Sales Trends and Forecast v.19

Build reports your way

• Create as many reports as you need, each based on Odoo sale order lines
• Save and reuse them, switch instantly, or experiment in real time without saving
• Choose the interval: yearly, quarterly, monthly, weekly, or daily
• Predict an unlimited number of future periods
• Show data as a table, line graph, or bar chart, or export it to Excel
• Grant access to any salesperson via the Sales > Sales Forecast group

Filter, group & measure

• Filter by product category, template, variant, sales team, salesperson, country, industry, customer, or company
• Add any custom rule with Odoo's dynamic domain constructor
• Measure total, untaxed total, amount to invoice, amount invoiced, gross weight, or volume
• Run several grouped series at once — by category, product, order state, team, country, customer, and more
• Show statistical measures: count, mean, standard deviation, min, max, and quartiles
• Shorten large figures (e.g. 192,217 → 200k) or display them precisely

Powered by proven statsmodels

• Forecasts run on the trusted Python statsmodels package
• Six models included: AutoReg, ARDL, ARIMA, SARIMAX, Holt-Winters (HWES), and Simple Exponential Smoothing (SES)
• Every model exposes its own parameters, with an in-app Help tab explaining each one
• Keep and reuse an unlimited number of configured models

You stay in control

• Nobody understands your sales tendencies better than you — choose the model and parameters that fit your market
• Compare several models side by side to see which predicts best
• Adjust and re-run flexibly from the forecast manager
• Expect to experiment: some parameter sets return no or unrealistic predictions, and that is perfectly normal

Backtest against real history

• Forecast a past period you already have actuals for and overlay the prediction against real values
• Turn on Hypothesis Testing to see how well a model and its parameters fit
• Example: train on 2017–2021, forecast 2022, then compare with what actually happened

Compare models with confidence

• Test several statistical models at the same time
• Instantly see which approach tracks reality most closely
• Decide whether a model is good enough or whether to keep experimenting

Enough history

Trends need a solid volume of past sales; a few recent days can't reveal one. Make sure every filter and grouped series has enough data.

Real, non-chaotic trends

Statistics extend patterns that already exist. Random, trendless data won't predict well, so know which assumption (e.g. seasonality) you are testing.

A statistical mindset

There is no one-click "general" forecast. Value comes from a basic grasp of statistics and a willingness to experiment with each unique case.