Credentials
Read a model's build record: what was built from your brief, what was left out and assumed, the checks it passed, and the independent review.
Where: Overview→Credentials
Credentials is a model's standing record of how it was built: the decision it makes, its capabilities and the words in your brief behind each, what was deliberately left out, the checks that proved it, and an independent review against your brief. When the build waits on you, you accept a design or your data here.
Open the record
Click the Credentials door card on model home; its preview counts the checks that passed and, until the model is ready, the review's confidence. Models built from a description or through the New Model steps have this door; a model imported from a workbook has a Model Design door instead. When a build pauses for you, the status card and Step 2's Review the design open the same record. It's headed with the model's design name and a status line. The refresh icon (Refresh dossier) reloads it; while the model is importing, mapping or solving it refreshes itself.
| Status line | Means |
|---|---|
| "Structure built and proven on test data — awaiting your review." | The design is waiting for you to accept it. |
| "Structure built and matched on sample values — awaiting your review." | The same, with the checks run on sample values. |
| "Structure built — not every check matched. Awaiting your review." | Some checks didn't match. See why. |
| "Your data is bound and validated — awaiting your approval." | Your data is mapped and clean. |
| "Your data is bound — blocking findings need your decision." | Your data has blocking findings. |
| "Your data has fatal findings — fix and re-upload." | Your data can't be built on. |
| "Structure accepted — running on sample data until you load yours." | Ready, on sample data. |
| "Built from your data on the verified structure — ready to use." | Ready, on your data. |
| "The build stopped — details below." | See When the build stopped. |
| Working… | Something is running. |
If the model was deleted, the record says "This model is no longer available — it may have been deleted." with Back to the model library. An older model built another way says it has no record.
Read the independent review
The seal on the right of the header shows the latest review's confidence: N% and confidence · after the design (or after the build, after the solve) · advisory. Green is 80% or more, amber 55% or more, red below. While the review runs, the seal says it's working in the background; you don't need to wait. REVIEW FAILED means the review itself errored. Click the seal to open Independent review, every round listed newest first:
- the reviewer's rationale;
- How this was scored: Matches your brief, Covers what you asked for, Built correctly and Honest about its limits, each out of 5;
- Missing: what your brief asked for that the model lacks, with the quote and why it matters;
- Added without a basis in your brief;
- Things we added, and whether your brief backs them, each backed, not backed or ?.
Here the review is read-only: you add or leave out a miss at the build (see Build and baseline), and each miss is also on the model's backlog.
Read what was built
What was built sums the model up in tiles: the goal (Max profit or Min cost), and how many sites, products and accounts it covers. Capability chips follow, with any added by AI rather than from your brief in a warning colour, then a count line (N capabilities · N added by AI — flagged for your judgment · N declared cuts) and the proof line, such as "All N capability checks proven on test data." Until the model is ready, a link here also opens the independent review.
Below it, the Model Design card shows the decision, scope, rules and assumptions. The folds that follow show their count in the header and open by themselves when they hold a problem, unless you closed them: eode remembers which folds you opened or closed, in this browser, for every model.
| Fold | Shows |
|---|---|
| Capabilities · evidence | Each capability, with up to two quotes from your brief behind it. AI-suggested marks one with no matching words in your brief: keep it only if the business really needs it. structural marks the model's own scaffolding. |
| Declared cuts · assumptions | The parts of your brief deliberately not modeled, and what was assumed. NONE when nothing was cut. |
| Completeness ledger | Dimensions a model like this usually needs, each marked can't model, usually needed, usually has or optional. "Open items are honest gaps, not errors." Answer them by refining your brief. COMPLETE when nothing is open. |
| Proof ledger · test data / Proof ledger · your data | The build's checks; see below. |
| Requirements | What you said the model must do; see Change a model. |
| Data to load | The tables to provide, with filled templates to download; see Load your data. |
| How this is built (for modelers) | The mechanics; see below. |
Planning horizon
When your brief gave a cadence but no count, Declared cuts · assumptions opens on "Planning horizon: I assumed …" with the same choices as on the design card; see Review the design.
The proof ledger
The banner states the verdict: N of M checks passed · verified in a fresh session, N of M checks matched on sample numbers, or Checks not passing · N of M passed. On sample numbers it adds "These checks ran on sample numbers, not yours. They prove the structure works. Your numbers are checked once your data is loaded." The pill reads VERIFIED, N/M MATCHED or FAILED.
The columns are Check, Part of the model, wanted and got. Checks are named in words, like "Every order is delivered" or "Equipment stays within its hours". A table that failed to load is listed under the ledger.
How this is built
How this is built (for modelers) holds the mechanics, each zone stating its counts in its header:
| Zone | Shows |
|---|---|
| Rules and how they are encoded | Each rule and whether it's proven on this model, proven on the catalog or not yet proven. |
| Where your data lands | Each sheet and the table it lands in, with row counts and whether they agree. |
| Checks | The build's checks, with wanted N · got N. |
| Build record | The state, time per phase, total time, the goal's value and the model's size. |
| Limitations | When the build stated any. |
Show engine names, off by default, adds the technical name behind each row. A footer identifies the record and when it was confirmed.
Decide on a pending design
While a design waits for you, a card explains it: "Here's what RIA built, in plain English — your last look before anything's committed." When checks didn't match, Why these checks didn't match explains the cause. Under Your decision:
| Button | Does |
|---|---|
| Accept structure | Accepts the design. Scenarios and what-ifs unlock, on sample data. |
| Load your data… | Opens the data tabs in place. On a model mapped from a workbook it reads Load a revised workbook…. |
| Use the numbers from your brief | When your brief states data, RIA transcribes it and binds it the same way. Transcribed from your brief then lists each table, what was skipped and the arithmetic applied. |
| Send back | Asks "What was wrong? (e.g. the capacity sheet had last year's numbers)". Send back with this reason returns the model to its questions with your answers kept. |
The design itself, with its own sign-off, is on Review the design.
Decide on your data
Once your data is mapped, Your decision shows a pill (FIX & RE-UPLOAD, BLOCKED or PENDING), a line counting tables bound, still on sample data and not bound, and the validation result. Data: links to the dataset version.
- Accept mapping · build with real data builds on your data.
- Upload corrected data… (or Load a revised workbook…) replaces it.
- Data validation lists rejected rows by table, up to 200 findings. Fatal findings hide Accept.
To build despite blocking findings, open Accept with exceptions, tick "I reviewed the N blocking finding(s) and accept them knowingly (recorded in the dossier + audit log).", type a reason, and click Accept with exceptions · build with real data. From then on the record carries a banner: "This model was built on data accepted with N blocking finding(s) outstanding." with your reason and the date.
Mapping manifest counts what's bound, still sample data and unbound; its Full mapping report fold holds the full report as text. How the mapping itself is reviewed is on Load your data.
When it's ready
- On sample data, Accepted reads "Structure accepted. It's running on sample data — usable for what-ifs now." with Load your data…, Model home and Scenarios →. Finishing your model shows the last steps: Reading your base-case results, Building the topology, Gathering market factors, then "Ready — explore the model below."
- On your data, Full-data build (READY) shows the goal's value and "Built from your data on the verified structure; every proof re-ran green." with Explore scenarios → and Open model home.
- While your data is being read, Your data offers Check the reading, which opens the reading step.
When the build stopped
Stopped (ERROR) names the stage and what went wrong, with the same recovery panel as model home's Let's get this unblocked. A Technical detail fold holds the raw error for support. When the build can pick up where it stopped, a button resumes it.
Limits
- Up to two brief quotes per capability.
- Data validation lists up to 200 findings; the counts above it are complete.
- The proof ledger scrolls past about nine rows.
Related
- Validation: the model's standing checks, re-run any time
- Review the design: the Model Design card and its sign-off
- Build and baseline: where the record is first written
- Change a model: the backlog, where cuts and misses can come back