How the reading works

The two halves, and which is which

The automated half. A photograph is passed to a feature-recognition model whose only job is to say which construction features from the reference table it can see: the width of a lapel, the height of the gorge, whether the hip pocket has a flap, whether the fly closes with buttons, a metal zip or a nylon coil. It proposes features. It proposes no dates.

The half that is not a model at all. Every interval comes from a reference table published on this site. The overlap between the intervals, and the detection of any pair that cannot overlap, are computed in your browser from that table.

This split is deliberate. It means a reading can be recomputed by hand from published numbers, and therefore that it can be checked and disputed. A date produced by a model that cannot show its working is a date nobody can argue with, and dating garments is a field in which being argued with is the normal and healthy state of affairs.

Dates are always intervals, never a year

The service reads construction features and gives the period ranges each one points to. Dates are always given as ranges and never as a specific year.

Each feature in the table carries an interval because that is what the evidence supports: a habit enters a workroom, spreads, and leaves, over years. The result you receive is the span where enough of your garment’s features can all be true at once.

Where several short intervals coincide the span is narrow, and the result says which features made it narrow. Where the features are long-band ones the span stays wide, and the result says that too rather than pretending to a precision it has not got.

This is not an authentication service

The service does not judge genuineness. It does not distinguish an original from a reproduction, does not identify a brand, and does not determine where an item was made.

Where several clues point to a later making or a later alteration — a cut that reads one way and a manufacturing method that cannot be as early as the cut — that is reported as an observation about construction and is not an authentication finding. The service does not decide which part of the clash is the original and which is not, because deciding that requires handling the object.

Nor is a reproduction treated as a lesser thing. Making a garment now to a pattern from a century ago is a craft, and a careful reproduction is a real object.

No valuation of any kind

The service gives no valuation of any kind. Period is only one of many factors in value. Condition, size, cloth, maker, completeness, provenance and demand all bear on it, and this service has a view on none of them.

It also has no view on whether you should buy or sell anything, and it will not be given one.

What distorts the clues, and is invisible to the service

Reproductions, later alterations, regional variation and unconventional cutting all distort the clues, and the service has no way of knowing that they have done so.

A garment made later to an earlier pattern reads as the earlier pattern.
Alterations during a garment’s life — a shortened hem, replaced buttons, a new lining, a re-cut waist — are read as if they were original.
A provincial workroom may be cutting to a shape a city house abandoned a decade earlier.
Bespoke, theatrical and one-off cutting sits outside the ordinary run the table describes.
Photographs flatten. Padding, cloth weight and the depth of a roll line are all judged less reliably from an image.

Every result carries this list. It is not fine print; it is part of the answer.

No people, and no styling advice

The service accepts no photographs containing people, generates no images of people, and offers no styling advice. Uploads that appear to contain a person are refused before any reading is produced, and nothing produced here contains a human figure. Whether a garment would suit you, fit you or go with anything else is not a question this service answers.

The limits travel with the reading

Every reading, on every plan, carries the notice that this is a reading of construction rather than of provenance, that it is not an authentication and not a valuation, and that a specialist should be consulted for anything being bought or sold. Every reading also carries its own list of what it cannot know.

Exports on the free plan carry a mark saying which plan produced them. That mark sits in the margin and is never placed over the notice or over the list of what the reading cannot know. A limitation that can be cropped out is not a limitation.

What you must not use a reading for

The terms of service prohibit using anything produced here in a sale description, an auction catalogue entry, an authenticity claim or an insurance valuation, or as the basis for buying or selling anything. For those purposes, consult a specialist.

A reading must also not be relied on for any decision about medical care, legal rights, financial or credit matters, safety, employment, housing, immigration or the administration of justice. The full list is in the acceptable use policy.

A worked example, including a clash

Four features. Three of them agree; the fourth cannot overlap any of them. The clash is reported at the top of a result, not buried below it.

Read this timeline as text

Lapel width — Notched lapel over 4 in (10 cm) — points to 1935–1952.
Shoulder line — Square, heavily padded shoulder extending past the joint — points to 1938–1952.
Hem — Permanent turn-ups — points to 1905–1960.
Fly — Nylon coil zip fly — points to 1965–2000.
The features that agree overlap across 1938–1952. That interval is the reading.
Clash: Lapel width points to 1935–1952 while Fly points to 1965–2000. These two intervals never meet.
Clash: Shoulder line points to 1938–1952 while Fly points to 1965–2000. These two intervals never meet.
Clash: Hem points to 1905–1960 while Fly points to 1965–2000. These two intervals never meet.

The three that agree leave 1938–1952. The nylon coil zip cannot be earlier than 1965–2000, because it is a manufacturing method rather than a style. The honest reading is that either the garment was made later to an earlier pattern, or the fly was replaced at some point. Which of those it is, this service does not claim to know.

The reference table, in full

Every interval used by the service is below — 59 features across 10 groups. It is published so that a reading can be checked against it, and so that anyone who thinks an interval is wrong can say which one and why. Corrections are welcome at support@darcyclothinguk.shop.

Lapel and collar

Observed	Points to
Notched lapel under 3 in (7.5 cm) Narrow lapels sit with the high buttoning points of the period before ready-to-wear widened the chest.	1900–1932
Notched lapel 3 to 3.5 in (7.5–9 cm) The long middle band. On its own this contributes very little narrowing.	1925–1975
Notched lapel over 4 in (10 cm) Wide lapels belong with the broad chest and square shoulder of the same years.	1935–1952
Notched lapel over 4.5 in with a low gorge The second wide-lapel period is told from the first by where the notch sits, not by width alone.	1968–1979
Broad peaked lapel on a double-breasted front The peak on a double-breasted front is one of the more tightly dated shapes in tailoring.	1928–1950
Three-button front pressed so the roll passes the top button Where the lapel is pressed to fold is a period habit, and it survives pressing.	1920–1942
High gorge, the notch sitting near the collarbone Gorge height moves slowly and steadily, which makes it a good anchor.	1900–1930
Low gorge, the notch sitting well down the chest A low gorge is hard to alter later, so it rarely lies.	1966–1980
Neckband cut for a detachable collar, with stud holes Stud holes are structural. They cannot be added convincingly afterwards.	1880–1935
Soft attached collar, no neckband studs A wide band, useful mainly for ruling the earliest decades out.	1930–2000

Shoulder

Observed	Points to
Narrow natural shoulder, no padding Padding thickness is one of the fastest-moving variables in the century.	1900–1925
Roped sleevehead standing above the shoulder line The rope is built into the sleevehead and is very seldom altered.	1930–1942
Square, heavily padded shoulder extending past the joint The widest shoulders of the century sit either side of the war years.	1938–1952
Narrow, lightly padded shoulder in a straight line The reaction against the padded shoulder is sharp and well dated.	1953–1965

Body and waist

Observed	Points to
High, strongly suppressed waist with a flared skirt The drape cut suppresses the waist hard and flares below it.	1930–1942
Straight, boxy body with little waist suppression Cloth rationing and then mass tailoring both favoured the straight body.	1946–1958
Hem falling below the seat, long skirt Jacket length shortens across the century almost monotonically.	1900–1935
Short hem, ending at the top of the seat Shortening cannot be undone, so a short hem may also be a later alteration.	1958–1972

Vents

Observed	Points to
No vent at the back hem The ventless back is the older default outside riding coats.	1900–1955
Single centre vent A wide band; contributes little on its own.	1930–1990
Double side vents Side vents move from riding wear into town tailoring at a datable point.	1955–1990

Pockets

Observed	Points to
Double-welt besom hip pockets, no flaps Flapless hip pockets are the formal older treatment.	1900–1935
Flapped hip pockets Very wide band. Flaps can also be tucked in, so read the pocket mouth.	1920–1990
Patch hip pockets on a tailored jacket Patch pockets on town tailoring, rather than on sports coats, are narrowly dated.	1925–1945
Ticket pocket above the right hip pocket The ticket pocket falls out of ordinary use and later returns as a quotation.	1900–1940
Slanted hacking hip pockets Slant angle is set by the pocket bag and is not alterable.	1930–1975

Buttons

Observed	Points to
Four-button front buttoning to a high point The high four-button front dies out early and does not come back.	1900–1922
Three-button front Very wide band on its own; read together with the roll line.	1900–1960
Two-button front Wide band. Useful mainly for ruling out the earliest decades.	1935–1990
Four-hole horn buttons with natural mottling Horn spans the century; it dates nothing on its own but rules out plastic.	1900–1990
Casein buttons, slightly translucent with a swirled grain Casein has a short commercial life, which makes it unusually informative.	1930–1955
Injection-moulded plastic buttons with a moulded shank and seam The mould seam is the tell. Buttons are also the most commonly replaced part.	1955–2000
Four cuff buttons set close together Wide band. Spacing matters more than count.	1900–1990
Two cuff buttons set wide apart Wide spacing on few buttons is an early habit.	1900–1935

Trousers

Observed	Points to
Waistband sitting at the natural waist, well above the hip Rise is set by the pattern and cannot be raised later.	1900–1955
Waistband sitting on the hip The drop in rise is one of the sharpest single changes in the century.	1960–2000
Two forward-facing pleats each side of the fly Pleat direction is a workroom convention that moves with the decade.	1922–1955
Reverse, outward-facing pleats Reverse pleats overlap the forward habit for two decades.	1935–1990
Flat front, no pleats, on a high rise A flat front on a high rise is early; a flat front on a low rise is late.	1900–1920
Permanent turn-ups Turn-ups can be let down, so their absence proves less than their presence.	1905–1960
Brace buttons on the outside of the waistband Outside brace buttons precede the move inside, which precedes their loss.	1900–1940
Side adjuster tabs with buckles Adjusters replace braces before belt loops become universal.	1930–1970
Belt loops on the waistband Loops are sometimes added later, so treat as weak evidence.	1925–2000

Closures

Observed	Points to
Button fly A button fly on a late garment is a choice; on an early one it is the default.	1900–1950
Metal zip fly The metal zip fly enters menswear at a well-recorded point.	1937–1975
Nylon coil zip fly The coil zip is the single most decisive late marker on a garment.	1965–2000

Internal construction

Observed	Points to
Hand-felled lining seams with irregular stitch length Irregular stitch length is the signature of a hand, not a machine.	1900–1950
Overlocked, serged raw edges Industrial overlocking arrives with volume ready-to-wear.	1950–2000
Pinked raw edges with no overlocking Pinking shears leave a zigzag that machines later replace.	1900–1955
Hand-worked buttonholes with silk gimp Hand buttonholes persist in bespoke work, so this dates the workroom, not only the year.	1900–1955
Machine-worked keyhole buttonholes The keyhole machine leaves an even, dense bar tack.	1930–2000
Full floating canvas chest piece A floating canvas moves independently of the cloth when pinched.	1900–1965
Fused, glued front interlining Fusing is the clearest post-war industrial marker inside a jacket.	1960–2000
Silk lining Silk gives way to viscose as the ordinary lining.	1900–1940
Viscose lining Wide band, but it rules out the earliest decades.	1925–1990
Half lining, sleeves lined only Half linings appear first in summer weights.	1930–1990

Cloth

Observed	Points to
Heavy woollen flannel, roughly 16 oz or above Cloth weight falls steadily once central heating spreads.	1900–1955
Light worsted, under 10 oz Light worsteds are a post-war possibility, not an earlier one.	1955–2000
Wool and polyester blend with a slight sheen A synthetic blend puts a hard floor under the whole reading.	1958–2000

Your photographs are not training data

Photographs you upload are not used to train or fine-tune any model, ours or anyone else’s. The recognition supplier processes an image on our instructions for the length of one request and is contractually prohibited from retaining it or using it for its own purposes. Retention periods and the deletion path are in the privacy notice.

No automated decisions about you

Nothing here makes a decision that produces legal effects for you or similarly significantly affects you. A reading is a description of a garment, not an assessment of a person. If you think a reading is wrong you can remove the feature you disagree with and the interval recomputes in front of you, and you can write to us at support@darcyclothinguk.shop and a person will look at it.