Predictive analytics sounds like something a large retailer does with a team of data scientists. In practice, a small South African business can do a surprising amount of it with a spreadsheet, twelve months of clean sales history and thirty minutes of thinking. The question predictive analytics answers is the one every owner asks: what is going to happen next, and how confident am I in that answer.
This session is the practical version, sized for a business that runs on a laptop rather than a data warehouse. Which forecasts are worth building, which are noise, and how to use them without pretending you have certainty you do not have.
The three forecasts that pay for themselves
Sales forecasting tells you what to stock, what to hire and what to spend on marketing. Cash forecasting tells you whether you can afford any of it. Customer churn or repeat purchase forecasting tells you which relationships are quietly ending, in time to save some of them.
You do not need a machine learning model for the first two. Twelve months of monthly data, a look at seasonality, and a modest allowance for outside factors gets you a forecast good enough to plan against. The macro backdrop that shapes it, including interest rates, unemployment and consumer confidence, is published quarterly by Statistics South Africa and the South African Reserve Bank.
What this session covers
- Building a monthly sales forecast that is useful. Simple methods that outperform sophisticated ones on small business data, including moving averages and basic seasonality.
- Forecasting cash rather than profit. Cash lags sales, and it is what actually decides whether you survive a quiet quarter. The horizon that matters is the next thirteen weeks.
- Spotting churn early. The buying pattern that says a repeat customer is about to leave, and what you can do about it before they go quiet.
- Confidence intervals in plain language. Why one number is dangerous and a range is honest, and how to plan for both the good and bad end.
- When to invest in better tooling. Signals that you have outgrown a spreadsheet forecast, and what the next step actually looks like.
- Avoiding the false precision trap. A forecast that is confident and wrong is worse than one that is uncertain and honest. How to talk about numbers you cannot fully guarantee.
The honest limits of predicting anything
Forecasts are wrong. The purpose is not to be right, it is to be less wrong than the guess you would otherwise make. An owner with a monthly forecast within twenty percent of reality has a better decision framework than one operating on gut feel, even if the forecast never lands exactly.
The second honest point is that predictive analytics amplifies whatever data you feed it. Clean records produce useful forecasts. Messy books produce confident nonsense. If your accounting is behind, fix that first.
Who should watch this session
- Owners who plan by extrapolating last month, and are surprised often.
- Retailers and product businesses making stocking decisions with real cash consequences.
- Service businesses trying to decide when to hire, where the risk of being early or late is high.
- Founders in front of investors or lenders who want to see forecasts that hold up to questions.
What to do after the session
Build the two forecasts this fortnight. A monthly sales forecast for the next six months, using your history and a written assumption about seasonality. A thirteen week cash forecast, updated weekly. Both in a spreadsheet, both simple enough to be maintained.
Then look at your five biggest recurring customers and check whether their order pattern is drifting. A ninety day repeat customer whose last order was one hundred and twenty days ago is telling you something. Our free templates and guides include cash flow and forecasting tools you can adapt, and if a forecast reveals a cash gap, our business funding pages cover options for working capital.
Frequently asked questions
Do I need software or AI to forecast for a small business?
Not at first. A well built spreadsheet outperforms most tools for a business with under a few hundred monthly transactions. Software earns its place when volume makes manual work unreliable.
How accurate should a forecast be?
Realistically, plus or minus twenty percent for sales at a monthly level, tighter for cash on a shorter horizon. Anyone claiming better accuracy for a small business is either lucky or overfitting.
What data do I need to start?
Twelve months of monthly sales, split by product line if possible. Bank statements for the same period. That is enough to build the two forecasts that matter.
How often should I update the forecast?
Sales monthly, cash weekly. The point is not the forecast, it is the discipline of comparing what happened against what you expected and asking why the gap exists.
Watch the session, then build the sales and cash forecasts this fortnight. Join the community to compare methods with other owners, or see the other sessions.
