I built TidyLedger to turn messy invoices into clean Excel reports
The problem I’m trying to solve is simple: collecting information from invoices and putting it into a usable spreadsheet can be repetitive and time-consuming. With TidyLedger, you can: • Upload invoices as PDFs or images, as well as text and CSV files • Choose, rename, and reorder report columns • Review and edit the extracted data • Download your results as Excel or CSV There’s also a free plan with up to 5 invoices per day. 🚀 Try it here: https://tidy-ledger-1.onrender.com/ I’m now looking for honest feedback from people who work with invoices, spreadsheets, or business data. I’d especially love to know: 1. Is this a problem you face in your work? 2. What would you change or improve? 3. What features would make you want to use a tool like this regularly? This is an early-stage product, and I’m building it to solve a real problem. Any feedback or suggestions would mean a lot. Thanks for checking it out! 🙏
Nice loop, especially the review-and-edit step. I'm not an invoice person, but I built a similar extraction flow (screenshots and pasted text into structured records), and a few lessons might carry over:
Tag where each field came from. I mark every value as read by the model, typed or edited by the user, or calculated by code. Once you have that, you can see which fields people correct most. That number tells you where extraction is weak, and accuracy on a test set won't show it.
Give the model today's date and the user's locale. Mine read "8th of July" and saved the wrong year, because nothing told it what year it was. That's the same silent failure as the 1.000,00 vs 1,000.00 problem raised above. The values look fine and are wrong.
Measure the cost per extraction before you set the quota. Mine came out around two cents, and that told me what a free tier could afford. A monthly pool is easy to size once you know that number.
Are you tracking how often people edit a field after extraction?
The column chooser is the part I’d protect. Once someone renames and reorders fields, the next invoice from that same vendor should land in those columns without them rebuilding the layout. I’d also flag fields the extractor had to guess, especially totals versus line items, so a bad read doesn’t get pasted into a monthly workbook unnoticed. Five free invoices a day is enough to prove the loop; the habit sticks when the second upload feels like a saved template, not a new setup.
You know, yeah, that's exactly where it breaks, at scale. The hardest part of pulling data out of invoices is that every vendor has their own format and field layout, scan quality varies a lot, and sometimes there are handwritten corrections on top. Would be curious to ask directly: how does TidyLedger handle that inconsistency across vendor formats - that's usually where these tools start breaking once you scale up.
We've run into a bunch of these edge cases extracting data from documents at scale — wrote some of it up here: https://data-ox.com/resources/blog/documets-scraping-parsing/
The invoice-to-spreadsheet grind is the right problem. Upload, fix the columns, download Excel is a clean first loop, and the 5-invoices-a-day free plan is a fair way to try it — though a monthly pile of receipts is how this work actually shows up.
Once those rows are in a sheet, the hole I kept hitting was turning them into a monthly P&L and a quarterly tax set-aside. Free lite for that habit is at https://solopreneurpl.com. If you want the one-file version that rolls monthly into quarterly plus owner pay, Pro is $19: https://buy.polar.sh/polar_cl_FNp5fjRfVQQkg1Xr5qXLGeus9hrmScjVMTwZG2mcHbr
When someone exports your Excel, do you expect them to stop there, or do you want a profit and tax view sitting on those columns?
Thanks for the thoughtful feedback! My current focus is making TidyLedger great at turning invoices into clean, editable Excel reports that fit users' existing workflows.
I see the value in building on those exported columns to generate monthly P&L and tax summaries. That's an interesting direction, but first I want to understand what users actually need and make sure the core workflow saves them time.
The bigger question is whether users want TidyLedger to stop at clean Excel exports or become part of their broader financial reporting workflow. Appreciate you sharing your perspective!
yes...! this post is exactly what i need! it helps me gain a deeper insight into relevant aspects.
Glad to hear that! I’m happy the post was useful and gave you some deeper insights. Thanks for taking the time to read and share your thoughts!
I'll try the free plan this week with my supplier receipts. Good luck!
Thank you so much for sharing this! The home bakery use case is exactly the kind of real-world problem I want TidyLedger to solve.
Your point about taking a photo with your phone and getting a ready-to-use report is especially helpful. Automatic categorization for ingredients, packaging, and other expenses, along with monthly totals, makes a lot of sense for small businesses preparing for tax time.
I'd love to hear how your experience goes with the free plan this week. If you’re comfortable sharing, I'd be particularly interested in whether the extracted data and monthly totals match what your tax person needs.
Thanks again for giving TidyLedger a try!
Solid use case. Your lesson (one specific use case, more user conversations) is the right one. Bookkeepers and small agencies processing supplier invoices monthly might be the sharpest first niche. Asking new users what they did before TidyLedger would tell you which workflow to nail.
Thanks for the thoughtful advice! I agree that focusing on bookkeepers and small agencies processing supplier invoices could be a good starting point for TidyLedger.
I really like your suggestion about asking new users what they did before using TidyLedger. Understanding their existing workflow will help me identify where the biggest time savings are and what I should improve first.
My next step is to talk to potential users, learn how they currently handle invoice data, and use that feedback to make TidyLedger more useful.
Appreciate the guidance! 🙌
Congrats on the launch! Invoice processing and manual data entry are definitely major pain points for many small businesses and finance teams.
A few quick thoughts:
Problem Relevance: Absolutely—anyone handling recurring client billing or receipts struggles with manual copy-pasting into spreadsheets.
Improvement / Feature Suggestion: Bulk upload support (uploading a batch of 10–20 PDFs at once) and auto-detecting key fields (Vendor, Date, Total Amount, Tax) without manual column mapping would save even more time.
What makes it sticky: An export integration directly into Google Sheets or accounting software (like QuickBooks/Xero) would make it an essential daily tool.
Building in a similar document-processing space myself (ClauseScan.AI), so I really appreciate smooth data-extraction tools. Wishing you great traction! 🚀
the extraction part is table stakes now, what decides whether people stay is the correction loop. when it reads a total wrong, how many clicks to fix it, and does the fix teach the next invoice or is it a one off edit.
also watch the free plan shape: 5 invoices a day sounds fine until you meet the actual use case, which is someone doing 200 in one sitting at tax time. batch jobs and daily quotas dont mix. a monthly pool fits how this work actually arrives.
This is really useful feedback. I agree that extraction alone isn't enough — the correction workflow could be much more important. I’m thinking about making corrections quick and, where possible, reusable for future invoices from the same supplier.
The batch-processing point is also something I hadn’t considered deeply enough. A daily limit may not match the way accounting teams actually process invoices, especially during month-end or tax time. A monthly invoice pool could make more sense.
Thanks for pointing out both of these. I’ll definitely look at the correction loop and usage limits more closely.
reusable per supplier is the moat in this category. extraction is a commodity, but "remembers how your suppliers format things" is a reason to stay. on limits: accounting work arrives in piles, so whatever the quota shape, it should survive a month-end pile.
The 5-invoices-per-day cap is worth rethinking. A_Bartholomew's point about the monthly pile is exactly right — the actual use pattern is "dump 40 invoices on the 3rd," not "trickle one per day." Reframing the free tier as 50/month instead of 5/day costs you nothing but matches how bookkeepers actually work.
One thing nobody's mentioned: date and number format parsing across locales. We run our site in seven languages and learned the hard way that 1.000,00 vs 1,000.00 and DD/MM vs MM/DD break extraction silently — the numbers look valid, just wrong. A locale flag on upload that sets parsing rules would catch those errors before they reach the spreadsheet.
The Render cold start is worth fixing first though. Sixty seconds of loading on a first visit means most people never see the product at all.
This is incredibly helpful feedback — thank you!
You're absolutely right about the usage pattern. Processing 40 invoices at once is much more realistic for someone doing their monthly bookkeeping than being limited to 5 invoices per day. A monthly limit could make the free plan much more useful, so I'll seriously consider that.
The locale parsing issue is also a great catch. Different number and date formats can create errors that aren't immediately obvious, which is especially important when the output is used for accounting and tax reporting.
And I agree about the Render cold start. Getting users to the product quickly is the first priority, because none of the other features matter if people leave before they can try it.
Really appreciate you sharing lessons from your own experience. This gives me some concrete things to prioritize!
Yes, this is a real monthly job for me, not a daily one. After the fields are extracted, the time goes to duplicates (the same invoice forwarded twice, or a copy PDF with a different filename) and to mixed currencies sitting in one amount column, which makes the spreadsheet total a lie. A check on supplier plus invoice number before a row is added, and separate amount and currency columns that do not sum across currencies, would change whether I used this every month. The 5 invoices a day cap also fights that workflow, since the real use is a pile of invoices at month end.
This is incredibly valuable feedback. Thank you for explaining the real monthly workflow!
You're absolutely right — extracting the fields is only part of the problem. Duplicate invoices and mixed currencies can make the final spreadsheet unreliable, even when the extracted data looks correct.
Checking supplier name + invoice number before adding a row is a practical way to catch duplicates. And keeping the amount and currency in separate columns, without adding different currencies into one total, is essential for trustworthy reports.
Your point about the 5-invoices-per-day limit is also well taken. Monthly batch processing makes much more sense for this workflow, so a monthly usage limit is something I'll seriously consider.
This gives me a much clearer picture of what would make TidyLedger genuinely useful for recurring bookkeeping. Thanks again for helping me identify these priorities!
This is great work — I'm a CS student also considering this kind of automation for small businesses. If you were starting from zero today, what would you focus on first, and what was the hardest part for you?
Thanks! If I were starting from zero today, I’d spend more time talking to potential users before building. The hardest part was narrowing the idea down to one specific workflow and making the output genuinely useful, rather than just technically impressive. That’s probably the biggest lesson I’ve learned so far.
Taking your three questions in order:
(1) The pain usually isn't typing the fields, it's the exceptions: the invoice with two VAT rates, the credit note that should be negative, the supplier who changes their layout. The tool that wins here is the one that makes those obvious.
(2) What I'd change: a confidence flag per field and a filter for "rows that need a look", so reviewing 50 invoices means checking 6 cells, not 50 rows. And flag any row where net + tax doesn't equal the total. That one check catches a lot of extraction errors for free.
(3) What would make it a habit: a saved column template, so every month's export matches the accountant's sheet exactly, and a forwarding email address, so invoices land in TidyLedger without downloading and re-uploading them.
One small thing: if it's on Render's free tier, the first visit after a quiet period can take close to a minute to wake up, which looks like a broken site. Worth a paid instance or a "waking up" message before you share the link widely.
The editable columns make this more than invoice OCR: the useful output is a report that fits someone's existing spreadsheet. I'd ask the first few users for one real export they already prepare each month, then see whether TidyLedger can reproduce it without manual cleanup. What report are they trying to get from those invoices today?
That's a great point. I want TidyLedger to do more than extract invoice data — the goal is to turn messy invoices into clean, structured Excel reports that fit into existing workflows. I'll focus on learning what reports early users actually prepare and how much manual cleanup they still need to do. That feedback will help shape the next steps for TidyLedger.
The problem is definitely real, but I think the biggest challenge will be making the value obvious beyond “extract invoices.” I’d be curious whether users care more about saving time, reducing manual errors, or fitting the output into their existing accounting workflow. That distinction could really shape the positioning.
That's a great point. I think the real value goes beyond extraction — it's about turning messy invoice data into clean, usable Excel reports with less manual work. I'm still learning which benefit matters most to users: saving time, reducing errors, or fitting into their existing accounting workflow. I'll make that a key question when talking to early users. Thanks for the insight!
Exactly. I think talking to a few real users will make that pretty clear. I’d also pay attention to the words they naturally use to describe the problem — those can be gold for your positioning and landing page.
Have early users shown that TidyLedger replaces a recurring spreadsheet workflow, or are people mainly testing it on a few invoices to see how accurate the extraction is?
That's exactly what I'm trying to validate right now. So far, I'm still learning whether users will make TidyLedger part of their recurring spreadsheet workflow or mainly use it to test extraction accuracy. My next step is to understand how often they process invoices, what they currently do with the extracted data, and whether TidyLedger saves them enough time to use it regularly.
Could be useful to compare notes as you see whether it becomes a recurring workflow — happy to continue by email sometime.
This is great work — reminds me of some of the calls I've had to make building Xstream4K. What would you do differently if you started over?
Thanks! I really appreciate that.
If I were starting over, I’d spend more time talking to potential users before building and focus on one specific use case first. I probably built more features than I needed before getting enough real-world feedback.
Now I’m trying to learn from actual users and improve TidyLedger based on their workflow and pain points.
What about Xstream4K? What’s one decision you’d make differently if you started over?