Retail demand swings with holidays, promotions, and calendar quirks. Learn how to encode seasonality and trend so your forecasts don’t drift or overreact.
Retail demand often exhibits multiplicative seasonality; a common stabilizing transform is the ______.
square transform
min‑max scaling
z‑score standardization
log transform
A robust way to separate trend and seasonality that works on non‑stationary retail data is ______.
binary encoding
Naive seasonal mean only
plain k‑means
STL decomposition
Moving holidays like Easter are best handled by ______ in forecasting.
discarding those weeks
increasing AR order only
dummy or regressor indicators for holiday windows
weekly seasonality alone
To model multiple seasonalities such as weekly and yearly patterns, you can add ______ terms.
dropout layers
one‑class SVMs
polynomial kernels only
Fourier series
Price changes that co‑move with demand should be treated as ______.
exogenous regressors
targets for differencing
output constraints only
seasonal periods
When stockouts occur, recorded sales understate demand; a typical fix is to ______.
set demand to zero permanently
flag and adjust or impute constrained periods
remove weekly seasonality
lower the forecast horizon
A quick diagnostic for multiplicative seasonality is a plot of seasonal strength increasing with level; this suggests using ______ models.
purely additive in raw units
no seasonal components
Poisson with identity link only
additive on a log scale
Retail has strong day‑of‑week effects; to capture them in regression you’d typically ______.
difference twice only
shuffle dates to remove order
use one‑hot encoded weekday dummies
apply PCA on dates
When trend has abrupt shifts due to assortment changes, a method tolerant to breaks is ______.
random shuffling of dates
constant mean assumption
strict linear trend only
piecewise trend with changepoints
To avoid leakage when evaluating retail forecasts, you should split data using ______.
forward‑chaining time splits
random holdout across all dates
stratified by price only
LOOCV after shuffling
Starter
Revisit calendar effects and basic transforms to stabilize demand.
Solid
Nice—dial in holiday windows, multiple seasonalities, and price effects.
Expert!
Excellent—your retail models balance trend, promos, and calendar like a pro.
