verify tone isn’t too generic.

  • Alerts not firing: Check phrasing and scope; confirm correct tag or action mapping.

  • Too many false positives: Add conditions (“if mentions pricing AND readiness”).


Pre-Flight Checklist

  • 3–5 products added with keywords and frequency = once per conversation.

  • Tags (Interested, Buyer, Not Yet Ready) created and active.

  • Alerts configured to detect upsell cues and tag accordingly.

  • Actions A–C tested in workspace.

  • Monthly Broadcast drafted.

  • Weekly product review and Alert-tuning block scheduled.


In Practice

Your Delphi becomes a silent salesperson — one that never pushes, only offers. Alerts read the room; Actions deliver relevance. Each conversation teaches your Delphi when to suggest, when to wait, and when to follow up.

“If a user asks about templates or scaling, tag as Interested and recommend the Starter Pack.” That’s not a campaign — that’s an intelligent upsell in real time.

Last updated