Skip to the content
TieProofRun a Check

The last check before you send

Know it ties before you send it.

TieProof checks a finished financial model, the deck made from it and its sources against each other. Every number that disagrees comes with the exact cell and slide on both sides.

Run a CheckSee a sample check

Start free with no card. The sample needs no account.

The dashboard after a check of Alder Ridge Systems, a made-up company, in 3 files. Its latest result: 6 findings need a reader, 4 Conflict, 2 Review and 244 Verified, and coverage of 100.0%, 3 of 3 files fully parsed. The check is opened at its first Conflict: Gross profit · FY2025 · estimate, 6,314.8 in the model and $7.02B in the deck, a difference of 705,244,000, outside the tolerance of ±5,000,000.

Why the numbers drift

A deal is never one file. Under a deadline, one is updated and another is not.

The model sits in Excel, the summary in PowerPoint and the sources in folders nearby.

  • An example. Revenue for FY27E is restated in the model from 842.3 to 861.9, while slide 17 still shows $842.3M. The two are compared, and the disagreement is a Conflict.

    One file moves, another stays

    The model is restated after the deck is built; slide 17 keeps the old figure.

  • An example. $842.3M in the model arrives on slide 17 as $824.3M, two digits swapped. The disagreement is a Conflict.

    A value is copied wrong

    The right number arrives with the wrong digits, unit, scale, sign or period.

  • An example. Revenue, FY27E in the model and Net sales · 2027 · estimate in the deck are matched as one fact: revenue, FY2027, estimate, in USD. Nothing was linked or tagged by hand.

    The same fact, written twice

    TieProof finds the same fact under two names, with nothing to link or tag.

How it works

Upload, verify, review. From the files to the proof.

01Upload

Add the files. Each is given a role.

Add the model, the deck and any sources. Each file is given a role, and you see what will be checked and the credits it uses before you start.

New Check, played once. The sample's three files arrive one after another and each is given its role: Alder Ridge model.xlsx as the Primary Model, Alder Ridge board deck.pptx as the Final Deck, Alder Ridge budget.xlsx as the Supporting Source. What will be checked: model-to-deck matching and reconciliation, and reconciliation of the model and the deck against each supporting source. This check will use 3 credits. Then Start check is pressed.

02Verify

The engine reads, matches and compares. Deterministic code decides.

Every number is read, put in one form and paired with the claims that describe the same fact. Deterministic code compares them, units, scale and tolerance included.

The check, played once: its four stages finish one after another, reading the files, putting the numbers in one form, finding the pairs that describe the same fact and comparing them, and it is completed with Review. From the sample: 6,314.8 in the model and $7.02B in the deck are each put in one form and matched as one fact, Gross profit · FY2025 · estimate, with a score of 0.9555.

03Review

Coverage first, then the proof. Then you decide.

Coverage comes first, then each finding with both sides, the difference and the tolerance. Resolve it, ignore it with a reason, or mark it not a match.

Review, played once, on the sample: coverage first, 100.0%, 3 of 3 files fully parsed, then the three verdicts' counts and the findings that need a reader. The first Conflict is opened: Gross profit · FY2025 · estimate, with both sides as written and as read, where each is and the cells or text around it, a difference of 705,244,000, outside the tolerance. It can be marked as resolved, ignored with a reason, or marked not a match; here it is ignored as known and accepted.

Fix the files and run it again.

A new check in the same project, charged only for new or changed files. A finding you ignored stays ignored while its values are unchanged.

Running the sample again, played once: the model has changed and the deck and the source have not, so this check will use 1 credit. The finding ignored on the first check is carried over, ignored earlier, since its values are unchanged.

Judge it on a real check. No account needed.

The sample deal is a made-up company's model, deck and budget, already checked, with every finding and its evidence.

See a sample checkRun a Check

What it catches

Four kinds of disagreement, today. Each from the sample deal.

  1. 01

    The model against the deck

    A figure in the deck that disagrees with the model it was made from.

    The model against the deck, from the sample deal: Gross profit · FY2025 · estimate. Alder Ridge model.xlsx, Model · I7, 6,314.8, against Alder Ridge board deck.pptx, Slide 5 · Shape 3, $7.02B. Conflict, a difference of 705,244,000.

  2. 02

    Inside the deck

    The same figure, different from one slide to another.

    Inside the deck, from the sample deal: Gross profit · FY2025 · estimate. Alder Ridge board deck.pptx, Slide 2 · Shape 3, $6.31B, against Alder Ridge board deck.pptx, Slide 5 · Shape 3, $7.02B. Conflict, a difference of 710,000,000.

  3. 03

    Against a source

    A supporting file says one figure; the model or deck says another.

    Against a source, from the sample deal: Gross profit · FY2025 · estimate. Alder Ridge budget.xlsx, Model · I7, 6,314.8, against Alder Ridge board deck.pptx, Slide 5 · Shape 3, $7.02B. Conflict, a difference of 705,244,000.

  4. 04

    Inside the model

    The same fact with two values in two places of one workbook.

    Inside the model, from the sample deal: Equity · FY2024 · estimate. Alder Ridge model.xlsx, Model · H28, 14,525.9, against Alder Ridge model.xlsx, Key figures · D6, 13,944.9. Conflict, a difference of 581,036,000.

Not yet: PDF and Word files are accepted as supporting sources, and a check lists them as not read. Formula patterns, typed-in constants, external links and drift from an earlier version are not checked yet.

The evidence

Every conclusion shows its proof. This is the product itself, not a picture of it.

The product's own screens, drawn from the sample deal's stored result.

Sample checkAlder Ridge Systems, a made-up companyChecked by TieProof's engine, version 0.0.0

What was read

3 of 3 files fully parsed

  • Conflict4
  • Review2
  • Verified244

Coverage100.0%

One finding's evidence

Conflict

Gross profit · FY2025 · estimate

high confidence
Primary ModelAlder Ridge model.xlsxModel · I7
As written
6,314.8
Normalized
Shown in millions, read as 6,314,756,000 USD
In the file
6314756000

To copy: Model!I7

Where it was found · Within Model · A1:I14

Alder Ridge Systems

Income statement · Currency: USD

Gross profit · FY25E

Formula I5-I6

  • Model · H7 5,649.8
  • Model · I6 7,594.4
  • Model · I8 1,404.8
Metric
gross_profit from “Gross profit”
Period
FY2025
Scenario
estimate
Unit
currency

Read by xlsx/0.5.0

Final DeckAlder Ridge …d deck.pptxSlide 5 · Shape 3
As written
$7.02B
Normalized
Shown in billions, read as 7,020,000,000 USD

To copy: Slide 5

Where it was found · Within Slide 5 · Shape 3

Summary

Gross profit of $7.02B in 2025E

  • Slide 5 · Shape 2 Summary
Metric
gross_profit from “Gross profit”
Period
FY2025
Scenario
estimate
Unit
currency

Read by pptx/0.2.0

Difference 705,244,00010.05%higher, outside the tolerance applied, tolerance ±5,000,000 (the precision shown)tolerance ±5,000,000 (the precision shown) · outside
Why these two were matched

Taken as the same fact with a score of 0.9555, high 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 1 (adds 0.018), headers 0.5 (adds 0.029), nearby_text 0.5 (adds 0.0155), acceptance_ceiling 0 (adds 0), hard_incompatibility 0 (adds 0).

The first side and the second side: Taken as the same fact, with values that disagree, score 0.9555, high confidence on provisional thresholds. Scored from its features alone. Features of version 9+09704bc1d85a.
FeatureValueAdds
Metric10.239
Label, as written10.061
Label, in meaning10.066
Period10.297
Scenario10.052
Entity10.092
Breakdown10.046
Kind of quantity10.037
Currency10.003
Scale10.018
Headers0.50.029
Words around the value0.50.0155
Held below acceptance00
Cannot be one fact00

Open this finding in the sampleSee every finding of the sample

  • Both sides, exactly where

    Each side's file, role and cell or slide, with what stood around it.

  • As written, and as read

    The raw value beside the value read, with its unit, currency and scale.

  • The comparison

    The difference, the tolerance, and why the two are one fact.

  • Kept with the results

    Stored when the check runs, so it outlives the raw files.

Three verdicts, never a guess. Each one here is a finding of the sample deal.

  • Verified

    They agree within the tolerance, with both sides proved.

    Gross profit · FY2025 · estimate

    Model · I76,314.8

    Slide 2 · Shape 3$6.31B

    Difference4,756,000inside ±5,000,000

  • Conflict

    They disagree beyond the tolerance, with exact evidence on both sides.

    Gross profit · FY2025 · estimate

    Model · I76,314.8

    Slide 5 · Shape 3$7.02B

    Difference705,244,000outside ±5,000,000

  • Review

    Not enough to decide, so a person looks: never a guessed Conflict.

    Operating profit · FY2023 · estimate

    Model · G113,708.1

    Slide 3 · Shape 3 · R6C43,708,097K

    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.

Coverage is reported apart and is never a verdict: 100.0%, 3 of 3 files fully parsed.

Security and privacy

Your files, handled as a deal's files should be.

  • Only your organization

    Only your organization can see your checks, enforced by the database itself.

  • Raw files deleted

    Raw files are deleted 7 days after a check ends by default; an admin can choose from 1 to 30.

  • Evidence kept

    The evidence of each finding is kept with its results, so they stay readable after the files are gone.

  • No external AI model

    No external AI model is used today: every finding is decided by deterministic code.

How TieProof handles your files

Pricing

Pay for the files you check. One credit checks one file.

Credits, played once. A first check of three files uses 3 credits, one for each file. Run again with one file changed, it uses 1 credit, and the unchanged files are free. A check that fails or is cancelled takes nothing.
  • A check shows its credits before it starts.
  • A failed or cancelled check costs nothing.
  • A check run again charges only for new or changed files.
  • Start free with no card; Pro starts at $29 a month.
  • Payments are not open yet: until they are, no card is asked for and no check is charged.
See pricing

Questions

Before your first check. Plain answers.

Rather look than read? The sample needs no account.

See a sample check
Which files does TieProof read?

Excel workbooks (.xlsx and .xlsm) and PowerPoint decks (.pptx): the model, the deck made from it, and supporting sources that are workbooks or decks. PDF and Word files are accepted as supporting sources, and for now a check lists them as not read.

Does it change my files?

No. It reads the copies you upload and writes nothing back to them. Macros, scripts and embedded programs are never run.

What happens to my files after a check?

Raw files are deleted 7 days after a check ends by default, and an admin can choose from 1 to 30. The evidence of each finding is kept with the results, so a finding can still be read after the files are gone.

Does AI decide the findings?

No. No external AI model is used today. Every finding is decided by deterministic code, and when the evidence is not enough the answer is Review, never a guessed Conflict.

What is a credit?

One credit checks one file. A check shows its credits before it starts, a check that fails or is cancelled costs nothing, and a check run again charges only for new or changed files.

Who can see my checks?

Only your organization. The database itself shows each organization only its own rows.

Can I try it without uploading anything?

Yes. The sample deal is a made-up company's model, deck and budget, already checked, and it opens with no account. When you want to check your own files, you can start free with no card.

Check the files before they go out.

Add a model and its deck, and read the proof of every number that disagrees.

Run a CheckSee a sample check