Segments and sizing
Who the 60 sampled creators are, cut five ways, and what that means for casting.
All figures measured from the dig Creator Panel: 60 Gen Z creators, 30,079 posts, Aug 2025 to Aug 2026.
One audience, cast from a narrow demographic band
54 of 60 are 20 to 29, 48 female, 6 male and 25 white, 17 asian, 11 black, 1 hispanic. Clustering the same 60 creators on their measured signal profiles put 51 of them in a single group, so the behavioural read and the demographic read agree: this is one audience.
Plan against one story. The useful segmentation for Spring is creator size and country, not audience type: 21 macro, 21 micro, 9 nano, 7 mega, 2 non influencer.
Five cuts of the panel
Gen Z by design. The panel was selected on category output within the age band.
Female-led casting matches where handbag conversation sits on this panel.
A mixed set, with white and asian creators carrying the largest shares.
Macro and micro creators are evenly matched, which gives Spring a two-tier casting option.
Panel composition by creator country is 21 US, 5 PH, 5 GB, 3 FR, 3 NG, 3 JP and a tail. Brand-level country cuts are not reported because brand-attributed posts concentrate almost entirely in the US.
Sizing here is panel reach
Sizing in this app is panel reach: creators, posts, views and followers. The panel composition page carries every reach figure, including 17.1M combined followers and 2.85B views across the window, now with the demographic columns for each creator.
For US population sizing, the next step is a purpose-built audience pull selected on audience behaviour rather than on category output. It is a separate study, and this read gives it a clear hypothesis to test: one audience, led by practicality, craft and earned luxury.
How these segments were resolved
Segments of the sampled creators themselves, by resolved demographic attribute. Attributes are resolved per creator, so counts below 60 mean the attribute is resolved for fewer creators rather than absent from the panel.
Segmentation was attempted by clustering all 60 panel creators on their measured signal profiles, testing three through seven segments and scoring each solution for separation. The best solution placed 51 of 60 creators in a single group and isolated one creator alone, with a separation score of 0.20 against a 0.25 threshold for usable structure. The panel is too small and too homogeneous to support segments, so none are reported.