Rollforward & time series

Overview

Use this category when you need time-based logic such as projections, running totals, moving windows, and timing effects. These functions operate along the time axis to help you forecast values into the future, build cumulative KPIs, smooth values over time, or model delayed impacts.

Start here if…

• You want to project values forward using drivers → start with ROLLFORWARD_ADVANCED.

• You want cumulative KPIs like YTD.

• You want rolling windows such as moving average or moving sum.

• You want to model timing effects (delays, adoption curves, stretched profiles).

• You need a previous-period building block for loop-logic.

Not here if…

• You want standard growth comparisons (YOY/…, deltas) → see Compare periods

• You mainly need to reshape dimensionality → see Dimensionality & hierarchies

Mental model

• Projection: calculate future periods based on a base and driver logic.

• Cumulative: sum (or product) from a start point to “now” within a period or across all time.

• Rolling window: aggregate across a sliding window of prior periods.

• Timing effects: shift or spread the impact of a change across time.

• Loop logic: reference previous period values with an explicit base case.

Projection (driver-based)

ROLLFORWARD_ADVANCED

Use when: you want the most flexible driver-based projection, including optional behaviors such as seasonal/average projection or aging-style logic.
Example: ROLLFORWARD_ADVANCED('Base', 'Driver1', 'Driver2', …)

ROLLFORWARD

Use when: future values depend on additive driver impacts.
Example: ROLLFORWARD('Base', 'Driver1', …)

ROLLFORWARD_MUL

Use when: drivers represent multiplicative effects.
Example: ROLLFORWARD_MUL('Base', 'Driver1', …)

Cumulative totals (to-date and running)

RUNNINGSUM / RUNNINGPROD

Use when: you want cumulative values across the full time axis (no periodic reset).
Example: RUNNINGSUM('Node'), RUNNINGPROD('Node')

MTD / QTD / YTD / WTD

Use when: you want cumulative totals that reset within each month/quarter/year/week.
Example: MTD('Node'), QTD('Node'), YTD('Node'), WTD('Node')

Rolling windows (smoothing)

MOVINGAVG

Use when: you want a moving average over a defined time window.
Example: MOVINGAVG('Node', "TimeLevel", WindowSize)

MOVINGSUM

Use when: you want a moving sum over a defined time window.
Example: MOVINGSUM('Node', "TimeLevel", WindowSize)

Timing effects (delays and profiles)

TIMELAG

Use when: an effect should show up after a delay at a chosen time granularity.
Example: TIMELAG('Node', 'TimeLagNode', "TimeLevel")

STRETCH_CURVE

Use when: you want to stretch or compress a time profile along the time axis.
Example: STRETCH_CURVE('Node', 'StretchNode')

SCHEDULEPROJECTS

Use when: you want to distribute a KPI over time based on project timing.
Example: SCHEDULEPROJECTS('ProjectKPI', 'TimingNode')

Loop logic

PREVIOUS

Use when: you need previous-period logic with an explicit base-case value.
Example: PREVIOUS('Node', "Level", 'Basenode', "Basevalue")

Statistical forecasting

ARIMA

Use when: you want a time-series forecast based on historical patterns.
Example: ARIMA('Node', …)

Functions in this category

Choosing between similar functions

• ROLLFORWARD_ADVANCED vs ROLLFORWARD / ROLLFORWARD_MUL

  • Use ROLLFORWARD_ADVANCED for richer projection options and more control.

  • Use ROLLFORWARD for additive drivers; ROLLFORWARD_MUL for multiplicative drivers.

• MTD/QTD/YTD/WTD vs RUNNINGSUM

  • Use to-date functions for totals that reset within periods.

  • Use RUNNINGSUM for a running total across the entire time axis.

• MOVINGAVG/MOVINGSUM vs to-date functions

  • Use moving functions for rolling windows.

  • Use to-date functions for cumulative totals within the current period.

• PREVIOUS vs Compare periods shortcuts

  • Use PREVIOUS when you need a base case and loop-style logic.

  • Use PM/PY/PQ/PW for simple previous-period references.

• TIMELAG vs STRETCH_CURVE

  • Use TIMELAG for delayed effects.

  • Use STRETCH_CURVE for stretching/compressing a profile.

Pitfalls & troubleshooting

• Future periods are missing in projections: confirm the base series has values in the last actual period and drivers are defined for the projection range.

• Projection results look wrong: validate the base node and each driver node separately, then add drivers back one by one.

• Cumulative totals are too large: confirm whether you used RUNNINGSUM (no reset) versus MTD/QTD/YTD/WTD (resets).

• Moving windows look shifted: verify the time level and window definition and confirm with spot checks in Data preview.

• PREVIOUS behaves unexpectedly at the first period: verify the base case and handle missing values explicitly if needed.

• Timing effects feel off: confirm the time granularity and whether you need a lag (TIMELAG) versus a profile stretch (STRETCH_CURVE) or project scheduling (SCHEDULEPROJECTS).

Related sections

• Compare periods standard growth comparisons and deltas

• Dimensionality & hierarchies prepare output shape; create empty simulation space with EXPAND(0, ...)

• Operators arithmetic and conditions used in forecasting logic

• Formula basics evaluation order and notation

• Troubleshooting guide missing values and unexpected results