Plan a pilot box
Menu

In-house or outsourced digitization: how to decide

Reader question: Should we digitize with our own staff and equipment, hire a vendor, or split the work between them?

By Dastok Editorial TeamIssue: Published

Digitize in-house when the originals are fragile or rare, the volume is modest and steady, and you can commit a trained person and a dedicated space for years. Hire a vendor when the volume is large, the deadline is fixed, or a format needs equipment you would use once. Most institutions end up splitting the work: they keep selection, rights, quality control and the review of descriptions, and send out high-volume capture or text recognition at scale.

A disclosure first. Dastok sells digitization, so we have an interest in this question. To keep it honest, this article gives the case for doing the work yourself at least as much room as the case for hiring help, uses published sources for every claim about cost and risk, and lists the costs of outsourcing as plainly as the costs of building a lab. Estimating the hours themselves is the job of the digitization budget calculator; this article is about which route those hours should take.

When each route wins: the short version

In-house work wins on control, handling and learning; outsourcing wins on volume, speed, specialized equipment and predictable unit prices. The table sorts the common factors by which route they favor.

Factor Favors in-house Favors outsourcing
Condition of originals Fragile, rare or unique items that should not travel Sturdy material, or surrogates such as microfilm
Volume Hundreds to a few thousand images a year, steadily Tens of thousands of images on a deadline
Uniformity Mixed material that needs judgment item by item Long runs of similar items that suit one setup
Formats Letters, photographs and small documents on common equipment Oversize maps, film, glass or audio needing gear you would use once
Time horizon An ongoing program over many years A one-time project or a grant with an end date
Staff Someone who can be trained and will stay No one with the time, or turnover you cannot absorb
Budget shape Capital money now, little cash for services later Project money that must be spent on deliverables
Learning You want the skills inside the institution You want the result, not the skills

No single row decides the question. If most of your answers fall in one column, that is your starting route; if they split, read the section on hybrid projects first.

Where to start: volume against condition In-house, with care Trained handling, cradles and time per item; or a provider working on site Hybrid On-site capture or planned transport; you keep selection and QC In-house, routine Common equipment, steady weekly batches, sampled QC Outsource capture Per-item pricing, several capture streams; you check a sample of every batch Small, steady volume Large volume, fixed deadline Fragile or rare Sturdy
A rule of thumb for a starting route, not a formula; formats that need special equipment can move a project to the right. Based on NEDCC Preservation Leaflet 6.7 and OCLC Research, Rapid Capture (2011), checked November 2026.

It is rarely all one or the other

Very few projects are fully in-house or fully outsourced. Some tasks stay with you whoever holds the camera, because they depend on knowing the collection.

The Northeast Document Conservation Center’s leaflet Outsourcing and Vendor Relations, written by Robin L. Dale, says of the either-or choice: “In reality, such a dichotomy does not exist.” It adds that “there will always be in-house components of any digitization project,” and names them: “selection, determining user needs, benchmarking (establishing required quality specifications for digitization), and quality control and verification are best done by staff members who know the collections.”

The evidence it cites is old but consistent. Researchers from Glasgow University’s Humanities Advanced Technology and Information Institute interviewed people from 36 major cultural heritage digitization projects in 2000 and 2001. More than half “both digitized in-house and used commercial vendors.” Daniel J. Cohen and Roy Rosenzweig, reporting the same study in their book Digital History, add that most of the rest did all the work on their own premises, and “only two relied exclusively on outside companies.”

Experience changes the answer

One observation in the NEDCC leaflet runs against intuition: “people and institutions experienced with digitization projects are more likely to outsource components of such projects than are those who lack experience.” First projects, it notes, are often seen as a chance to learn skills and buy equipment, so “the initial urge is to accomplish everything in-house.”

Neither impulse is wrong. Experience shows where an institution’s own hands add value and where they only add hours, and that can only be learned by doing some of the work. The right question is therefore which tasks to keep and which to send out, which is the subject of a section below.

The case for digitizing in-house

Doing the work yourself gives you control over handling, quality and pace, keeps originals in the building, and builds skills that stay after the project. For steady, modest volumes of common material, it can also cost less than the fee per image suggests.

Control over handling and quality

The NEDCC leaflet quotes Janet Gertz of Columbia University Libraries: “The primary argument for digitizing in-house is that it gives the institution close control over all procedures, handling of materials, and quality of products. There is no need to send valuable or fragile originals off-site and no worry about working with a vendor who turns out to be incompetent, provides something other than what was required, or goes out of business.”

Handling is the strongest of those reasons. Cohen and Rosenzweig report that “the most common explanation” project directors gave for doing work onsite “was that the original materials were rare or fragile,” citing one institute’s concern about “the cost and risk of transporting the original materials.” Every trip out of the building is a trip that can go wrong, and the collection care page lists what has to be agreed before any original moves.

Requirements you can change as you learn

A vendor contract fixes the specification at the start. The NEDCC leaflet notes that “changes to requirements along the way may require contract renegotiations and possibly price increases,” while in-house teams “can adjust requirements incrementally, making changes to specifications based on gradual experience with the original materials or early project outputs.” For a collection nobody has looked at closely, that flexibility matters: the third box often reveals what the first two did not.

Skills that stay with the institution

Gertz, as quoted by Cohen and Rosenzweig, notes that “working in-house is a good way to learn the technical side of digitization,” which may prove useful “even when most work in the future will be sent to vendors.” Staff who have captured a few hundred cabinet cards write better specifications, spot problems in a vendor’s batch faster, and answer researchers’ questions about the files with authority.

Routine material does not need a large lab

The OCLC Research report Rapid Capture (April 2011) described several in-house programs that reached real volume with modest means:

  • A single sheet-fed scanner. The University of Minnesota Archives scanned 219,074 pages of legacy university publications in a year “on a single scanner, a $1,300 Fujitsu fi-6230,” run by student workers averaging 500 pages per hour, scanning and OCR. (Those were duplicates and publications it then recycled, so the speed came from material that needed no special care.)
  • Photocopiers and students. Princeton’s Mudd Manuscript Library, in a pilot, scanned “over 20 linear feet” of 20th-century manuscripts, about 70,000 images, using student workers and two photocopiers at about 300 pages per hour. Its assistant university archivist said, “Where we really save is on training, since we use undergrads and equipment that they are already familiar with.”
  • An in-house production center. The University of North Carolina’s Southern Historical Collection did its digitization in-house in the library’s Digital Production Center, with about 70 hours a week of graduate student help and typical capture of 90 images per hour.

These figures are from 2011 and from university settings, so treat them as evidence that routine capture is achievable, not as rates to plan with. Your own rates come from timing your own material, as the article on running a digitization pilot explains.

Text recognition you can run yourself

The published comparisons of in-house and outsourced work predate today’s handwriting recognition, and that changes the in-house case. Printed and typed text can be read with Tesseract, an open-source OCR engine released under the Apache License 2.0. Handwriting recognition is available on platforms an institution can subscribe to directly; Transkribus, for example, sells credit-based plans and, as of November 2026, offers a free plan with 50 credits a month.

Running the engine is the easy part. The work is measuring the error rate on your own pages, correcting what matters for search, and labeling machine output until a person has checked it, as our articles on AI handwriting recognition accuracy and labeling AI-generated metadata describe. A small team that can do those three things can keep text recognition in-house.

Volunteers can carry real work

Many historical societies run on volunteers, and an in-house program can put them to good use. Preparation, capture of sturdy material on a fixed setup, checking images against a short list, and correcting machine text are all tasks a trained volunteer can do well. What volunteers need is a written procedure, a fixed station so settings do not drift, and a staff member or senior volunteer who checks a sample of their work every session.

Plan for the limits as well. Volunteer hours are irregular, so set targets per month rather than per day, and keep fragile items for the most experienced hands. The article on volunteers, AI or both compares volunteer transcription with machine recognition in more detail.

A basic capture station

Atla’s guide to digitizing collections lists what a starting imaging workstation usually needs: a digital camera with a 50 mm lens, a copy stand, lights and light stands, a book cradle with weights and other tools to hold items in place, a computer with image-processing software, hard drive storage with a cloud backup, a flatbed scanner and a film or slide scanner. The space, it adds, “needs to be secure and safe for your equipment & materials,” and “lighting needs to be consistent, typically low or no room light.”

The NEDCC leaflet adds the facility questions that are easy to skip: whether the room meets “minimal electrical requirements,” whether it has “environmental controls appropriate to the materials and staff,” and whether its layout “facilitates a productive work flow.” A corner of the reading room with a window behind the copy stand will fail the lighting test every sunny afternoon.

When in-house is the clear choice

  • The originals are too fragile, rare or sensitive to leave the building, and no provider can work on site.
  • The volume is a few boxes a month, indefinitely, with no deadline.
  • Requests from researchers drive what gets scanned, one item at a time.
  • You already have a trained person, a quiet room and basic equipment.
  • The point of the project is to build skills, for example in a teaching collection or a volunteer program.

What in-house digitization really costs

The in-house bill is mostly people’s time, plus equipment that ages, a room that is taken out of other use, and the cost of learning. None of it arrives as a single invoice, which is why it is easy to underestimate.

The NEDCC leaflet’s advice is blunt: “The adage for determining in-house digitization costs is to estimate costs and then double them,” and to double the result again for a first project. It also warns that many of the costs “are not one-time costs; most of them are ongoing to some degree,” including “annual equipment costs, maintenance licenses, system-monitoring costs, replacement costs associated with technological obsolescence.”

The cost lines to count

Cost line What to count Easy to forget
Equipment Camera or scanner, lens, copy stand, lights, cradle, calibrated monitor, computer, software, color targets Replacement when it ages, and repairs
Space A room with stable light, power, security and climate suited to the originals The other use that room no longer has
Capture labor Hours of capture at a loaded rate (wages plus taxes and benefits) Recapture of failed images
Preparation Removing fasteners, flattening, sorting, condition checks Returning items to order afterward
Quality control Checking images against the specification and targets Time to fix and recheck
Post-processing Cropping, rotation, derivatives, file naming Storage while files wait
Description Catalog records, review of machine suggestions Authority work for names and places
Training and turnover Learning time, supervision, retraining a replacement Slow output during the learning curve
Storage and IT Preservation copies, backups, checks The years after the project ends
Management Planning, reporting, grant administration Meetings

The Digital Library Federation’s Digitization Cost Calculator breaks a project into similar steps: condition review, fastener removal, sorting, rights review, capture, quality control at several levels, post-processing (alignment, clean-up, color correction, cropping), descriptive metadata at several levels, and putting the material back. Its own banner carries a warning worth repeating: the web version “will not be maintained as of 1/1/2025” and “there may be inaccuracies in the data and results,” since it is built on contributed community data. Its maintainers published a spreadsheet version on OSF.

Start-up time is a cost too

Equipment does not produce images the day it arrives. In the OCLC report, UCLA described “a lot of startup delays,” including installation, training and “several months” of learning new workflow software. The NEDCC leaflet adds that an in-house team “will not be ready to start your project at the moment you receive the funding,” and that between the first and last day “the organization faces variables that can affect quality, cost, and the ability to meet deadlines,” including staff who leave once they are trained.

Capture is not the bottleneck

In-house teams in the OCLC report kept finding that the camera was the fast part. At the Archives of American Art, “processing takes about the same amount of time as capture.” In a University of Minnesota project, “quality control continues to provide the largest bottleneck.” At the Walters Art Museum, “copying back-up files and storing the images takes a surprising amount of time and attention.” Budget the whole chain, not the shutter.

The case for outsourcing

A good vendor brings capacity, equipment and routine that a small institution cannot build for one project, and it prices the work per item so the budget is known in advance. It also carries the cost of equipment that ages and staff who leave.

Capacity and speed

Vendors can run several capture streams at once. In the OCLC report, the Archives of American Art, which had captured from originals only in-house until 2010, noted that its vendor “can have more than one capture stream working on different sections of the collection,” so it could “complete far more than can be scanned in-house.” The NEDCC leaflet says the “volume and throughput of a specially designed and staffed production facility is far greater than any cultural heritage organization is likely to have.” For a grant with an end date, that capacity can decide whether the project finishes.

Equipment you would use once

Some formats need gear a small institution would buy for one project and then store: wide-bed scanners or large copy stands for maps and drawings, transmissive setups for film and glass, microfilm scanners. Cohen and Rosenzweig note that vendors often have an advantage with nonstandard materials such as large-format maps “because they own expensive specialized equipment.” Our articles on scanning large maps and drawings and digitizing glass plate negatives show how specialized those setups are.

Known prices and shifted risk

With a vendor, the NEDCC leaflet notes, “the cost of digitization is almost always related in per-item charges that allow for easy comparison and understanding,” and “obsolescence costs are borne by the vendor,” as are “problems related to staff expertise and staff retention.” In the OCLC report, the Library of Congress observed that “vendors can quickly upgrade or change technology as requirements change from project to project.” For a grant budget, a per-item price from a signed quote is easier to defend than an estimate of your own hours.

Surrogates travel without the risk

Some material can leave the building with little at stake. The Bancroft Library’s note in the OCLC report puts it simply: microfilm reels “are not original materials, so can be shipped off-site without the high cost (insurance, specialized shipping) of sending original materials.” The same reasoning applies to sturdy modern records held in duplicate.

Help beyond capture

Vendors also offer services around the camera. The NEDCC leaflet lists preparation and flattening, OCR, encoding, derivative files, extra levels of quality control, and storage, noting that “simply receiving copies of files on CDs, DVDs, magnetic tape, or disk drives is not enough.” Today the same list includes handwriting recognition, machine-suggested catalog fields and preservation packaging, which is where many small institutions lack both tools and time.

What outsourcing really costs

The vendor’s invoice is only part of the outsourced bill. Your staff still select and prepare material, write the specification, choose the vendor, check the work, and receive and re-shelve the originals.

Cost line What to count Easy to forget
Vendor fees Price per image, page, item or hour, by format Set-up or minimum charges; rework terms
Selection and preparation Choosing, counting, listing and preparing items to ship Rights review and restrictions
Specification and procurement Writing the specification, the request for proposals and the contract Answering bidders’ questions
Transport and insurance Packing, courier or vendor pickup, coverage in transit and on the vendor’s premises Condition checks before and after
Oversight Approving first batches, answering questions, tracking schedule Staff time over the whole contract
Acceptance checks Your own quality control on a sample of every batch Re-inspection when a batch fails
Receiving files Validating, copying and storing deliveries Storage for the years after
Description Review of any machine-suggested records Authority work, as in-house

Procurement and oversight take real hours

Cohen and Rosenzweig list “the staff time to solicit bids, select a vendor, and monitor a contract” among outsourcing’s costs. They cite an older study of the University of Michigan’s Making of America project, where preparing the request for proposals “consumed several days of the project manager’s time,” for a contract covering only page images. The same request drew fourteen bids, “with prices ranging from ten cents to four dollars per page,” which shows both why careful procurement pays and how far apart quotes can be. Our guide to writing a digitization RFP covers how to get bids you can compare.

Your own quality control stays

Handing over capture does not hand over responsibility for quality. The NEDCC leaflet says “local quality control, for at least a sample of returned files, is necessary to validate the vendor’s work,” and Cohen and Rosenzweig advise checking “the quality of the work produced by the vendor, preferably before a major portion of the digital conversion has been done.” Budget a sample from every batch, and the hours to re-inspect when one fails.

What to check before you sign

A short list of checks avoids most of the problems that make outsourcing expensive later. Atla’s guide suggests asking what security a vendor provides for your objects, considering a site visit, asking about its experience with cultural heritage collections and its equipment, confirming it can meet your file standards, being clear about “who holds ownership of the digital files,” establishing a timeline, asking for a test and having a signed contract. The NEDCC leaflet adds a warning about businesses that sell generic “scanning” services: they “may try to convince you that their stated technical requirements for digitization are ‘more than necessary’” and propose lower quality.

Ask, too, where the work will be done, who will handle the originals, how the provider reports problems found mid-batch, and how long it keeps working copies after delivery. Put the answers in the contract, not in an email.

Small jobs cost more per item

Fixed set-up work spreads thinly over a small order. Cohen and Rosenzweig note that because vendors’ set-up costs can run high, “you will wind up paying much more per page for a small job than a large one.” If your project is small, ask how the price changes with volume, and consider joining another institution’s order.

Risks on each side, and how to reduce them

Both routes carry real risks to originals, files and schedules. The table sets them side by side with a way to reduce each one.

Risk In-house Outsourced How to reduce it
Handling damage Inexperienced staff and volunteers handling fragile items Vendor staff unfamiliar with heritage material Training, written handling rules, a condition baseline before work starts
Transport None Loss or damage in transit Vendor pickup or on-site work, packing rules, insurance in writing
Quality drift Settings slip as staff change or tire Vendor meets the letter of the specification, not its intent Targets in every session, sampled acceptance checks, a pilot first
People One trained person leaves The vendor’s experienced staff are not on your job Written procedures; named staff and supervision in the contract
Continuity Equipment fails with no budget to replace it A small vendor closes mid-project Replacement fund; staged payments and delivery batch by batch
Where the work happens Not applicable Material or files sent offshore or to a subcontractor A contract clause naming the location and any subcontractors
Ownership and exit Files stuck on one staff laptop Files held on a vendor platform Masters delivered to you, with ownership and licensing set in the agreement
Schedule Other duties crowd out scanning Long queues at popular vendors A realistic plan; ask for the queue time in the quote

Two of those rows come straight from the NEDCC leaflet. It warns that “some vendors offer cheaper services by sending the work offshore” and tells readers to know where their materials are going. It also notes that “choosing to work with a smaller or lesser-known vendor could place your contracting organization at greater risk if the vendor becomes financially unstable,” while the most recommended vendors “have longer waiting times and may not meet your schedule.”

Transport, the strongest objection to outsourcing, “need not be a problem,” the leaflet adds: “Some vendors drive to an organization for pickup, and still others coordinate special shipping.” Some providers will also work at your premises; Cohen and Rosenzweig mention vendors who “set up shop at your location.” Dastok can work on site when originals cannot travel, and the collection care page lists the questions to settle in writing with any provider.

Which tasks to keep and which to send out

Keep the tasks that depend on knowing the collection and its users; consider sending out the ones that depend on volume, special equipment or tools you do not have. The split below is a starting point, task by task.

Task Usually keep in-house Often sent out Notes
Selection and priorities Yes Rarely Depends on knowing the collection and its users
Rights review and restrictions Yes Advice only Your institution carries the legal decision
Specification of quality Yes, with advice Advice You approve it; a vendor or consultant can draft it
Preparation (fasteners, sorting) Often Sometimes Cheap to do in-house; vendors charge for it
Capture of fragile or rare items Often With on-site work or careful transport Decide item by item
Capture of sturdy, uniform runs Sometimes Often Volume favors several capture streams
Oversize, film, glass, audio Rarely Often Equipment you would use once
Quality control Acceptance sample always Full inspection by the provider You check a sample of every batch
OCR of print and typescript Feasible Often bundled with capture Measure error rates either way
Handwriting recognition at scale Feasible on a platform Often Needs measured accuracy and correction
Machine-suggested catalog fields Review always Drafting often People review; machine fields stay labeled
Description and authority work Yes Sometimes drafting Names, places and context need local knowledge
Publishing and import Depends on your system Often Test import before full delivery
Preservation packaging Storage always Packaging often You keep the copies and run the checks

The pattern follows the NEDCC list of tasks “best done by staff members who know the collections,” with one change the older guidance could not foresee. Machine text recognition and machine-suggested description can now be bought as services or run on platforms you subscribe to, but in both cases a person who knows the collection still has to review the output before it is trusted. That review is an in-house task in almost every hybrid plan.

Three common hybrid patterns

  1. Vendor capture, in-house description. The vendor captures and runs text recognition; your staff review records and correct the names. This suits a large run of letters or ledgers with a small cataloging team.
  2. In-house capture, outsourced specialty work. Your team captures loose papers and photographs on a copy stand; maps, film and glass go out. This suits a steady program with a few awkward formats.
  3. On-site provider, in-house selection and checks. The provider brings equipment to your building for fragile material; your staff select, prepare and accept each batch. This suits collections that cannot travel and are too large for one person.

A worked comparison for a small historical society

The comparison that matters is the cost of the same output by each route, including your own staff time on both sides. Count only what differs between the routes; preparation, description and storage often cost about the same either way.

Example (fictional). The Aster County Historical Society plans to digitize about 30,000 images of loose family papers. Its pilot timed letters at 48 images per hour, including handling, with 2.2 percent recapture. It values staff time at a loaded rate of $38 per hour, the same rate it uses in the budget calculator.

In-house capture. At 48 images per hour, 30,000 images take 625 hours, plus 14 hours of recapture. Quality control at 15 percent of capture time adds about 94 hours. That is about 733 hours, or about $27,850 at $38 per hour, before equipment. The society’s own equipment quotes, a fictional figure for this example, come to $9,000, so in-house capture costs about $36,850, or about $1.23 per image. The equipment remains for later boxes, but so does its upkeep.

Outsourced capture. The society would still prepare the boxes, write the specification, check a sample from every batch and receive the files. It estimates 120 hours for procurement, oversight and acceptance checks, or $4,560, plus $1,500 for packing and insured transport. Done in-house, the 733 hours of capture and checking are about 18 weeks of one person’s full-time work.

The break-even. Outsourcing capture costs less if the vendor’s all-in price is below about $31,000, or about $1.03 per image. The society will request quotes in that unit and compare each against that line, not against the bare per-image fee.

Example (fictional), continued. The society’s Wren Gulch Mercantile day books and a folder of oversize plat maps do not fit its copy stand. It decides on a hybrid: volunteers capture the loose letters and cabinet cards in-house, the maps and day books go to a provider that will capture them on site, and the society’s collections manager reviews every machine-suggested catalog field before records are published. The board minute records two reasons beyond cost: the volunteers want to learn the work, and the day books should not leave the building.

The arithmetic is simple; the judgment is not. A quote slightly above the break-even may still win if it frees a staff member for description, the slower stage in most projects. A quote below it may still lose if the originals are too fragile to travel. Write down which factors outweighed the numbers, so the board can see the reasoning.

Run the numbers on your own material

Measured times beat published ones. Rates from other institutions “travel badly,” as the pilot article puts it, because material, resolution and handling differ. The OCLC report warns that “comparing throughput rates across the vignettes must take other factors into account,” noting that “for some of these projects 300 ppi was adequate resolution; others required as much as 2500 ppi.” Time a representative sample, enter the times in the digitization budget calculator, and compare the result with real quotes.

How to decide: questions and next steps

Decide with measured times, real quotes and a written list of what matters beyond cost. A short pilot, run in-house or with a vendor, gives you the first two; your board supplies the third.

Questions to answer before choosing

  1. Can these originals travel? If not, the choice is in-house or a provider working on site.
  2. What volume, by when? Count images by format and set the date the work must finish.
  3. Who will do it, for how long? Name the person, and say what happens if they leave.
  4. What equipment does each format need? Separate the common formats from the one-off ones.
  5. What will stay in-house either way? Selection, rights, acceptance checks and description review usually do.
  6. What is the break-even? Compare all-in in-house cost with all-in outsourced cost, including oversight.
  7. What does the funder require? Some grants favor building capacity, others favor deliverables.
  8. Who owns the files, and how do you leave? Settle ownership, formats and exit in the agreement.

Next steps

Dastok offers that test as a pilot box, and our engagements page and services describe what we do and what you keep. The same questions apply to any provider, and to the copy stand in your own back room.

Working through this with a real collection?

If this article describes a decision you are facing, tell us about the collection. A pilot box answers most questions with your own material.

Plan a pilot box

Studio and contacts

24 South Weber Street, Suite 225
Colorado Springs, CO 80903
(719) 244-8450sales@dastok.com