Trend Analysis for Forecasting

STL Decomposition for Time Series Analysis 101 👇🏼 When working with time series data, one of the most powerful ways to understand underlying patterns is STL decomposition — Seasonal and Trend decomposition using LOESS. This is one of my favorite tools for articulating modeling decisions to stakeholders 🎯. What is STL? STL is a flexible method that breaks a time series into three main components: 🔹 Trend – the long-term direction of the series 🔹 Seasonal – repeating patterns that occur at fixed intervals 🔹 Remainder (Irregular) – what’s left after removing trend and seasonality Unlike traditional decomposition, STL uses LOESS smoothing, which makes it highly adaptable to complex, nonlinear patterns. Why use STL instead of classical decomposition? Compared to classical methods (additive or multiplicative), STL offers several advantages: 🔹 Works well when seasonality changes over time using a window function 🔹 More robust to outliers 🔹 Handles nonlinear trends better 🔹 Requires fewer strict assumptions about fixed patterns 🔹 Unlike classical decomposition, there is no loss of observations from the tails of the series (due to the trend smoothing) This makes STL a strong default choice for real-world, messy time series data. Key components in an STL plot: ✅ Actual (Observed): The original time series ✅ Trend: The smoothed long-term movement ✅ Seasonal: The repeating cyclic pattern ✅ Seasonally Adjusted: The series with the seasonal component removed (Observed − Seasonal) ✅ Irregular (Remainder): Random noise and unexplained variation left after removing both trend and seasonality Pro Tip: Overlay the irregular component standard deviation on the Actual plot. I use a range of ±2σ to ±3σ (orange) and bands above ±3σ to immediately spot points where a large variation is observed that the seasonal and trend components cannot explain. This makes it easier to diagnose potential outliers. #timeseries #forecasting #datascience

Forecasts are worthless if they don’t drive action. This document shows how to turn forecast errors into insights: # 1 - Compare Forecast vs Actual Pattern, Not Just Values Look for trend breaks: promotions, seasonality shifts, competitive actions Insight: shows whether the model or the business behavior changed # 2 - Separate Volume Error from Mix Error Your total forecast may be right but SKU mix is wrong Insight: points to cannibalization, launches, or customer preference shifts # 3 - Slice the MAPE (forecast error) MAPE at total level hides the real problem; slice by SKU, region, channel, and planner Insight: find where the system is breaking, not the average # 4 - Track Bias Consistently MAPE shows how much you miss; bias shows how you think Insight: positive bias = optimism; negative bias = fear of stockouts # 5 - Connect Error Spikes to Events Overlay error trend with business events; launches, stockouts, price changes and map everything Insight: turns disconnected numbers into cause-and-effect stories # 6 - Use FVA (forecast value added) to Check If Adjustments Helped or Hurt Measure whether human overrides improved or worsened accuracy Insight: helps remove emotional adjustments from the process # 7 - Build an Error Heatmap One view showing where the biggest misses are by SKU, month, region Insight: quickly identifies where planning attention is needed # 8 - Weekly Error Deep Dive Pick the top 5 SKUs with the biggest misses; ask: “what changed?” and “who owns the correction?” Insight: makes forecasting a feedback loop, not a ritual Any others to add?

🌟 Day 4 – Forecasting Basics How do you know how many calls, chats, or emails to expect tomorrow? 👉 That’s Forecasting—the foundation of Workforce Management (WFM). At its core: Forecast = History (baseline) + Trend + Seasonality + Event Adjustments + Judgment 🔍 Why Forecasting matters (in plain terms) Without a forecast, everything else is guesswork: Capacity Planning: You can’t know how many people you need. Scheduling: You don’t know which hours need extra coverage. Real-Time Management: You can’t tell if you’re off-track or on-target. Reporting: You can’t measure if the plan was realistic. 🧭 What goes into a good forecast - Baseline History- Start with apples-to-apples data (same channel, same handle type). Use the closest comparable days (e.g., last 6–8 Mondays for next Monday). - Trend - Are volumes growing or shrinking month over month? Apply a gentle up/down adjustment (e.g., +2% MoM). - Seasonality - Intra-week: Mondays heavier than Fridays? Intra-day: 11:00–13:00 peak every day? Keep a pattern profile so you can shape the daily forecast by 15/30-minute intervals. - Events & External Drivers - Holidays, promos, product launches, price changes, outages, weather. Each can add/subtract volume. Use an uplift/deflation percentage based on past, similar events. - Judgment & Business Intel - Talk to Marketing, Product, and Ops. Numbers + context beats numbers alone. 🧪 Mini example (numbers you can follow) Baseline: Last 4 Mondays ≈ 10,000 calls This Monday is a holiday: Past similar holiday = +15% uplift Marketing email scheduled 10:30: Past emails add +8% for 2 hours Day total: 10,000 × 1.15 = 11,500 base for the day Apply short, time-boxed +8% uplift 10:30–12:30 to those intervals only. Shaping by intraday pattern (illustration): If 12% of Monday’s calls typically arrive 11:00–12:00, that hour ≈ 11,500 × 12% = 1,380 calls (then layer the +8% marketing effect inside that window). You now have a time-sliced forecast (by 15/30/60-min intervals), not just a day total—this is what schedulers need. 🎯 How to check if your forecast is any good MAPE (Mean Absolute Percentage Error): Average error size. Bias (Over/Under): Do you consistently over- or under-forecast? Hit Rate: % of intervals within a target error band (e.g., ±10%). Track these by channel and by interval, not just daily totals. A perfect day can still hide ugly peaks. 📌 Takeaway: Forecasting is educated prediction—never perfect, always essential. Get close, shape it by interval, adjust for real-world events, and learn fast from misses. That’s how you keep customers happy, and costs controlled. #WorkforceManagement #WFM #Forecasting #ContactCenter #CustomerExperience #BusinessEfficiency #Scheduling #CapacityPlanning #RTA #OperationsExcellence #Analytics #DataDriven

𝗧𝗵𝗲 𝘁𝗿𝗲𝗻𝗱 𝘄𝗮𝘀 𝗵𝗶𝗱𝗶𝗻𝗴 𝗶𝗻 𝗽𝗹𝗮𝗶𝗻 𝘀𝗶𝗴𝗵𝘁. 𝗡𝗼 𝗼𝗻𝗲 𝗲𝗹𝘀𝗲 𝘀𝗮𝘄 𝗶𝘁. But one sock brand did. A sock brand I spoke with spotted a tiny shift in consumer behavior—𝘣𝘦𝘧𝘰𝘳𝘦 it blew up—and turned it into their lowest CPA campaign ever. The wildest part: they found the trend using a 100% free tool. 𝗪𝗵𝘆 𝗱𝗶𝗱 𝗶𝘁 𝘄𝗼𝗿𝗸? 𝗧𝗶𝗺𝗶𝗻𝗴. The biggest difference between a winning DTC brand and a struggling one isn’t budget—it’s timing. 👉 Move too late, and you’re in a price war. 👉 Move early, and you print money. Here’s what happened: On a call last week, I casually mentioned Pinterest Predicts—because Pinterest has an 80% accuracy rate in forecasting viral trends months before they peak. 𝗔𝗞𝗔: 𝟴𝟬% 𝗼𝗳 𝘁𝗵𝗲 𝘁𝗶𝗺𝗲, 𝗣𝗶𝗻𝘁𝗲𝗿𝗲𝘀𝘁 𝗰𝗮𝗻 𝘀𝗲𝗲 𝘄𝗵𝗮𝘁 𝘁𝗿𝗲𝗻𝗱𝘀 𝘄𝗶𝗹𝗹 𝘁𝗮𝗸𝗲 𝗼𝗳𝗳 𝗯𝗲𝗳𝗼𝗿𝗲 𝗮𝗻𝘆𝗼𝗻𝗲 𝗲𝗹𝘀𝗲. This sock brand took action. We noticed that "cherry" patterns were quietly trending up. Within 24 hours, they launched a cherry sock collection, tied their creative to the aesthetic, and started testing new ads. 🚀 Today: → The product/ad combo is their 3rd highest spender → Lowest CPA across all campaigns → Crushing their new customer acquisition costs 𝗧𝗵𝗶𝘀 𝗶𝘀 𝘄𝗵𝘆 𝗰𝗼𝗻𝘀𝘂𝗺𝗲𝗿 𝘁𝗿𝗲𝗻𝗱 𝗺𝗶𝗻𝗶𝗻𝗴 𝗶𝘀 𝗮 𝗰𝗵𝗲𝗮𝘁 𝗰𝗼𝗱𝗲. 🚀 Brands that move fast have: ✅ Ads that convert immediately ✅ Higher margins (first-mover advantage) ✅ Less reliance on discounts & promos 🐢 Brands that wait too long: ❌ Spend more to compete ❌ Launch when the market is saturated ❌ End up in a price war just to survive 𝗪𝗮𝗻𝘁 𝘁𝗼 𝘀𝗽𝗼𝘁 𝘁𝗿𝗲𝗻𝗱𝘀 before they blow up? Here are 4 free tools I use daily to track consumer trends before they hit mainstream: 📌 Pinterest Trends/Predicts → Forecasts viral trends months in advance 🛍️ Etsy & Amazon Search Data → Shows what niche buyers are actively searching for 🎥 TikTok Comments → Raw, unfiltered consumer obsession in real time 💬 Reddit Threads → Where micro-trends are born Most brands have access to this data—but never use it (or aren’t using it enough). 𝗕𝗲𝗰𝗮𝘂𝘀𝗲 𝗴𝗿𝗼𝘄𝘁𝗵 𝗶𝘀𝗻’𝘁 𝗮𝗯𝗼𝘂𝘁 𝘀𝗽𝗲𝗻𝗱𝗶𝗻𝗴 𝗺𝗼𝗿𝗲. 𝗜𝘁’𝘀 𝗮𝗯𝗼𝘂𝘁 𝗹𝗮𝘂𝗻𝗰𝗵𝗶𝗻𝗴 𝘁𝗵𝗲 𝗿𝗶𝗴𝗵𝘁 𝗮𝗱 𝗮𝘁 𝘁𝗵𝗲 𝗿𝗶𝗴𝗵𝘁 𝘁𝗶𝗺𝗲.

Most demand forecasts are built on a single method chosen by habit. Simple moving average because it is familiar. Exponential smoothing because someone set it up years ago. The method stays even when the data changes. The problem is that no single forecasting method works best for every demand pattern. Stable demand with no trend behaves differently than demand with a clear upward trend. Seasonal products need a completely different approach than items with flat, irregular consumption. Using the wrong method does not just produce a less accurate forecast. It produces systematically biased safety stock levels, reorder points, and procurement timing. The Demand Forecasting Tool runs five methods simultaneously on your historical data: Simple Moving Average, Weighted Moving Average, Single Exponential Smoothing, Holt's Double Exponential Smoothing for trending data, and Holt-Winters Triple Exponential Smoothing for data with both trend and seasonality. For each method, it automatically optimizes the smoothing parameters to minimize error on your specific data rather than using defaults. It then scores all five methods against your history using three error metrics: MAPE, MAD, and MSE. The best-fit method is identified automatically and used to generate the forward forecast. The Safety Stock tab takes the forecast error directly from the best method and calculates safety stock and reorder point across four service level targets using the standard formula. Paste your data, set your lead time and service level, and get a defensible stocking recommendation in under two minutes. Link in the comments. #SupplyChain #DemandForecasting #InventoryManagement #ProcurementAnalytics #CPSM

Forecasting is hard. Finding analysts who do it well is even harder. Too often, I see forecasting either: 1. Overcomplicated: Applying complex ML models just to predict a moving average (?!), or 2. Oversimplified: Running regressions without understanding what the coefficients even mean. I personally use 4 forecasting methods to model a range of outcomes, from conservative to aggressive: 1. ARIMA - Smooths time series data, w/o seasonality adjustment. 2. SARIMAX -  Like ARIMA, but accounts for seasonality. Likely to be the safest and conservative forecast. 3. Prophet -  Captures non-linear trends and seasonality. Often the most accurate. My favorite model for growth forecasts. 4. Manual Projection – aka Olga's secret, overly complicated manual projection. I plot every available metric’s historical D/D, W/W, M/M, and Y/Y % change and analyze their: (a) correlations and relationships (b) seasonal thresholds. It takes ages to complete, but it delivers the most precise forecast. If done right. If I can account for everything the teams are doing. Which is rarely the case. 😬 When reporting, I typically present only Prophet alongside my Projection, keeping ARIMA and its variations for myself as checks. There are many time series models out there: MA, AR, ARMA, ARIMA, SARIMA, Exponential Smoothing, VAR, and more. Forecasts are fun.

𝗦𝘁𝗲𝗽 𝗻𝘂𝗺𝗯𝗲𝗿 𝟭 in any good projection: calculate future Revenue. As accurate as possible. That's mandatory!! 𝗣𝗼𝗽𝘂𝗹𝗮𝗿 𝗠𝗲𝘁𝗵𝗼𝗱𝘀 ✔️Historical Trend Analysis - Leveraging past performance to predict future trends. ✔️Market Analysis - Understanding market segments and potential impacts on revenue. ✔️Customer Segmentation - Analyzing different customer groups to tailor marketing and sales strategies. ✔️Sales Funnel Analysis - Monitoring progression through the sales funnel to anticipate revenue generation. ✔️Product Lifecycle Analysis - Assessing the stages of a product's life to forecast sales and revenue. ✔️Econometric Models - Using statistical methods to forecast revenue based on economic and market variables. 𝗢𝘁𝗵𝗲𝗿 𝗶𝗺𝗽𝗼𝗿𝘁𝗮𝗻𝘁 𝗺𝗲𝘁𝗵𝗼𝗱𝘀 ➡️ Driver-Based Forecasting: Focusing on key business drivers like unit sales, market share, or operational efficiency, this method provides a granular view of forecasted revenue, allowing for more targeted strategy adjustments. ➡️ Rolling Forecasts: Instead of static annual forecasts, rolling forecasts update throughout the year to reflect real-time market conditions and business outcomes, providing a more dynamic financial outlook. Curious to know how you all manage forecasting? What methods do you find most useful?

Here are 5 machine learning algorithms used for FP&A and #finance time series analysis: ✅ ARIMA/SARIMA: Forecast future revenues and expenses by identifying trends and seasonality. ✅ LSTM: Analyze complex patterns in cash flow or sales data to improve financial planning. ✅ Prophet: Handle unpredictable markets and still make reliable forecasts. ✅ GARCH: Assess and predict market volatility to make more informed investment or budgeting decisions. More detail below ↓ 1. ARIMA (Auto-Regressive Integrated Moving Average) ARIMA helps predict future values by analyzing past data to identify patterns like trends or seasonality. For example, you can use ARIMA to forecast next year’s monthly revenue by recognizing historical trends and seasonal variations, such as higher sales during holiday seasons. 2. LSTM (Long Short-Term Memory) Networks LSTM is an artificial intelligence technique that learns from past data and remembers long-term patterns. It can be used in FP&A to forecast cash flow by identifying recurring inflows and outflows over time, like specific project payments or seasonal cash patterns. 3. SARIMA (Seasonal ARIMA) SARIMA extends ARIMA by incorporating seasonality, making it ideal for forecasting data with regular patterns. For example, you can predict quarterly expenses more accurately if certain quarters have consistently higher costs due to contracts or seasonal demand. 4. Prophet Prophet, developed by Facebook, handles missing data and outliers well, making it useful for complex datasets. To get the code and example for implement it, go here: https://lnkd.in/eJKcHzqU You could use Prophet to forecast annual sales even when your data is incomplete or affected by irregular events like economic shifts. 5. GARCH (Generalized Autoregressive Conditional Heteroskedasticity) GARCH models volatility and is great for predicting how much financial data varies over time. You can apply it in FP&A to assess and predict the volatility of stock prices in your investment portfolio, helping in risk management and budgeting.

Want to use machine learning for time series forecasting? The best models will identify the drivers of trends. I once worked with a KPI like the image below. My ML model identified a serious problem. First, let's establish a working definition of "trend" when it comes to time series forecasting: The tendency of the KPI to increase/decrease over time. Like the image above, my real-world KPI exhibited a strong upward trend. Additionally, as shown in the image above, the trend was linear (i.e., a straight line). Finance loved this KPI because it could be easily forecasted with high accuracy. Executives loved this KPI because it kept going up and up. I didn't like it all. The problem was that traditional forecasting techniques rely only on the historical KPI values. These forecasting techniques may implement additional calculations (e.g., moving averages) to enhance accuracy. However, these calculations are based solely on historical KPI values. So, it's no wonder that Finance was able to easily forecast the KPI. However, I wanted to know what the drivers of the KPI were. Enter machine learning forecasting models. Machine learning forecasting models can not only use historical KPI values, but can also include any other data that might impact KPI values: Month of the year Day of the week Economic data Promotions Weather Etc. In the case of my KPI, I was examining activities originating from the marketing team (e.g., promotions and digital ads). That's when my ML forecasting model uncovered a serious problem. The ML model identified that the primary external driver of KPI values was the marketing team's digital advertising spend. I dove into the data and found that digital ad spend increased over the same time period as the KPI. However, the digital ad spend was increasing at a higher rate. The digital ads were experiencing diminishing returns. We were burning budget to prop up the KPI. That's the power of ML forecasting models. BTW - Millions of professionals now have access to the tools to craft powerful ML forecasting models. Python in Excel is included with M365 subscriptions and provides access to libraries such as scikit-learn and statsmodels. Everything you need to go far beyond Microsoft Excel's forecast worksheet.