# Footwear Advertising Benchmarks 2026 — data dictionary

Published under CC BY 4.0. Attribute to Scaletrics with a link to
https://www.scaletrics.com/blog/footwear-advertising-benchmarks

## Columns

| Column | Meaning |
|---|---|
| `table` | Which table on the page the row belongs to. |
| `metric` | Machine-readable metric name. |
| `segment` | The slice the value applies to, including any held-constant inputs. |
| `value` | The value as published. Units are in `unit`, never inferred. |
| `unit` | USD, percent, ratio, multiple, USD/order or USD/return. |
| `tier` | reported, modelled, derived or assumed. There are no untagged figures. |
| `source_id` | Key into the `sources` array of the JSON file. Empty for derived rows, which are reproducible from the model instead. |
| `period` | The period the figure covers. |
| `notes` | Held constants, denominators and caveats. |

## Reproducing the derived rows

```
CM% = m(1-r) - (1-m)*r*(1-k) - p - (f + r*f_r)/A
break-even reported ROAS = 1 / CM%
max affordable CPA       = CM% * A
max affordable CPC       = CM% * A * CVR_nonbrand
contribution-margin ROAS = reported ROAS * CM%
```

Every `derived` row is reproducible from the archetype inputs in the same
file plus this formula. If a row does not reproduce, it is an error and we
want to hear about it.

## What this dataset is not

Scaletrics does not operate a proprietary footwear account panel. Rows tagged
`reported` and `modelled` belong to the publishers named in `sources`; rows
tagged `derived` are our arithmetic on them; rows tagged `assumed` are
modelling inputs with no available source. Reusing a `derived` or `assumed`
row as though it were measured footwear data would misrepresent it.