Product
Upload the finished model, the deck made from it and its sources. TieProof finds the financial claims in each, works out which describe the same fact, checks that their numbers agree and shows the evidence behind every conclusion, whatever tool made the files and with nothing to link or tag.
The files
Read in full: each value as the sheet shows it and as it is stored, with its formula, its headers and its neighbours. A macro is never opened or run.
Read in full: the text and the tables of every slide, each value with its slide, table, row and column.
Accepted as supporting sources. For now a check lists them as not read, and says so above its findings.
.xls, .xlsb, .ppt, .csv, Apple Numbers, Pages and Keynote files, and password-protected files, each with a request to export it and upload the export.
Up to 12 files a check, each up to 50 MB.
Roles and checks
Each file has one role. A check needs a Primary Model or a Final Deck, and New Check says which checks will run before you start.
The model checked against itself: the same fact with two values in one workbook.
The deck checked against itself: a figure that differs from one slide to another.
All of that, and every figure of the deck matched to the model and compared.
The model and the deck compared with each source that is a workbook or a deck.
Accepted, each with the file it replaces. Drift from an earlier version is not checked yet.
Verdicts and coverage
Every finding carries exactly one verdict. Coverage is reported beside the verdicts and is never one of them.
Agreement proven within the tolerance that applies.
Disagreement proven, with exact evidence for both sides.
No claim is made: the evidence is not enough, or the match is not certain. A Review is not a weaker Conflict.
A file, or a check, that TieProof could not fully read. It stands in the coverage above the findings, never on a finding.
A check's own state is drawn in neutral tints: a finished check is not verified, and a failed one is not a conflict.
The evidence of a Review
This pair from the sample deal was matched without enough confidence to claim either way, so it is a Review: both values, where each was found and why they were matched, and no verdict until a person decides.
Operating profit · FY2023 · estimate
medium confidenceTo copy: Model!G11
Where it was found · Within Model · A1:I14
Alder Ridge Systems
Income statement · Currency: USD
Operating profit · FY23E
Formula G9-G10
Model · F11 3,236.8Model · H11 3,803.5Model · G10 580.5Model · G12 94.8Read by xlsx/0.5.0
To copy: Slide 3, table, row 6, column 4
Where it was found · Within Slide 3 · Shape 3
Summary income statement
USD m
Income from operations · 2023E
Slide 3 · Shape 3 · R6C3 3,236.8Slide 3 · Shape 3 · R6C5 3,803.5Slide 3 · Shape 3 · R5C4 (580.5)Slide 3 · Shape 3 · R7C4 2,807.55Read by pptx/0.2.0
The two were not compared, so no agreement or disagreement is claimed.
The scale of the value in Alder Ridge board deck.pptx (Slide 3, table, row 6, column 4) is ambiguous, because two sources in its file state different scales.
Taken as the same fact with a score of 0.9465, medium confidence on provisional thresholds. Features: metric 1 (adds 0.239), label 1 (adds 0.061), embedding 1 (adds 0.066), period 1 (adds 0.297), scenario 1 (adds 0.052), entity 1 (adds 0.092), dimensions 1 (adds 0.046), unit_type 1 (adds 0.037), currency 1 (adds 0.003), scale 0.5 (adds 0.009), headers 0.5 (adds 0.029), nearby_text 0.5 (adds 0.0155), acceptance_ceiling 0 (adds 0), hard_incompatibility 0 (adds 0).
| Feature | Value | Adds |
|---|---|---|
| Metric | 1 | 0.239 |
| Label, as written | 1 | 0.061 |
| Label, in meaning | 1 | 0.066 |
| Period | 1 | 0.297 |
| Scenario | 1 | 0.052 |
| Entity | 1 | 0.092 |
| Breakdown | 1 | 0.046 |
| Kind of quantity | 1 | 0.037 |
| Currency | 1 | 0.003 |
| Scale | 0.5 | 0.009 |
| Headers | 0.5 | 0.029 |
| Words around the value | 0.5 | 0.0155 |
| Held below acceptance | 0 | 0 |
| Cannot be one fact | 0 | 0 |
After the check
Resolve it, ignore it with a reason, mark it not a match or confirm the match, and reopen any of these. Each decision is kept with who made it and when, and none changes a verdict.
A new check in the same project. An ignored finding stays ignored while its values are unchanged, and a pair marked not a match counts against matching next time.
A rule tells TieProof that a label your team uses means a known metric, for one project or the whole organization. A rule never rewrites a verdict.
Invite members as analysts, reviewers or admins. Reviewers and admins read the organization's audit log.
What it does not do