15 minutes, live on a call. Bring a real report. No contract to sign first.
Built by Lido, the document platform used by accounting and audit teams
And of the tool that cries wolf until you stop listening.
Legacy footing software sums a budget variance column like a total, then flags it. On every page. Staff stop trusting the flags and go back to the ten-key, while the licenses keep billing.
Every revision that comes back from the preparer has to foot again. On big engagements the rounds run into the 40s and 50s, right through the filing deadline.
In oversight states like Louisiana, the legislative auditor can bounce a report over one number that does not foot. Hard deadline, full revision round, your firm's name on the opinion.
Auditors on the footing software they pay for today, from recorded calls:
It says everything doesn't foot. I'm looking at it with my eyes saying, I think it foots. I go through it, and sure enough it does.
It's always telling me that column 4 is not the total of the first 3 columns. Which of course it's not. It's not supposed to be.
The report was about 110 pages. 30 of them had schedules that needed footing. I had to manually do 28 of them.
Footing tools fail in one of two ways: they misread the page, or they let a language model do arithmetic. This one splits the job.
A first pass reads each schedule and writes down which figures feed every printed total. Variance columns are read as variances, not sums, based on what the schedule actually presents.
Arithmetic runs deterministically, on the server. A language model never adds two numbers. If a total does not foot, it is because the numbers on the page do not foot.
Every failed formula gets reviewed under auditor rules, so the model cannot invent a formula that happens to make a bad column look right.
Real errors with page references, rounding differences kept apart so they never bury a finding, and everything that foots so you can move on.
A plain-text instructions field covers firm-specific quirks, like the client that presents every variance as a positive number.
What happened when a government audit firm put a real report through it
The whole report in about ten minutes. Not just the schedules someone had time for.
Not a page of false positives to re-check by hand.
In a report the firm had already issued.
From a recorded trial call, July 2026. The reviewer had spent about six hours on that report by hand.
Their footing module is also named Financial Statement Suite. The resemblance ends at the name.
DataSnipper is incapable of doing budgets for governments. The amount of money you pay for a program like that, it's frustrating. Incredibly frustrating.
I don't remember it working this well on DataSnipper. We just issued this report in June, so it's been less than two months since I looked at it.
The second check auditors asked us for: internal consistency. Total long-term liabilities shows up on the face, in the notes, and again in the MD&A, and it has to be the same figure everywhere. On a 100 page report, some numbers repeat a dozen times. Checking that by hand is a full review day.
The suite indexes every value in the document. Click one and you see each page where it appears, and where a figure that should match does not. It catches disagreements between the face and the notes, and it points at real printed pages instead of cross-reference codes that mean nothing on a PDF.
I showed this to someone else I work with and they were like, oh yeah, that's awesome. If the number is called one thing in one part, it's got to be the exact same number everywhere throughout the report. With a report that's 100 pages, some of these numbers appear quite a few times. It's very time-consuming to check.
+ every fund statement in between. No templates to build first.
Reports do not get footed once. Every revision that comes back from the preparer has to foot again, and on big engagements those rounds stack up right through the filing deadline. A re-run here takes minutes, so the check that used to be the bottleneck becomes the default. Comment in Adobe the way you already do, send it back, run the next version.
For small, easy audits, it might be 3 or 4 times back and forth. For some, I've gone into the 40s and 50s.
Uploads are wiped automatically. The annotated PDF carries the results, and re-uploading it restores the view without reprocessing.
Your clients' statements are not training data. Not for our models, not for anyone's.
Report available for your security review.
Available for firm-wide rollout.
Send the security questionnaire before the demo and we will show up with it answered.
Pick a slot. If IT wants answers first, send the security questionnaire ahead and we will show up with it filled in.
Upload a statement from your own shop on the call. No sample files, no staged demo. Bring the one your current software chokes on.
Errors, rounding, and passing come back in about 10 minutes. You look at the exceptions, not the whole report.
The annotated copy is yours either way, along with per-report pricing to put next to your renewal quote.
If it does not catch anything your current process misses, you are out 15 minutes.
This page will host a live version of the tool. Upload a statement, watch the buckets come back.
No. DataSnipper sells an Excel add-in module by that name. This is an independent tool built by Lido that runs in the browser on the issued PDF. Firms evaluating both usually run the same statement through each and compare the exception lists side by side.
Each schedule is read contextually, so a variance column is computed as final budget minus actual rather than summed like a total. There are no hand-coded column rules to maintain, which means the fifth column GASB 103 requires for fiscal 2026 year ends is read the same way. A plain-text instructions field covers firm-specific presentation quirks.
No. A first pass reads each schedule and writes every total as a formula. The arithmetic then runs on the server, deterministically. A second pass re-examines failed formulas under consistency rules, for example requiring the same formula across columns of one schedule, so the model cannot invent a formula that happens to make a bad column pass. Results come back as errors, rounding differences, and passing.
Uploaded documents are deleted automatically within 24 hours and are never used to train AI models. A SOC 2 report is available for your security review, and SSO is available for firm rollouts. The annotated PDF you download carries the results, so re-uploading it restores the review view without reprocessing.
Up to about 10 minutes for a report near 100 pages. It runs unattended, so you upload, go back to work, and come back to the buckets. Statements drafted in Word are fine: you run the exported PDF, which is the version that actually gets issued.
Pricing is per report, tiered by page count. There are no seat licenses, so you never pay for staff who do not use it, and there is no per-page meter to predict. Bring your annual audit count to a demo and leave with a number to put next to your current renewal quote.
A 15 minute call. Upload a real report from your shop, watch the buckets come back, and take the annotated PDF with you. If it does not catch anything your current process misses, you are out 15 minutes.
Renewal coming up? Bring the quote.