Data licensing resources: guides, programme comparison, data readiness

The two programmes, side by side

Both license internal records from established companies. They differ on who they want, how they price and what you hand over. Many companies qualify for both.

	micro1 Company Data Partnerships	Mode Data Partnerships
Published payouts	Tiers from $100K+, $500K+ and $1M+	$100K–$5M
Backing	$1bn, 12-month commitment announced 9 Oct 2026, with capital from Citi and Hercules Capital	Publishes a pricing formula based on staff numbers and years of records
Company size	30+ staff	20+ office staff; accounting firms from 10, law firms from 6
Where	Many countries; strongest demand in the US, UK and Canada	Mainly US companies
What they look for	English-language documentation of how work gets done	Years of internal records from office teams
What you hand over	Agreed in scope after review; micro1 states companies keep ownership	Mode exports and de-identifies an agreed copy; no passwords handed over
Process	Short assessment on their site, review, then a discovery call	Estimate, then a non-binding letter of intent, then an offer
Timeline	Not published	About three months from first conversation to payment
For advisers	Referral scheme with a sub-affiliate feature	Referral scheme with a second-level referral feature
	Apply to micro1 →	Apply to Mode →

From each programme's own published material, October 2026. Terms can change; check their sites before relying on anything. These buttons are our referral links: the programmes may pay us a referral fee, which never comes out of your payout. Not sure which fits? The 60-second fit check matches you.

What licence-ready data looks like

Programmes pay for the work: the problem, the steps, the judgement calls and the outcome. Personal details have to come out first. Flip between the stages to see where people go wrong.

Illustrative Every record on this page is invented by TrackImpact. No real data is shown.

Links between records are where the value is

A message on its own teaches little. A message that leads to a meeting, a decision, a document and an automation shows how work actually gets done. Commercial datasets are built synthetically to imitate this. Your company already has the real thing.

Teams thread

A request

user: staff_B
act: request
"Can you review the Q2 budget before Friday?"

Meeting notes

A decision

attendees: staff_B, staff_G
decision: proceed with
Q2 plan, cut travel 10%

Action item

An owner and a date

owner: staff_G
due: 2026-04-05
"update budget + share"

Document

The work, versioned

doc: budget_q2
v1 → v2 (staff_G)
approved: staff_B

Outcome

What good looked like

status: approved
time: 3 days
automation: monthly
reminder created

Illustrative Invented example. Note the consistent stand-ins (staff_B, staff_G): who did what survives without anyone's name.

Six ways to make your data worth more

What we saw again and again in commercially licensed datasets.

01 · Structure

Tidy beats big

Consistent fields, stable IDs and timestamps on every record. A clean export of one system is worth more than a dump of five.

02 · Links

Keep the threads together

Ticket to fix to review; brief to draft to sign-off. Don't break the IDs that connect them when you export.

03 · Outcomes

Label what happened

Resolved, approved, placed, rejected, escalated. Reviewer comments too. That's how an AI agent learns what good looks like.

04 · Stand-ins

Replace names consistently

The same person or client gets the same tag everywhere ([client_9], staff_B), so the story survives without the identity.

05 · Datasheet

Describe each source

One row per source: what it is, language, date range, size, contributors and how the team worked. Template below.

06 · Sample

Start with a small sample

Ten to fifty representative records, in the same structure as the full set, is normal for a first review.

A one-row datasheet for each source

Buyers expect a short description with every dataset. Fill one row per system you plan to license.

Source	What it is	Language	Date range	Size	Contributors	How the team works	Personal data
Service desk	Customer tickets, notes and resolutions	English	2018–2026	~48,000 tickets	22 agents	Two-tier support, weekly QA review	Replaced with stand-ins
Process wiki	SOPs, checklists, onboarding	English	2016–2026	~1,900 pages	35 editors	Owned by ops, reviewed quarterly	None found
Project tracker	Projects, tasks, comments, status changes	English	2019–2026	~6,200 projects	60 staff	Small project teams, PM sign-off	Names replaced

Illustrative Invented example for a 60-person services firm.

Check these before anything leaves the building

The four most common ways company data leaks after it has been "cleaned". Easy to miss, hard to undo.

The key file goes with the data

A log of what each tag replaced ("customer_1 = Sarah…") undoes the whole redaction. Keep it internal, or don't create one.

Names in free text

Fields get redacted; notes, quotes, reviews and file names don't. So do business names, phone numbers and short forms of names.

Secrets in code and logs

Passwords, API keys, private keys and database dumps turn up in repositories and debug logs. Scan for them before any export.

Scans instead of exports

Scanned documents carry text-recognition errors. Export from the system itself wherever you can.

The programmes state they remove or de-identify personal information. One exports and de-identifies the agreed copy itself. Agree in writing who does what, and have counsel check your contracts and data-protection obligations first.

Questions to get answered before you sign

Much of what matters sits in the licence, not the headline number.

Term: is the licence perpetual, or for a fixed period? Commercial dataset licences are often perpetual.
Use: can the buyer commercialise models trained on your data? Usually yes; check for limits.
Exclusivity: can you license the same material to anyone else?

Onward use: can they resell or sublicense your data? To whom?
Restrictions: can you exclude use cases, competitors or regions?
Security: ask about their GDPR position and certifications such as ISO 27001, ISO 27701 and ISO 42001.

The full checklist is in section 8 of our open guide. Not legal advice.

Everything you need to license your company's data to AI programmes.

I run a company

I advise companies

I own several companies