Know a company with years of records? Refer it and earn up to $100K
Unused and archived records

Dark Data Monetization: Turning Unused Records Into Income

Dark data is what a company stores and never uses. Which archived records AI developers license, which have no market, and how to find and count them.

A signed licensing agreement with a fountain pen and archive boxes
Short answer

Dark data is information a company collects and stores but never uses. Some of it can earn income: AI developers license records of real work with the outcome attached, such as closed tickets, engineering files, inspection records and call recordings. Much of it cannot: duplicates, logs with no outcome, bought-in lists and anything you have no right to share. Monetizing it means a licence, so you keep the records. Start with a one-page inventory.

  • Value comes from records that show a problem, the work done and the result. Volume alone is not value.
  • Age is rarely the obstacle. Rights and personal data are.
  • Finding dark data is a counting exercise, system by system.
  • A licence is not a sale. The archive stays where it is.

What is dark data?

Data a company keeps and does not use. IBM defines it as "the information that organizations accumulate but often never use for analytics or decision-making" (IBM). It includes old project folders, closed tickets, archived email, call recordings, sensor logs and the databases of retired systems.

There is a lot of it. IBM cites a Splunk survey of more than 1,300 decision makers in which "60% of respondents reported that half or more of their organization's data is considered dark". IBM puts this partly down to "the advent of inexpensive storage": keeping everything became easier than deciding what to keep. For some record types the archive has since found a use nobody stored it for.

Is unused company data worth anything?

Some of it is. AI developers pay for records that show how skilled work was done and how it turned out, and those are often the records a company stopped looking at when the job closed. Most of the rest has no market.

What decides value is not size. It is whether each record links a situation, the work and the result, whether there are years of them, and whether you have the right to license them. Published prices and deals are on the price index and the deals tracker. The guide to data valuation explains the methods.

Which unused records do AI buyers pay for?

Closed records of expert work. The common thread is that a person with training made decisions, and the file shows what they decided and what happened next.

Dark record setWhere it usually sitsWhy it has a marketGuide
Closed support and service ticketsHelp desk archiveProblem, diagnosis and fix in one threadSupport tickets
Engineering and CAD files with revisionsPDM vault, shared drivesDesign decisions and the changes that followedCAD and engineering data
Quality, inspection and non-conformance recordsQMS, spreadsheets, paperDefect, cause and corrective actionManufacturing quality records
Field service and maintenance historyService management systemFault, parts used, time taken, whether it heldField service records
Call recordings with outcomesTelephony or contact centre platformReal conversations tied to a resultCall recordings
Source code and its historyRepositories, old serversHow software was built and fixed over timeSource code
Email and chat of closed projectsMail and chat archivesHow decisions were reachedSlack and email archives

The full list by record type is in the guide to what company data can be licensed.

Which dark data has no market?

Most of the bytes. Knowing what to leave out saves the inventory from becoming a storage audit.

  • Duplicates, backups of backups and exported copies of the same system.
  • Machine logs with no label for what was happening or what went wrong.
  • Contact lists and marketing databases. They are personal data, they age fast and buyers of training data do not want them.
  • Data bought or licensed from a vendor, which is the vendor's.
  • Content customers stored in your product, which is theirs.
  • Scraped web data, which a buyer can collect without you.
  • Records so fragmentary that no case can be followed from start to finish.
  • Files nobody can open because the software is gone.

Contact lists carry a legal problem as well as a commercial one. In California, to sell covers making a consumer's personal information available "to a third party for monetary or other valuable consideration" (Cal. Civ. Code 1798.140).

Is old data too old?

Not because of age alone. A ten-year run of closed cases is worth more than last quarter's, because it has variety and known outcomes. Age hurts only when the work itself has changed so much that the records no longer describe how it is done.

Two things do get worse with time. Formats rot, so export while the system can still be read. And personal data should not be kept forever: European law requires it to be "kept in a form which permits identification of data subjects for no longer than is necessary for the purposes for which the personal data are processed" (GDPR, Article 5). De-identified records are outside that limit, which is one more reason de-identification comes first. See how to keep company data licensable.

How do you find dark data?

Ask each department what it stopped looking at, then count. Nothing has to be exported or opened. System names, record counts and date ranges are enough for a first pass.

Where to lookWhat is usually thereHow to count it
Systems the company still pays for but rarely opensClosed tickets, old jobs, finished projectsRecord count by year from the admin console or a report
Retired systems and their final exportsThe whole history of a product or processDatabase row counts. See retired products and dormant entities
Shared drives and document managementProject folders, drawings, reportsFolder count by project, file count by type
Mail, chat and telephony archivesYears of working conversations and callsMailboxes, channels, hours of audio
Machines and instrumentsLogs, measurements, imagesFiles per machine per year, and whether results are linked
Paper and off-site storageOlder job files and inspection recordsBoxes by year and record type

The company data inventory template has a sheet for this.

What does it cost to prepare dark data, and who pays?

The inventory costs staff hours. Preparation costs more, and who bears it is negotiated. There is no standard split and no published figure, so agree it in the terms before work starts.

Preparation has three parts: export from the source system, removal of personal and customer-confidential information, and a description of the fields. Some buyers take a raw export into a secured environment and do the processing themselves under strict terms. Others expect a cleaned set. The first lowers your cost and raises the care needed in the contract. The second is the reverse.

Do not spend on preparation before a buyer has seen a description and a small sample. The guide to de-identifying company data covers the middle step.

Do you lose the records when you monetize them?

No. Monetizing dark data almost always means a licence, not a sale. The buyer gets a right to use a copy for a stated purpose. The company keeps the records and can license them again.

Keep the licence confidential, too. Federal law treats information as a trade secret only if "the owner thereof has taken reasonable measures to keep such information secret" (18 U.S.C. 1839). See exclusive and non-exclusive data licences and the overview of data monetization.

Example: one-page dark data inventory

A fictional 200-person equipment service company lists what it keeps and no longer uses. The figures are invented to show the method.

Record setSystemVolume and yearsOutcome linked?RightsFirst view
Closed service jobsService management system410,000 jobs, 2009 to 2024Yes: fault, parts, return visitsCompany records with customer namesStrong. De-identify
Technician photos of faultsMobile app storage1.8 million imagesYes, by job numberCompany records. Some show peopleStrong after review
Dispatch callsTelephony platform22,000 hours, 2019 to 2024Partly, by job numberRecording consent to checkPossible
Old CRMRetired in 2018, final export kept60,000 contactsNoPersonal dataLeave out
Machine telemetryVendor cloud3 TBNo link to faultsVendor terms applyLeave out unless linked
Email of the service deskMail archive14 mailboxes, 12 yearsPartlyCompany records with personal dataPossible, with care

Fictional example. Two of six sets carry most of the value, and both were records nobody had opened since the job closed.

Common mistakes with dark data

The first is thinking the problem is storage.

  • Measuring the archive in terabytes instead of cases with outcomes.
  • Deleting closed records to save storage before anyone has counted them.
  • Cleaning everything before a buyer has seen a description.
  • Including customer content or bought-in data because it sits in the same system.
  • Assuming old means worthless.
  • Letting a retired system's export become unreadable.
  • Treating a contact list as the valuable part.

Dark data as a data asset

The records your company closed and forgot are a history of real decisions, and that is what AI developers lack. An archive nobody opens does not earn anything by sitting there, and it does not have to be sold to earn. The first step is the one-page inventory above.

Origin works with buyers and helps a company describe what its archive holds, find out what it could be worth and decide whether to go further. Nothing is exported at the start. Origin Data Partners is an independent data partner that helps companies get their data bought. A buyer may pay Origin when a deal closes.

Does your company hold records like these? See what your data is worth.

Common questions

What is dark data monetization?

Earning income from records a company stores and never uses. In practice it means licensing closed records of real work, such as tickets, engineering files and inspection records, to AI developers for model training, while the company keeps the originals.

How do you monetize archived data?

Inventory the archive by record type and year, pick the sets where each record links a problem, the work and the result, check you have the right to share them, remove personal data, and license a copy for a defined use. Do not clean anything before a buyer has seen a description.

How do you monetize legacy data from old systems?

Export it in an open format while the system still runs, with the schema and audit history. Then treat it like any other record set: count it, check contracts and personal data, and license what passes. A closed, complete history can be more useful than a live one.

Is unused company data worth anything to AI companies?

Some of it. Records of skilled work with the outcome attached have a market. Duplicates, unlabelled logs, contact lists, bought-in data and customer content generally do not, or are not yours to license.

Can you sell old company data?

You can usually license it, which is better, because you keep it. Age is rarely the obstacle. What matters is whether you own the records, what old contracts say, and whether personal data can be removed.

Who pays to prepare dark data for licensing?

It is negotiated. Some buyers take a raw export into a secured environment and process it themselves. Others expect a de-identified set. There is no standard split, so put the answer in the terms before any preparation work starts.

Sources and further reading

What is your company’s data worth?

AI companies pay $20K to $5M for company records. Six questions, no files.