Codebook for app-store-screenshots-ledger-2026-09-22.csv
Source: the current US App Store screenshot sets of 37 iPhone apps, retrieved on September 22, 2026 through Apple's iTunes Search API (6.9-inch iPhone slot) and read from the first six frames of each. Screenshots are the property of their developers. Every column is the author's reading of the images.

app, seller, genre, app_store_id: as returned by the API.
sheet: which of the nine ledger sheets carries this set's frames (/static/images/blog/app-store-screenshots/ledger-01.jpg to ledger-09.jpg; four sets to a sheet, five on the ninth, in ledger order).
frames_in_set: screenshots in the 6.9-inch slot. frames_read: the first six (or all, when fewer).
first_caption_frame: the first frame carrying marketing text outside the interface; 0 when the set has no captions.
opener: the kind of caption on the first captioned frame. claim = a point of view (an outcome, a stance, a superlative, what the app is for); feature = a capability; title = the app's name and category and nothing else; none = no captions in the set.
headline: the lead line of that caption, set apart by size or weight; wordmarks and logos are not headlines. headline_words: whitespace-separated words in it. subline: smaller or lighter text under the headline, not counted.
caption_position_first_captioned_frame: top, bottom, or fill (type that occupies a frame with no interface).
caption_position_set: where the captions sit across the frames read. top; opens below = the first captioned frame carries its caption under the phone or logo and later frames carry theirs at the top.
proof_frames_1_2: kinds of proof on frames one or two: apple (an Apple-designated recognition: Apple Design Award, App of the Year, Editors' Choice, App of the Day, App Store Awards finalist), press (a publication quote or logo), rating (stars or a rating count), user (a user quote), claim (an unattributed accolade such as "award-winning"), none. proof_apple_designated: yes when any of it is Apple's. proof_detail: what and where.
compositions_per_frame: one code per frame read. P full straight device; T tilted device; Cb, Ct, Cs device cropped by the bottom, top, or side edge; Z zoomed detail crop; F floating card or element with no device; X text or logo only; H photograph, product shot, or hand-held device; M two or more devices; W watch only; L landscape device; R frameless or raw screenshot; O illustration, mascot, or 3D object. A trailing + marks a card floating over a device; it does not make a new composition.
distinct_compositions: the number of distinct codes, ignoring +. Three or more is what the article calls a composition that moves from frame to frame.
device: bezel (an Apple-style device frame), frameless (interface crops with no frame), raw (the screenshot is the asset), photo (a device in a photograph), mixed.
ground: dark, light, colored, gradient, photo, raw (no designed ground).
notes: what the columns cannot hold.