Published · The automation team at Gentle Reminder
Fixed-Format vs Open-Ended Document Work: Where Rules-Based Extraction Fits
Rules-based extraction works best for a business with serious, recurring volumes of the same kinds of documents, like repeat vendor invoices and the bank statement formats your clients use. It replaces repetitive work, not random one-off documents from random senders, which a person handles better.
Two kinds of document work
Document work is not one thing. Two situations look similar from the outside and need different approaches.
Fixed-format work. The same vendor sends an invoice in the same layout every month. A client's bank sends statements in the same structure every time. The values you need are in the same places, or follow the same patterns, each run.
Open-ended work. Any document may arrive, from anyone, in any layout. Next week's batch might include layouts nobody has seen before, and each one has to be handled as it comes.
Most accounting offices do some of each. The useful step is to find out how much of your volume is which.
Why rules suit fixed formats
For recurring layouts, a script with fixed rules is a natural fit. The rule says where to find the invoice number, how to read the line items, how the tax is shown and where the total sits. Once it is written and tested against real samples, it does the same thing every run.
That gives you some practical things:
- Consistent output. The same input gives the same result, so you can compare week to week.
- Checks that attach cleanly. Because the fields are known, you can set line math, tax math and GSTIN checks against them.
- Plain failures. If a field cannot be found, the script says which field on which document, instead of passing a guess along.
This is how we work: a script per workflow, built around your layouts, set up once and run every week on your own computer.
Who this is for
This approach works best for a business that handles serious, recurring volumes of the same kinds of documents. Think of a firm that processes invoices from the same set of vendors every month, or statements from the same set of banks for the same set of clients.
What we are replacing is repetitive work: the same fields, copied from the same places, week after week. We are not replacing truly random work, where each document is different and someone has to read it, guess what it is and decide what to do on the spot. That is a judgement call, and it is better done by a person.
What about bank statements?
Bank statements are a good fit. They are structured documents, and each bank's layout is consistent from one statement to the next, in PDF or Excel.
That means a script can be built for the fixed set of formats your firm actually receives, and for a fixed set the cost can be kept low. If your clients use a defined set of banks and layouts, building that set is realistic.
We do not claim to cover every bank. We build for the formats you use. If you need a tool that handles any statement from any bank without being told about it first, that is open-ended work and a different conversation.
Where open-ended work is harder
If most of your documents are one-offs from random vendors, such as a different kind of paper every day, then a script per layout is not efficient. Writing rules for something you will see once costs more than reading it by hand.
Truly random work, where each document needs to be looked at and decided on the spot, is better done by a person. We would rather tell you that than build something that does not pay off. Rules can still help around the edges, for instance by running checks on whatever a person or another tool produces.
A quick way to tell which you have
Take a month of documents and sort them:
- Group the invoices by vendor.
- Within each vendor, check whether the layout is the same from month to month.
- Group the bank statements by bank and format.
- Count the distinct layouts.
You will often find that the number is smaller than expected. A handful of vendors may make up most of the volume, and a few bank formats may cover most of the statements. The long tail is where most of the one-offs live.
Decide what to do with the long tail
Even in a mostly fixed-format office, a few documents will not match any rule. There are sensible options:
- Build rules for the formats that make up most of your volume first.
- Let unmatched documents be flagged by name, so a person handles them.
- Add a format later when it starts to recur.
That keeps the setup honest. The script handles what it was built for and tells you plainly when something else turns up. It does not stretch to cover documents it was not built for.
How formats change over time
Formats are fixed for a while, not forever. A vendor redesigns an invoice, or a bank updates a statement. We cover what happens then in our post on what happens when a document format changes.
Questions to ask any vendor
- How many distinct formats do I really handle, and which make up most of my volume?
- Does your tool work from fixed rules or from a general reading of any document?
- What happens when a document does not match a known format?
- Which bank statement formats can you handle, specifically?
- What does adding one more format involve?
Rules-based work is one way to automate PDF data extraction when the same vendors send the same layouts.
If you can list the vendors and banks that make up most of your volume, you can show us your workflow and we will tell you which formats look straightforward and which do not. You can list the documents, the monthly volume and a sample in the document automation assessment.
FAQ
Does rules-based extraction work for bank statements?
Yes. Bank statements are structured, so a script can be built for the fixed set of formats your firm actually receives, at low cost. We do not claim to cover every bank.
What is the difference between fixed-format and open-ended document work?
Fixed-format work involves recurring layouts that arrive repeatedly, and suits a business with serious recurring volumes. Open-ended work involves random documents that differ every time, and is better done by a person.
How do I count how many formats I handle?
Group a month of documents by vendor or bank and check whether each group's layout stays the same. The number of distinct layouts is your count.
Mostly the same vendors and banks each month?
List the formats that make up most of your volume. We will say which look like a fit for rules-based extraction.
Show us your workflow