> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getnovaplan.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Variance Guide

> How to read a variance, when a percent is material, how to read a waterfall, and how to drill to root cause.

A variance is the difference between what you planned and what happened. This guide explains how to read one correctly, how to judge whether it matters, how to follow a waterfall, and how to chase a gap down to its cause. It applies anywhere variance appears: [Comparison Center](/reporting/comparison-center), [Briefs](/reporting/briefs), and [Financial Performance](/reporting/financial-performance).

## Reading a Single Variance

A variance is always comparison minus base. Base is what you measure against (budget, target, or a prior version); comparison is what you are testing (actuals or the live forecast).

| You see | Read it as |
| - | - |
| Variance (\$) | The absolute gap in base currency: comparison minus base |
| Variance (%) | The gap as a share of base, so size is comparable across lines |
| Color (cost direction) | Whether the move helps or hurts, not just its sign |

<Warning>
  Sign alone does not tell you good or bad. On a cost line, spending more than budget (a positive variance) is unfavorable. On a revenue line, earning more than plan (also a positive variance) is favorable. Always read variance by cost direction.
</Warning>

## When Is a Percent Material?

There is no single threshold, but a practical way to judge materiality is to weigh three things together: the percent, the absolute dollars, and whether the line is volatile by nature.

* A small percent on a large line can still be a large dollar amount worth explaining.
* A large percent on a tiny line is often noise, not a story.
* A line that swings every period (timing-sensitive accruals, for example) tolerates a wider band than a stable, predictable line.

As an illustrative rule of thumb only, many finance teams flag anything beyond roughly 5 to 10 percent on a material line for review, tighten that band on stable lines, and loosen it on naturally lumpy ones. Use your own plan's tolerance; the point is to read percent and dollars and line behavior together, never percent alone.

<Note>
  A 12% variance is not automatically bad. On a small or volatile line it may be expected. On a large, stable cost line it is usually worth a look. Judge it against the dollar size and the line's normal behavior.
</Note>

## Reading a Waterfall

A waterfall (also called a bridge) decomposes a single change into named movements, so an ending number becomes a story. Read it left to right:

| Bar | What it represents | Color |
| - | - | - |
| Beginning anchor | The starting balance (BoP) | Blue |
| Positive movements | Drivers that add (new, expansion, favorable variances) | Green |
| Negative movements | Drivers that subtract (churn, contraction, unfavorable variances) | Red |
| Ending anchor | The closing balance (EoP) | Blue |

The green bars build up from the blue beginning anchor, the red bars pull back down, and they net to the blue ending anchor. A correct waterfall always reconciles: beginning plus the green minus the red equals the ending. If it does not net, the beginning balance is usually being mis-counted as a movement instead of anchoring the bridge.

## Common Variance Drivers

When a cost variance is decomposed (for example in a headcount brief), the named buckets explain the cause:

| Driver | What moved |
| - | - |
| Rate | Cost per unit changed (cost per person, price per seat) |
| Volume / FTE | How many units changed (more or fewer people, more or fewer deals) |
| Mix | The blend shifted toward more or less expensive components |
| Fringe | Loaded-cost surcharges on top of base (benefits, employer cost) |
| FX | Currency translation moved the base-currency figure |
| Residual | The part not attributed to a named driver |

<Warning>
  A large residual bucket is a signal, not an answer. It usually means the variance is concentrated in a dimension the decomposition did not name, so drill into it rather than accepting it.
</Warning>

## Drilling to Root Cause

<Steps>
  <Step title="Start at the headline">
    Read the total variance and its direction. Confirm it is material by dollars and percent together.
  </Step>

  <Step title="Find where it concentrates">
    Pivot by department, account, or entity in [Data Explorer](/reporting/data-explorer) to see which slice carries the gap.
  </Step>

  <Step title="Drill to the transactions">
    Open the line to the individual transactions, including vendor and memo, so the cause is a specific posting, not an abstraction.
  </Step>

  <Step title="Name it">
    Tie the gap to a driver (rate, volume, mix, fringe, FX) so it can be explained in one sentence, then capture it in a [Brief](/reporting/briefs).
  </Step>
</Steps>

## If the Numbers Look Off

* A variance that will not reconcile is almost always a scope mismatch: confirm both sides cover the same entity and period.
* A blank on one side means that account or department is not mapped in that version, not a true zero.
* Two surfaces that disagree usually differ on universe (for example GL P\&L versus a domain cost model), which overlap but measure different things, so the gap is the universe difference, not an error.

<Tip>
  When you have read the variance and named the driver, generate a [Brief](/reporting/briefs) to write the story, or return to [Comparison Center](/reporting/comparison-center) to widen or narrow the scope.
</Tip>