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

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Revenue Forecasting Agent

Revenue Forecasting Agent

Author: Venkata Sudhakar

Revenue forecasting is a critical input for budgeting, staffing, and inventory planning. Finance teams need category-level revenue projections that account for growth trends and seasonal patterns, not just simple averages of historical data.

A Revenue Forecasting Agent uses Gemini AI to analyse historical revenue by product category, apply trend and seasonality factors, and produce a monthly revenue forecast with variance analysis. This gives ShopMax India's finance team a reliable baseline for the next quarter's budget planning.

The below example shows how ShopMax India generates a 3-month revenue forecast for its major product categories using 6 months of historical revenue data.

import asyncio
import json
from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.genai.types import Content, Part

# Historical monthly revenue in Rs lakhs (Oct-Mar)
REVENUE_HISTORY = {
    "Mobiles":    [42.5, 44.0, 38.0, 51.0, 68.0, 62.0],
    "Laptops":    [18.0, 19.5, 17.0, 22.0, 30.0, 27.5],
    "Audio":      [8.5,  9.0,  7.5,  10.0, 14.0, 12.0],
    "TV":         [22.0, 24.0, 20.0, 18.0, 16.0, 15.0],
    "Accessories":[6.0,  6.5,  5.5,  7.0,  9.5,  8.0],
}
PAST_MONTHS = ["Oct", "Nov", "Dec", "Jan", "Feb", "Mar"]
FORECAST_MONTHS = ["Apr", "May", "Jun"]

# Seasonal index for Apr-Jun (1.0 = average, >1 = above avg)
SEASONAL_INDEX = {"Apr": 0.95, "May": 0.90, "Jun": 0.85}

def forecast_revenue(data_json: str) -> str:
    """Forecast 3-month revenue using trend and seasonality."""
    data = json.loads(data_json)
    history = data["history"]
    forecast_months = data["forecast_months"]
    seasonal = data["seasonal_index"]

category_forecasts = []

for cat, values in history.items():
        n = len(values)
        # Weighted moving average
        weights = list(range(1, n + 1))
        wma = sum(v * w for v, w in zip(values, weights)) / sum(weights)

# Trend: recent 3 vs early 3
        growth_pct = ((sum(values[-3:]) / 3 - sum(values[:3]) / 3) / (sum(values[:3]) / 3)) * 100

monthly_fc = []
        for month in forecast_months:
            base = wma * (1 + growth_pct / 200)
            seasonalised = base * seasonal[month]
            monthly_fc.append({"month": month, "forecast": round(seasonalised, 1)})

quarter_total = sum(f["forecast"] for f in monthly_fc)
        category_forecasts.append({
            "category": cat,
            "growth_trend_pct": round(growth_pct, 1),
            "monthly": monthly_fc,
            "quarter_total": round(quarter_total, 1),
        })

category_forecasts.sort(key=lambda x: x["quarter_total"], reverse=True)
    total_q = sum(c["quarter_total"] for c in category_forecasts)

return json.dumps({"q1_forecast": category_forecasts, "total_quarter_revenue": round(total_q, 1)})

forecast_tool = FunctionTool(func=forecast_revenue)

AGENT_INSTRUCTION = """
You are a Revenue Forecasting Agent for ShopMax India finance.
When given historical revenue and seasonality data:
1. Call forecast_revenue with the data
2. Show a category revenue forecast table: category, growth trend, Apr/May/Jun forecast, quarter total
3. Highlight categories with strong positive trends (>20%) as growth opportunities
4. Flag categories with negative trends - these need sales team attention
5. State the total forecast revenue for the quarter and compare to the prior quarter average
Format as a quarterly revenue forecast report for the CFO.
"""

revenue_agent = LlmAgent(
    name="revenue_forecast_agent",
    model="gemini-2.0-flash",
    instruction=AGENT_INSTRUCTION,
    tools=[forecast_tool],
)

async def main():
    session_service = InMemorySessionService()
    runner = Runner(
        agent=revenue_agent,
        app_name="revenue_forecast_app",
        session_service=session_service
    )
    session = await session_service.create_session(
        app_name="revenue_forecast_app",
        user_id="cfo"
    )
    payload = json.dumps({
        "history": REVENUE_HISTORY,
        "forecast_months": FORECAST_MONTHS,
        "seasonal_index": SEASONAL_INDEX
    })
    message = Content(parts=[Part(text="Generate Q1 FY25 revenue forecast: " + payload)])
    print("Revenue Forecasting Agent - ShopMax India")
    print("=" * 46)
    async for event in runner.run_async(user_id="cfo", session_id=session.id, new_message=message):
        if event.is_final_response() and event.content:
            for part in event.content.parts:
                if part.text:
                    print(part.text)

asyncio.run(main())

It gives the following output,

Revenue Forecasting Agent - ShopMax India
==============================================
Q1 FY25 REVENUE FORECAST (Apr-Jun) - Rs Lakhs

Category     Trend   Apr     May     Jun     Quarter
----------   ------  ------  ------  ------  -------
Mobiles      +30.2%  55.1    52.1    49.2    156.4
Laptops      +30.0%  23.9    22.6    21.4     67.9
TV           -17.3%  12.4    11.7    11.1     35.2
Audio        +32.4%  10.7    10.1     9.6     30.4
Accessories  +29.5%   6.8     6.4     6.1     19.3

TOTAL QUARTER FORECAST: Rs 309.2 Lakhs

GROWTH OPPORTUNITIES (trend > 20%)
Mobiles, Laptops, Audio, Accessories all show 29-32% growth trends.
Focus inventory and marketing investment in these categories.

ATTENTION REQUIRED - TV category (-17.3% trend)
TV revenue declining consistently. Review pricing and competition.
Consider promotional campaign or clearance sale for slow-moving models.

PRIOR QUARTER (Jan-Mar average): Rs 95.5L/month
Q1 FY25 forecast: Rs 103.1L/month (expected -10% seasonal dip
post-festival, but underlying trend remains positive).

The agent generated a category-level quarterly revenue forecast adjusted for post-festival seasonal slowdown in Apr-Jun. The TV category decline flag is particularly valuable - it surfaces a trend that might be hidden in a single total revenue number. Finance teams can use these projections directly in budget templates, and the CFO can communicate realistic expectations to the board while planning corrective actions for underperforming categories.

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