Yearly running cost Assuming 1,200 models a year
By hand $54K / year 900 hours of work
Subscription n.a.
This workflow $20 / year Machine usage only; setup, hosting and review are extra. How should it receive your financial model? Find formula issues, mismatched figures and unsupported assumptions before your analyst signs off.
One model includes up to 5,000 populated cells and 20 agreed model checks.
Upload my models Check a workbook before analyst review $1,152 – 3,240 setup Connect my SharePoint folder Check saved models in my existing process $1,512 – 4,320 setup
Which folder should it read?My SharePoint folder Your analyst reviews every model. No automatic formula changes, investment recommendations, trades, source-file overwrites or external report sending.
What is included Apply the agreed checks to versioned workbook copies. Show cell-level findings, calculation evidence and unresolved areas; do not change the source model.
Start with ordinary unencrypted Excel workbooks and saved calculation results. Add an approved folder connection for repeat reviews. Macros, live feeds and full model construction are outside the starting scope.
Workbooks per review: up to 1 workbook.
Worksheets per workbook: up to 12 worksheets.
Formula cells per workbook: up to 2,000 formula cells.
Compressed workbook size: up to 10 MiB.
Monetary calculation tolerance: up to 1 cent.
Percentage calculation tolerance: up to 1 basis point.
How reliably should it find model issues? Find real model issues, minimise false flags and show exactly what was checked.
Standard Find the agreed issues and show cell evidence, with unknown results left for analyst review. Model issues found: ≥95% Correct issue flags: ≥95% Included Strict Add more difficult models, independent verification and checks for misleading assumptions and links. Model issues found: ≥98% Correct issue flags: ≥98% Setup +$288 – 720 My own targets Agree more complex workbooks, formula handling and financial checks with your provider. Quote separately
How these standards are measured Use unseen financial models with independently labelled issues and correct cases. Check missed issues, false flags, calculations and exact cell evidence. Unsupported or stale results stay unresolved, not passed.
Targets for your selected standard What is checked Target Model issues foundFlagged true issues divided by all labelled in-scope issues. Keep missed and deferred issues in the denominator. An unusual formula or hardcoded value is not automatically an error. ≥95% Correct issue flagsConfirmed issues divided by all proposed issue flags. Fix the rule definitions before testing; legitimate exceptions and intentional modelling choices must not become invented errors. ≥95% Checked financial arithmeticCompare every agreed numeric check with an independently computed reference under the same units, periods and rounding. Missing or unverified results cannot pass as correct. ≥100% Valid cell evidenceEvery finding resolves to the exact workbook version, sheet and cell with the rule, formula or calculation needed to verify it. ≥100% Unresolved checks shownShow unsupported formulas, missing external sources, stale or absent saved results and inaccessible inputs. A partial review must not be described as a clean whole-workbook audit. ≥100%
These are proposed pilot targets, not a full model audit or proof that an investment is sound. Analysts review every model and resolve unsupported areas.
How quickly do you need each model checked? Choose how soon the findings should be ready for your analyst.
Within 3 minutes Included Within 90 seconds Setup +$72 – 360 Within 45 seconds Setup +$216 – 600
Timing details Time isolated parsing, agreed checks, evidence, model explanations, queues, retries and stored findings. Buyer-side workbook recalculation and analyst sign-off are separate.
The target applies to at least 95% of agreed test runs, with 2 in progress at a time.
How much do you want to spend per model? Choose the machine budget for checking the model and explaining its findings.
Up to $0.50 Included Up to $0.30 Setup +$72 – 360 Up to $0.15 Setup +$216 – 600
Cost details Includes local parsing, deterministic checks, model explanations, retries and shared hosting at the reference workload. Finance software, data subscriptions and analyst review are separate.
Reference machine cost: $0.22 – 0.42 per model at 100 models a month. The selected cap is a target to test, not a replacement for this estimate.
Where do you want it to run? Run in your cloud or on your machine. Keep checks local; send only approved extracts to an external model for explanations.
Run it onYour cloud Local environment AI model accessApproved model API No external calls
Data and access details Runs in an account you control with isolated parsing and a review log.
Use only authorised formula and finding extracts for explanations. Checks and financial arithmetic remain local.
Do not send confidential deal data without permission. Private mode uses local files and templated explanations; it cannot read a live SharePoint folder. Missing calculation results need analyst attention.
How do you want to use it? Review the findings on a web page, ask your own agent or use a dedicated model-review agent.
A model review page Upload a workbook and inspect findings beside the relevant sheet and cell. Included My existing AI agent Ask your agent to check a model and return cell-level evidence. Setup +$72 – 240 A dedicated review agent Use a separate agent to organise model checks and unresolved findings. Setup +$144 – 480