BRENNEN

Industries / Finance and business services / Printing

Offset printing

Offset printing inks a flat plate that is kept damp, so that only the image areas take the greasy ink, passes the image to a rubber blanket and from there to paper, on presses fed with single sheets (sheetfed) or from a reel (web).

4

Milestones achieved, from 1796 to 2011.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1796

    Printing from flat stone

    Alois Senefelder, a Bavarian actor and playwright looking for a cheap way to publish his plays, found that a drawing in greasy ink on Solnhofen limestone could be printed from the flat surface of the stone, which he called chemical printing and the French called lithography. By his own account, he and the composer Franz Gleissner printed 120 copies of 12 songs in under two weeks in 1796, spent no more than 30 florins and sold 100 florins' worth; he received British patent No. 2518 for the process in 1801.

    LessonA new process needs printers trained in it; in 1797 a Munich music publisher's untrained printers "spoiled a great deal of paper", so he went back to copper plates, by Senefelder's account.

    historyofinformation.com

    Credit

    Alois Senefelder

  2. 1903

    Printing through a rubber blanket

    Ira Washington Rubel, who ran a small paper mill and lithographic shop in Nutley, New Jersey, built a rotary press that printed from the plate onto a rubber-covered cylinder and from there onto paper, the method now called offset. The Smithsonian's National Museum of American History, which holds the press, says it ran in Rubel's New York plant in 1904, was sold in 1905 to the Union Lithographic Company of San Francisco for $5,500 and printed about 2,500 sheets an hour. Charles and Alfred Harris sold their first production offset press, at about 3,000 sheets an hour, in July 1906, and Harris presses soon became the most popular offset presses in the United States, by the museum's account.

    LessonRubel's syndicate, set up early in 1906 as a monopoly to sell the press, failed later that year, while the Harris brothers' production machines took the market.

    americanhistory.si.eduamericanhistory.si.edu

    Credit

    Ira W. Rubel; Charles and Alfred Harris

  3. 2009

    State guarantees for Heidelberg

    Heidelberger Druckmaschinen (Heidelberg), a German press maker, reported sales down 18.3% to about €3 billion and a pre-tax loss of €347 million for the year to March 2009; its chief executive called it one of the most difficult years in the company's history, as printers struggled to get loans to buy presses. Its banks approved new credit on condition that the German federal government and the states of Baden-Württemberg and Brandenburg guaranteed it, along with a loan from KfW, the state development bank, by the end of August 2009, by the company's annual report. In October 2009 Heidelberg said it would cut about 4,000 jobs worldwide.

    LessonPrinters buy presses on credit, so when banks stopped lending, a press maker's orders fell at once.

    sprinter.com.auheidelberg.compiworld.com

    Credit

    Heidelberger Druckmaschinen AG; the German federal and state governments

  4. 2011

    A press maker goes insolvent

    manroland AG, which the IG Metall union called Germany's third-largest printing press maker, filed for insolvency at the district court in Augsburg on 25 November 2011, after an investor withdrew and talks on a loan failed. The company said orders had fallen sharply since mid-July 2011 and that its market was only half its size before the 2008 crisis. It employed about 6,500 people by its own count and 6,600 by IG Metall's. Its sheetfed business passed to Langley Holdings in 2012.

    LessonWhen demand for new presses halved, not every maker survived; manroland's business was split up and sold rather than rescued whole.

    printcan.comigmetall.delangleyholdings.com

    Credit

    manroland AG

Watch for

  1. Inks that waste fewer sheets

    hubergroup, an ink maker, says a new resin it makes in-house gives its conventional offset inks less misting (a fine spray of ink thrown off fast rollers), a wider water window (the range of dampening water over which the press prints cleanly), lower dampening levels and "significantly fewer" make-ready sheets, the sheets printed while setting up the press before saleable copies come off. It cites "extensive long-term production testing" and gives no figures, and it planned to bring the resin into all its conventional offset inks in the first half of 2026.

    Achieved whena printer or test lab outside hubergroup publishes before-and-after counts of make-ready sheets and misting on the same press and jobs.

    hubergroup.com

    Claimed by

    hubergroup, May 2026

  2. AI sets up the press

    Heidelberg says Brodnax 21C Printers in Dallas, which installed a Speedmaster XL 106 sheetfed offset press, can complete setup in less than 30 seconds with automatic plate changing and an AI-controlled system, Intellistart 3, that plans every step of each job change "without operator intervention", and that its sheets are right for color and register "after just 60 sheets". The release gives no figure for the press it replaced and no waste figures.

    Achieved whenthe printer or an outside party publishes measured job-change times and make-ready sheets over a run of real jobs, against a baseline.

    heidelberg.com

    Claimed by

    Heidelberg, July 2026

Open

  1. 01

    Public offset defect image set

    A public, labeled image set of the faults that spoil offset sheets: hickeys (spots with a white ring, from specks of dirt or paper on the plate or blanket), ghosting (a faint second image), streaks, scumming (ink in areas meant to stay blank), set-off (wet ink marking the next sheet), misregister (colors out of line) and mottle (blotchy solids). Label each fault with its cause where the printer knows it, and publish a baseline detector with its scores for each fault type.

    The brief
    Credit · unclaimed
  2. 02

    Open make-ready waste log

    A public, anonymized log of how many sheets offset jobs waste before and during the run, with the job details that drive waste (number of colors, sheet size, paper, run length, press model, ink coverage, whether the previous job used the same paper), gathered from several printers, plus a baseline model that predicts the waste before a job runs.

    The brief
    Credit · unclaimed

Open 01 · the brief

Public offset defect image set

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public, labeled image set of the faults that spoil offset sheets: hickeys (spots with a white ring, from specks of dirt or paper on the plate or blanket), ghosting (a faint second image), streaks, scumming (ink in areas meant to stay blank), set-off (wet ink marking the next sheet), misregister (colors out of line) and mottle (blotchy solids). Label each fault with its cause where the printer knows it, and publish a baseline detector with its scores for each fault type.

Why it is still open

Press makers and inspection firms sell camera systems that catch faults on press, and researchers train on private or synthetic samples: Valente and colleagues (2020) trained on two simulated defects and tested on their own printed images. The closest public sets do not cover offset sheets: a 2022 Mendeley set has 299 images of lines from a gravure offset system for printed electronics, and the "Printing Errors" set on Roboflow Universe (639 images, six classes such as streak and misregistration) does not say what process or press made them. Searches of Hugging Face, Zenodo, Crossref and GitHub (November 2026) found no documented set of sheetfed offset faults; on GitHub the nearest are a nameplate print checker and a list of print-defect papers.

arxiv.orgdata.mendeley.comuniverse.roboflow.comgithub.com

Built with what exists

Make-ready and spoiled sheets collected from print shops; a flatbed scanner at 600 dpi or a phone on a copy stand; labeling in CVAT or Label Studio; an open detector such as Ultralytics YOLO, or a segmentation model such as DeepLabv3+, trained in PyTorch.

  1. First step

    Ask one local commercial printer for a week of make-ready and spoiled sheets, scan 200 of them and label the hickeys.

  2. Done looks like

    At least 5,000 labeled faults across the seven types, from sheets printed at three or more printers, released under an open license, with the baseline's precision and recall for each fault type reported on sheets from a printer held out of training.

Open 02 · the brief

Open make-ready waste log

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public, anonymized log of how many sheets offset jobs waste before and during the run, with the job details that drive waste (number of colors, sheet size, paper, run length, press model, ink coverage, whether the previous job used the same paper), gathered from several printers, plus a baseline model that predicts the waste before a job runs.

Why it is still open

Kalafatelis and colleagues (2023) predicted paper waste per order with machine learning, but from four months of one Greek printer's records (Pressious Arvanitidis), which were not published; the formulas they cite, such as Hamerliński and Pyr'yev (2014), estimate waste from a few factors. Ink and press makers state make-ready savings without job-level numbers, as in hubergroup's and Heidelberg's 2026 releases above. Searches of Zenodo, Hugging Face, Crossref and GitHub found no public multi-printer waste data.

zenodo.orghubergroup.com

Built with what exists

Job records exported from print shops' management software, a simple data-sharing agreement, Python with pandas, scikit-learn or XGBoost for the baseline, and Zenodo to publish the data with a fixed DOI.

  1. First step

    Draft a one-page form with about 12 fields and ask three printers to fill it in for one week of jobs.

  2. Done looks like

    5,000 jobs from at least five printers and three makes of press, published with a data dictionary, and a baseline whose average error in waste sheets is reported on a printer held out of training.