Anticipating peaks and troughs in demand lets you adjust budgets, stock and teams before they arrive, rather than reacting afterwards. With your own data (sales, visits, brand searches), even an SME can make simple, useful forecasts without complex models.
The essentials in a definition
Demand forecasting with your own data uses sales, search and campaign history to anticipate peaks and troughs, and plan stock, budget and team. Simple models (seasonal averages, trends) usually suffice. Studies such as those from Think with Google help identify seasonality, Meridian offers mix modelling, and smart bidding already incorporates short-term forecasts.
Why forecasting helps
Badly distributed budgets waste money in weak seasons and lose sales in strong ones. Insufficient or excess stock costs margin. Teams without preparation lower quality at peaks. A forecast, even approximate, improves all these decisions.
Demand is rarely random: it has weekly, monthly and annual patterns.
What data to use
Sales or leads by week over the last two to three years, site visits, brand searches, a calendar of promotions and events (holidays, vacations, launches). If there is little data, use the last year and add external indicators, such as search trends.
Always record what happened: a promotion, a stock-out, a site failure.
A simple demand forecast
- Gather weekly data for two to three years
- Mark peaks, troughs and their causes
- Project with low, base and high scenarios
- Compare with reality every month
A simple method
Calculate the weekly average and the variation against the average in each week of the year. Adjust for business growth. Mark the peaks and troughs and their causes. Project to the next year, with a range (low, base, high scenario).
A spreadsheet is enough. What matters is to review it against reality every month.
Whoever knows when demand arrives prepares; whoever does not is caught by surprise.
Using the forecast
Plan budget by week or month, reinforcing strong-demand periods and saving in weak ones, if the return justifies it. Anticipate production, stock and teams. Prepare creatives and campaigns before the peak. Compare forecast and reality to improve the method.
Stay flexible: the forecast guides, it does not guarantee.
Decision table
| Method | Data | Usefulness |
|---|---|---|
| Seasonal averages | Sales by week or month, at least two years | Predictable peaks and troughs |
| Trend and growth | Recent evolution | Adjust the overall target |
| Leading indicators | Searches, visits, waiting lists | Early signs of change |
| Scenarios | Prudent, base, ambitious hypotheses | Planning with ranges |
Common mistakes to avoid
- Forecasting without enough history. With little history, the forecast is fragile.
- Ignoring one-off events. Promotions and failures distort averages; flag them.
- Treating the forecast as certainty. Build a safety margin.
How to measure the result
Compare forecast and actual monthly, and calculate the percentage error; record events that explain deviations.
Limits and exceptions
Forecasts fail in the face of unexpected events (crises, platform changes, competition) and in new businesses with little history. Over-fitting to past data can mislead. Treat the forecast as guidance, update it often and prepare alternative plans for scenarios in which demand departs far from the expected.
Illustrative example
Illustrative example: a shop analyses three years of weekly sales and identifies a demand peak in two seasons and a summer dip. It plans stock and budget in advance, and tracks leading indicators (brand searches and visits). When indicators rise earlier than expected, it brings the campaign forward, and avoids stockouts.
In practice: checkpoints
Start with two or three years of weekly sales and search data, add known promotions and holidays, and build a simple forecast by week. Compare the forecast with actual results each month and adjust.
Frequently asked questions
Do I need advanced tools?
No. A spreadsheet and organised data already give a useful forecast.
What if I have little data?
Use the last year and external indicators, with more caution.
How often should I update?
Monthly, comparing forecast and reality.
How much history do I need?
Ideally two or more years to capture seasonality; with less, use prudent scenarios.
Which tools should I use?
A spreadsheet with seasonal averages usually suffices; more complex models only pay off with lots of data.
Forecast demand with your own history, seasonal averages and leading indicators, using scenarios and a safety margin, and always compare with actuals.
Sources and recommended reading
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.