This gap analysis belongs to the guide on context engineering in a real workflow, which walks through one service credit and the three dates on its ticket. Here the job is different: comparing what a review would expect against what a typical decision archive holds, and turning the difference into honest sentences for an introduction.
The six items a context review looks for
A context review asks one question of each past decision: could someone rebuild exactly what the decider saw, and nothing more? Anthropic describes context engineering as curating the best set of information during a model's work, covering instructions, tools, external data and history. A training or evaluation task built from a company decision needs the historical version of that set, so the reviewer looks for these items in the record.
| # | Expected evidence | Why the review needs it |
|---|---|---|
| 1 | Trigger linked to the decision (incident, request, order) | Fixes the facts and the date the rule keys on |
| 2 | Policy or schedule version with its effective dates | Shows which rule was in force, not which one is current |
| 3 | The clause or rule the decider cited | Separates the governing document from the merely relevant ones |
| 4 | Approver identity and authority limit | Shows whether a discretionary step was allowed |
| 5 | Amount or fee in force at the time | Lets a reviewer recompute the result |
| 6 | Final outcome, including any dispute or reversal | Confirms the decision stood, or shows it did not |
A reviewer using a tool such as micro1's flow, which micro1's blog (checked October 18, 2026) says compares a model run's steps, tool calls and output with a task rubric, needs items 2, 3 and 6 to write that rubric at all. Without them there is no answer key.
Expected, typically held, usually missing
Most archives were built to run the business, not to explain decisions later. The table sets the expectation against what operations leads commonly describe, and the shape of the gap.
| Evidence item | What companies typically hold | What is usually missing | Shape of the gap |
|---|---|---|---|
| Linked trigger | Incident and request in one ticketing tool, related by a link | Links for requests that arrived by email | Older years, email-only channels |
| Dated policy version | The current schedule, perhaps a few prior versions | The revision a past decision depended on | Policies edited in place, or version limits reached |
| Cited clause | An outcome status: approved, declined | A note naming the document or clause relied on | Free-text notes, or none |
| Approver and limit | Who closed the ticket | An approval field and the limit table in force then | Goodwill gestures with no approver |
| Amount at the time | Billing history with monthly fees | Rarely missing | Small |
| Outcome and disputes | The credit note or declined status | Disputes and reversals linked back to the decision | Disputes handled in email or a separate system |
The version gap is the one introducers most often miss. Microsoft's documentation explains that SharePoint libraries keep a version history from which earlier versions can be viewed and restored, but also that a library saves only up to a set number of versions and deletes the oldest when the limit is reached. A busy policy library edited dozens of times may have lost the revision a two-year-old decision used. Ask whether policies are published as separate dated documents or edited in place, and whether anyone changed the version limit.
The service credit archive, gap by gap (Illustrative)
A fictional 260-person managed IT provider grants or declines service credits after outages. Its service delivery lead was asked, for every 10 credit decisions in a recent year, roughly how many carry each item. Nobody opened a ticket; these are the lead's estimates.
| Evidence item | Expected per 10 decisions | Available per 10 | Missing per 10 | Cause the lead gave |
|---|---|---|---|---|
| Linked trigger | 10 | 9 | 1 | Some requests still arrive by email |
| Dated policy version | 10 | 4 | 6 | Schedule edited in place until a policy library was set up |
| Cited clause | 10 | 3 | 7 | Directors write "declined" without a reason |
| Approver and limit | 10 | 6 | 4 | Goodwill credits skip the approval field |
| Amount at the time | 10 | 10 | 0 | Billing history is complete |
| Outcome and disputes | 10 | 3 | 7 | Disputes live in the account manager's email |
Read across the rows and the picture is specific. Nearly every decision can be placed in time and priced. Fewer than half can be tied to the rule version that governed them, and only about three in ten record why the director decided as they did. That does not make the archive worthless; it means the reconstructable share is smaller than the ticket count suggests, and that the years after the policy library was set up are likely stronger than the years before.
How to describe each gap in an introduction
An honest introduction names the gap, its scope and its cause. It does not guess at value. These sentences work as templates.
- Dated policy version. "Policies were edited in place until 2023; versions with effective dates exist only from then."
- Cited clause. "Decision notes name the governing clause in some cases, mostly for declined requests; approvals often carry no reason."
- Approver and limit. "Contractual credits record the approver; goodwill credits do not, and the authority table for earlier years was not kept."
- Outcome and disputes. "Disputes are handled by account managers in email and are not linked to the original ticket."
Each sentence tells SourceX where to look and what to expect, which is more useful than a confident claim that "all the history is there". For the related check on access and sign-off before a call, see how to identify missing permissions before a discovery call.
Gaps that narrow a review, and gaps that stop it
Most context gaps narrow the usable share of decisions. A few stop the conversation until they are resolved, and none of them are about the records' content.
| Gap | Effect | What the introducer does |
|---|---|---|
| Missing versions, missing notes, unlinked disputes | Narrows the reconstructable share | Describe scope and years |
| Approver not recorded for one decision type | Narrows that decision type | Name the type |
| Company under 50 full-time employees at peak (contractors excluded) | Stops the introduction | Do not introduce |
| Customer contracts or a parent company control the records | Stops until rights are clear | Ask who can authorize a licensing discussion |
| No authorized contact | Stops until one is named | Find the CEO, or the owner in an owner-run company |
The company also needs documented operating history and rights to license the records. 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.
What stays unknown until SourceX reviews it: the exact reconstructable share, whether any buyer wants this decision type, the de-identification needed for customer names and fees, and any offer. When the company prepares for that review, the context engineering review packet lists what to assemble, and the discovery questions for context engineering cover the first call. Engineering teams with similar decision trails are covered in assessing issue histories and review decisions, and finance-side context in the MCP CFO dashboard example. Other record families sit under data types.
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 describe a client's context gaps in a few honest sentences and the gates above are clear, register as a partner, check who qualifies and read the terms before you introduce the decision-maker.