Home / Guides / Data Merge Preflight

Production guide · variable data

Data Merge Production Preflight

A practical release sequence for source records, linked images, difficult names and cut-and-stack output.

Purpose and production context

Data merge is a production handoff between a structured source, a layout template and a finished physical run. The most expensive failures are not always syntax errors. A valid name can be paired with the wrong portrait, a correctly linked image can be obsolete, and a numerical sequence can become unusable after guillotine cutting. This guide establishes a controlled path from source approval to release without pretending that a browser checker replaces the compositor, RIP or operator. Assign a data owner and an artwork owner before production. Keep one immutable received source and identify every approved revision; do not let several corrected spreadsheets circulate under the same filename.

Definitions and distinctions

A record is one logical data item, even if a quoted field contains line breaks. A physical CSV line is not necessarily a record. A field is a named source column; an image field refers to a file rather than containing its pixels. A source ID must remain stable through sorting, diagnostics and output. A physical sheet is the complete press layout, while a cell is one finished-piece position on that sheet. Cut-and-stack sequence describes how records are distributed through a lift so that collecting the cut piles produces the intended order. It is different from calculating how many pieces fit. Text-fit screening estimates plain-text geometry; final overset checking examines the actual document's composition and fonts.

Inputs, units and method

Receive a header-bearing CSV or tab-delimited export and the corresponding image package. Identify delimiter and encoding from the export procedure, not from the computer's language. Import identifiers as strings: 001 and 1 may represent different printed values, and a sixteen-digit number can lose precision in a spreadsheet. Write required fields and uniqueness rules into the job ticket. For images, define the root directory and exact relative paths. For text, measure outside frame width, height and inset in mm; use the specified regular font, size and fixed leading in pt. For sequencing, specify across/down cell counts, maximum sheets per cutting lift, pile collection order and the within-pile sheet reading direction. These last quantities are counts and routes, not dimensions.

Source and asset approval sequence

Run CSV Preflight before opening the source inside a template. Resolve malformed quotes and inconsistent field counts before assessing business warnings. Send a record-level report to the data owner instead of silently deleting unusual punctuation or converting IDs. Then run the Image Asset Matcher for every variable-image field. An exact filename match proves only that the selected path exists in the manifest; open representative images in the compositor and compare them with the approved ID/name pair. Investigate case-only matches, duplicate basenames in different directories and blank image fields. Do not delete unused assets automatically, because another field or approved version may still need them.

Worked production example

An event job has eight numbered passes and an approved two-cell sheet layout. The cutter plans two sheets per lift and collects the left pile before the right, reading each pile in emitted-first sheet order. The CSV Sequencer emits source record pairs 1/3 and 2/4 for lift one, then 5/7 and 6/8 for lift two. Its production CSV contains ID__C1, Name__C1, ID__C2 and Name__C2 rather than a single ID column. The designer maps each cell's field set in a full-sheet template and performs a single-record merge. For a seven-pass run, one final cell is blank; cutting and removing that blank must still yield records 1 through 7 exactly once. Reversing the actual delivered stack without changing the selected sheet-read direction invalidates this example, so the operator signs off a paper dummy.

Text proof and output interpretation

Screen the Name field with the Variable Text Fit Checker using the intended frame and optional local font. A 40 by 12 mm frame with 1 mm inset provides about 107.72 by 28.35 pt of usable area. Two 14 pt line intervals occupy 28 pt; a third does not fit the declared model. Review the worst measured strings, empty cells, accented names, explicit newlines and each writing system. A FIT ESTIMATE is not a production approval: a browser can substitute glyphs, and InDesign can use different line-breaking, baseline, table or paragraph rules. Never automatically shrink all names to hide one problematic record. Agree a readable exception style or correct the frame with the artwork owner, then repeat the actual-layout proof.

Decision checklist

Release only after every source record can be reconciled with the approved job quantity. Confirm that required values and record IDs meet the written contract; that selected images correspond to the right people/products; and that final PDF fonts, bleed, trim and picture resolution pass the normal preflight. Prove first and last sheets, each lift boundary, the partial last lift and at least one known difficult text record. Retain the source revision, diagnostic reports, expanded CSV, cell map and signed dummy instructions. Confirm the template uses single-record sheet rows and that the RIP will not apply a second cut-stack arrangement. Specify how blank cells, rejected pieces and reprints are separated so a restart cannot create duplicate credentials or mismatched records.

Common mistakes and cautions

CSV quoting is not protection against spreadsheet formula execution. An untrusted cell beginning with a formula initiator can remain dangerous even when enclosed in quotes. Diagnostic reports can prefix such text, but doing that silently to a production source would change what prints, including an @ image header. Keep production exports lossless, inspect trusted data and use a text-preserving import workflow. Do not upload private recipient lists into unrelated online tools merely to test them. Browser-local processing avoids a server transfer but does not replace workstation access controls or safe download handling. Another frequent mistake is treating a parser warning as proof of an application defect: encoding/import behavior varies by version, and a valid CSV cannot repair a compositor regression.

Related workflow and source documents

Return to the Data Merge Production Prep hub to repeat checks on the approved revision, not on an obsolete working copy. Use Print Imposition for sheet geometry and Bleed and Safe Area when setting the artwork frame. Technical references include Adobe's data-merge overview, RFC 4180 CSV conventions and OWASP's CSV injection guidance. These tools prepare data and expose assumptions; final production evidence comes from the approved merged file and physical collection proof. Last reviewed: 2026-10-26.