Abstention examples evidence gaps: what a review expects, what contract archives keep, and how to say what is missing

Short answer

Five gaps make a history of contract holds hard to assess: the package was not kept as it first arrived, the hold reason is blank or vague, the escalation went somewhere that left no role on record, earlier versions were overwritten, and the signed outcome sits in another system with no link back. None of these rules a company out. Each one narrows which years or contract types a review can use, so the honest introduction names it.

Visual briefFull-size diagram
Five pairs of vertical columns compare the evidence a review expects with what a fictional IT services firm holds for its contract holds: arrival state 35%, hold reason 60%, escalation role 75%, version history 45% and linked outcome 55%, with each missing share outlined in orange.
Illustrative gap analysis for a fictional 240-person IT services firm. Percentages are invented for this page.

This page is part of abstention examples: five questions for introducers. Its job is a gap analysis: what a review would expect, what companies usually keep, where the difference lies and how to describe it without rounding up. To see a complete chain first, read the step-by-step workflow example.

The evidence a review expects from a history of holds

An abstention example is only assessable when a reviewer can answer three things: was something really missing when the person stopped, did the stop follow the company's own rule, and did stopping lead to a better result than proceeding would have? Five kinds of evidence answer them.

Expected evidenceThe question it answers
Arrival state of the packageWas the item actually missing or mismatched at the moment of the hold?
Hold reason naming each gapWhich rule triggered the stop, in the words of the person who stopped?
Escalation recipient and roleDid the case reach someone allowed to resolve it?
Version history of the documentsDid two versions really coexist, and which one was reviewed?
Outcome linked to the holdWas the contract signed, amended or lost, and does the executed set match the reviewed one?

The interest behind this list is a research one. OpenAI's explainer on hallucination argues that scoring by accuracy alone rewards guessing over admitting uncertainty, and a 2025 survey in Transactions of the ACL reviews how abstention is studied and measured. A review of business records is looking for the human version: a documented stop that was right. Vendors such as Appen market failure-mode taxonomies and rubric design for agent evaluation (Appen's agentic AI page, checked October 18, 2026), which shows the category exists, not that any one archive is wanted.

Expected versus typically kept: the gap table

Companies rarely hold all five cleanly. The pattern below is what ordinary contract archives tend to look like, described from how the tools are commonly set up, not from a survey.

GapWhat a reviewer cannot tellCommon causeMetadata question to ask
Arrival state not keptWhether anything was missing when the hold was placedThe contract tool stores only the latest attachment set"Does the system keep the document list as it was on first upload?"
Hold reason blank or free textWhich gap triggered the stopStatus can change without a required reason field"Is a reason required when a contract goes on hold?"
Escalation role not recordedWhether the case reached someone allowed to resolve itEscalations sent by personal email"Is legal intake a queue or a named role, and is the assignment logged?"
Versions overwrittenWhether two versions of an exhibit ever coexistedFiles replaced in place under the same name"Are earlier versions retained, or replaced?"
Outcome stored elsewhereWhether the contract was signed, amended or abandonedExecuted copies kept in a shared drive or the CRM"Can the signed copy be traced back to the hold?"

The first gap is the most damaging because nothing else can repair it. If the record shows only the final, complete package, a careful hold and a needless delay are indistinguishable.

A gap analysis worked through (Illustrative)

Illustrative and fictional. A 240-person IT services firm runs about 400 customer renewals and statements of work a year through a contract tool it adopted in 2021. Before that, contracts lived in a shared drive. An introducer asks the head of commercial operations the five metadata questions and records rough shares of the roughly 600 holds logged since 2021.

EvidenceExpectedHeld (rough share of holds)The gap in one line
Arrival stateEvery hold35%First uploads are kept only where staff attached the original email
Hold reasonEvery hold60%A reason field became mandatory in mid-2023
Escalation roleEvery hold75%Most go through the legal queue; some went by direct email
Version historyEvery hold45%Exhibits are often re-uploaded under the same file name
Linked outcomeEvery hold55%Executed copies for older deals sit in the CRM without a contract tool ID

Read across, the firm has a usable core: holds since mid-2023 with a mandatory reason, routed through the legal queue, where the original email was attached. The shared-drive years before 2021 carry almost no hold history at all. That is still a reasonable introduction, provided nobody describes the whole archive as complete. The graphic on this page plots the same five shares.

Describing a gap honestly in an introduction

A gap described plainly is useful to a review. A gap rounded up becomes a problem later. The right column is what to write instead.

Instead ofSay
"They keep every contract.""Contracts since 2021 are in a contract tool; earlier ones are in a shared drive."
"Every hold has a reason.""A hold reason has been mandatory since mid-2023; earlier holds often have none."
"Legal handles all escalations.""Most escalations go through a legal queue; some went by direct email."
"Full version history.""Exhibits are sometimes replaced in place, so older versions may be gone."
"Outcomes are tracked.""Signed copies for recent deals link back; older ones sit in the CRM."

Two habits help. Give years rather than adjectives ("since 2021", not "recent"), and name the person who gave each answer, so the review knows whom to ask next. The M&A contract review resource shows how contract records are described in a deal setting, and assessing business templates and their completed examples covers a related check on paired records.

What gaps do not change, and what stays unknown

Gaps do not change who must own and authorize. Contract records usually belong to legal or a contracts team, and authority to discuss licensing sits with the CEO, or the owner in an owner-run company. Customer contracts can limit reuse of their terms, so a rights review comes either way.

They also do not change the company test. The company needs 50+ full-time employees at peak (contractors excluded), documented operating history, rights to license and an authorized contact. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English.

Unknown until SourceX reviews it: the true share of usable holds, whether control cases (packages routed cleanly) survive alongside them, what redaction customer names need, and whether any buyer wants this task type. Rare, well-judged stops can be valuable precisely because they are uncommon; why AI models need rare examples explains the logic. What the company would later assemble is set out in the review packet contents list, and the wider record family sits under contract review workflows.

Next step

Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.

If you can state a client's hold-history gaps in the plain terms above and name who owns the records, register as a partner, check who qualifies, and read the program terms before you make the introduction.

  1. Step 1Share your linkSend your personal link to a company you know.
  2. Step 2Company appliesThe company applies itself at /apply.
  3. Step 3Company signs and is paidThe company signs, delivers its data and is paid by SourceX under its agreement.
  4. Step 4You get your rewardthe greater of $50,000 plus 5% and 15% of the actual supplier payout, with no cap. A reward becomes payable after SourceX has paid the referred company under its signed agreement and the payout conditions in the program terms are met, including finance approval and your tax and payment details.

Common questions

Does a company with gaps still deserve an introduction?

Often, yes. Gaps narrow which years or contract types can be assessed; they rarely erase the whole archive. A company that moved to a contract lifecycle tool in 2021 may hold usable holds from that year on, and saying so plainly helps the review.

Which gap does the most damage?

A missing arrival state. If the archive shows only the final, complete package, a well-judged hold and a pointless delay look the same, and no other field can repair that.

Can the introducer check the gaps by looking at a few held contracts?

No. Partners never handle client records. The introducer asks how the system behaves (does it keep first uploads, is a hold reason required) and writes down the answers and who gave them.

Sources

  1. openai.com/index/why-language-models-hallucinate
  2. aclanthology.org/2025.tacl-1.26
  3. appen.com/agentic-ai
  4. refer.sourcex.si/terms

Free resources

Know a company with valuable proprietary data?

Become a referral partner, get your link and introduce the company’s CEO (in an owner-run company, usually the owner). Referral partners can join from most countries. We can't work with partners in countries or regions under U.S. sanctions or on U.S. restricted-party lists. The companies you introduce should be based in the U.S., Canada and the EU, or in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia with at least 70% of their data in English.

Refer a company →

I own a business

Explore licensing your company's data directly to SourceX for AI training and evaluation. Start a short assessment; no uploads needed.

Start an assessment