	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Win-Back Agent for Lapsed Customers

Win-Back Agent for Lapsed Customers

Author: Venkata Sudhakar

ShopMax India has thousands of customers who were once active buyers but have not placed an order in over 90 days. Some drifted to Amazon or Flipkart. Others simply forgot. A win-back agent that identifies these lapsed customers, understands what they bought before, and sends a relevant, personalised offer can recover a meaningful percentage of this dormant revenue at a fraction of the cost of acquiring new customers.

This tutorial builds a Win-Back Agent using ADK and Gemini. The agent segments lapsed customers by recency and lifetime value, selects the highest-potential win-back candidates, and generates personalised re-engagement messages with tailored incentives based on their purchase history.

The below example shows the win-back workflow for ShopMax India lapsed customers.

from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService

# --- Lapsed customer database ---
LAPSED_CUSTOMERS = [
    {
        "id": "C101", "name": "Naveen Sharma", "city": "Delhi",
        "days_since_last_order": 120,
        "lifetime_value": 145000,
        "total_orders": 9,
        "favourite_categories": ["tv", "audio"],
        "last_product": "LG OLED TV",
        "tier": "Gold",
    },
    {
        "id": "C102", "name": "Pooja Menon", "city": "Bangalore",
        "days_since_last_order": 95,
        "lifetime_value": 38000,
        "total_orders": 3,
        "favourite_categories": ["tablet", "accessories"],
        "last_product": "iPad case and keyboard",
        "tier": "Silver",
    },
    {
        "id": "C103", "name": "Harish Babu", "city": "Chennai",
        "days_since_last_order": 200,
        "lifetime_value": 12000,
        "total_orders": 1,
        "favourite_categories": ["audio"],
        "last_product": "Bluetooth speaker",
        "tier": "Bronze",
    },
]

# --- Tool: get lapsed customers ---
def get_lapsed_customers(min_days_lapsed: int = 90) -> dict:
    """Fetch customers who have not purchased for a given number of days.
    Args:
        min_days_lapsed: Minimum days since last order (default 90)
    Returns:
        Segmented list of lapsed customers
    """
    lapsed = [c for c in LAPSED_CUSTOMERS if c["days_since_last_order"] >= min_days_lapsed]
    return {"total": len(lapsed), "customers": lapsed}

# --- Tool: score win-back potential and generate offer ---
def score_winback(customer_id: str) -> dict:
    """Score a lapsed customer for win-back potential and suggest incentive.
    Args:
        customer_id: Customer identifier
    Returns:
        Win-back score, recommended offer, and message tone
    """
    customer = next((c for c in LAPSED_CUSTOMERS if c["id"] == customer_id), None)
    if not customer:
        return {"error": f"Customer not found: {customer_id}"}
    score = 0
    # Lifetime value
    if customer["lifetime_value"] >= 100000:
        score += 40
    elif customer["lifetime_value"] >= 30000:
        score += 25
    else:
        score += 10
    # Order frequency
    if customer["total_orders"] >= 5:
        score += 30
    elif customer["total_orders"] >= 2:
        score += 15
    # Recency penalty (longer lapse = harder to win back)
    if customer["days_since_last_order"] <= 120:
        score += 20
    elif customer["days_since_last_order"] <= 180:
        score += 10
    # Select incentive
    if score >= 60:
        offer = "8% off your next order + free express delivery"
        tone = "warm and VIP"
    elif score >= 35:
        offer = "5% off coupon COMEBACK5"
        tone = "friendly and casual"
    else:
        offer = "free delivery on your next order"
        tone = "light and low-pressure"
    return {
        "customer_id": customer_id,
        "name": customer["name"],
        "winback_score": score,
        "priority": "HIGH" if score >= 60 else ("MEDIUM" if score >= 35 else "LOW"),
        "offer": offer,
        "tone": tone,
        "favourite_categories": customer["favourite_categories"],
        "last_product": customer["last_product"],
    }

# --- Agent setup ---
lapsed_tool = FunctionTool(func=get_lapsed_customers)
winback_tool = FunctionTool(func=score_winback)

agent = LlmAgent(
    name="winback_agent",
    model="gemini-2.0-flash",
    instruction="""You are a customer win-back agent for ShopMax India.
1. Call get_lapsed_customers to fetch all lapsed customers.
2. For each customer, call score_winback to get priority, offer, and tone guidance.
3. Write a re-engagement message matching the recommended tone:
   - Mention their last purchase or favourite category to show you remember them
   - Include the specific offer with a 7-day expiry
   - HIGH priority: warmer, more personal, mention their Gold/Silver status
   - LOW priority: brief and non-intrusive
Address customers by first name. Keep messages under 80 words.""",
    tools=[lapsed_tool, winback_tool],
)

session_service = InMemorySessionService()
runner = Runner(agent=agent, app_name="winback", session_service=session_service)

import asyncio

async def run_winback(query: str) -> str:
    session = await session_service.create_session(app_name="winback", user_id="crm_team")
    from google.genai.types import Content, Part
    msg = Content(role="user", parts=[Part(text=query)])
    result = ""
    async for event in runner.run_async(user_id="crm_team", session_id=session.id, new_message=msg):
        if event.is_final_response() and event.content:
            result = event.content.parts[0].text
    return result

response = asyncio.run(run_winback("Run the win-back campaign for all customers lapsed 90+ days."))
print(response)

It gives the following output,

Win-Back Campaign - ShopMax India (3 lapsed customers)

--- C101 | Naveen Sharma, Delhi | HIGH PRIORITY (score: 70) ---
Offer: 8% off + free express delivery

Hi Naveen, we miss you! As one of our valued Gold members and a big fan of
home entertainment, we wanted to personally reach out. Your next order gets
8% off plus free express delivery - just for you. Explore the latest TVs and
soundbars at shopmax.in. Offer valid for 7 days. Welcome back!

--- C102 | Pooja Menon, Bangalore | MEDIUM PRIORITY (score: 40) ---
Offer: 5% off coupon COMEBACK5

Hi Pooja, it has been a while! We have some great new tablets and accessories
you might love. Use code COMEBACK5 for 5% off your next order.
Valid for 7 days: shopmax.in

--- C103 | Harish Babu, Chennai | LOW PRIORITY (score: 20) ---
Offer: Free delivery

Hi Harish, free delivery on your next ShopMax order - no minimum spend.
Check out our latest audio deals: shopmax.in/audio

Run this win-back campaign monthly via Cloud Scheduler. Suppress customers who have already responded to a win-back message in the last 60 days by checking a Firestore log. A/B test the incentive levels - 5% vs 8% - across customer segments using BigQuery experiment tracking to find the minimum incentive needed for each tier.

Send your comments, suggestions or queries regarding this site to [email protected].