Time series and forecasting · insight-1238
Check that the method tolerates gaps, and model systematic closures with dummies rather than interpolating them
- Evidence type
- documented fact
- Difficulty
- easy
- Roles
- Data analyst, Data scientist
- Industries
- retail, healthcare, public-sector
Summary: situation, action, reason, boundary
Situation
A store's daily sales are missing on public holidays when it is closed, and the pipeline either fails inside an exponential smoothing fit or silently interpolates the closures.
Action
Separate systematic gaps from random ones. Model closures with holiday and next-day dummies in a method that accepts gaps, such as dynamic regression with ARIMA errors; for random gaps interpolate with a model or train only on data after the last gap if enough remains.
Reason
Systematic missingness is informative about the forecast, while random missingness is not. Methods differ in whether they can fit through gaps at all.
Boundary
Retail, branch and clinic series with closures, sensor outages or batch-load failures.
Symptom or risk
Interpolating a closure invents sales that never happened, and a method that cannot handle gaps either stops or quietly uses a distorted series.
Exceptions
The method-support list depends on the implementation; check your library instead of assuming it matches the textbook's software.
Trade-offs
Shortening history to a gap-free tail loses seasonal cycles; interpolation keeps them but adds a model's assumptions.
Pitfalls
Filling closures with zero and letting the model learn that the shop is slow on holidays.
Lessons and concepts
Claims and their sources
FPP3 states that naive methods, ARIMA, dynamic regression and neural network autoregression handle missing values while ETS and STL do not, and recommends modelling systematic missingness such as holiday closures with dummy variables.