April 14, 2021
The order-to-cash cycle serves as both fuel and purpose of nearly every business. If your existing order management system struggles with processing orders received from nonstandard channels, like email and fax, then Impira AutoML is your friend. Impira relieves bottlenecks and helps you build a speedway on-ramp into your OM, AR, and ERP systems.
A heavy inflow of orders and sales is always a good thing, but if they’re coming in faster than you’re able to send them through to your fulfillment and billing systems, clogs and backups can occur. Pulling up each order, going through each field to manually rekey data into your existing systems in order to generate invoices is a laborious, albeit necessary, process. You’re not the first person to yearn for a more efficient way.
Speed is great, but it’s not worth as much if you don’t have accuracy. Impira AutoML works with you to create custom machine learning (ML) models that extracts pertinent data out of your orders and sends them into whatever downstream systems you have in place. We don’t use pre-trained ML models predestined to perform a specific task — Impira AutoML learns and adapts according to you and your needs. Each collection of different files (e.g., orders) has a custom-tailored machine learning model that continuously learns from every interaction you have with it. This means you could be sending more accurate, easy-to-process data downstream to the rest of your team, ensuring that fewer and fewer errors surface at the end of the line.
Many of our readers will be familiar with optical character recognition (OCR) and many will know its strengths and weaknesses. Impira takes traditional OCR and elevates it through the power of AutoML by building in a layer of adaptable intelligence. We reduce the number of errors that typical OCR generates by allowing users to be directly involved in teaching the AutoML to actively learn from its mistakes.
Our easy-to-use interface mimics the look and feel of the ever-familiar spreadsheet, so anyone can use Impira AutoML to immediately realize productivity gains. You don’t need expertise in optical character recognition or machine learning to take advantage of what we offer. Spend less time manually rekeying order information into your ERP system, shorten your days to invoice, and increase downstream accuracy by having your data easily accessible.
If the phrase “training a machine learning model” sounds daunting, take a quick look at the steps below to understand how fast, straightforward, and easy it can be. Implementing Impira can mean saving time and money, and gaining freedom from the burden of data bottlenecks.
Getting up to top speed with Impira just takes a few steps and a few minutes.
Go to impira.com/signup to create your free account. There’s no need for any credit card or personal information to get started, just an email address. Your free account can process up to 200 file units (pages), so you can jump straight into the rest of this tutorial.
Start by creating a collection, a folder that holds and organizes your different file types. Create a new collection for each task or data extraction goal. Click the + symbol in the left-hand sidebar next to Collections.
Give each collection a unique name that describes what types of files you’re uploading, e.g., “Purchase Orders.”
Click the name of your newly created collection in the left-hand sidebar. Choose files with the same layout and common information that you’re looking to retrieve (e.g., Ship To Address, P.O Date, Total amount due).
One way to upload documents into Impira is to drag and drop your files directly onto the browser window. You also have the option to upload files via Amazon S3 or Dropbox. You can even email files directly from your inbox to Impira.
As files upload, Impira AutoML will immediately get to work to process the data within your files. As soon as it’s done, you’ll see each file in a row in a spreadsheet.
Impira AutoML takes standard optical character recognition to the next level by giving you the freedom to make the changes you need. Training this collection’s machine model to fit your needs is simple. Let’s begin by grabbing the first bit of data you want from each document.
1. Double-click on the first file to open up the mark-up interface.
2. Let’s start with identifying “Ship To Address.”
The machine learning model for this collection immediately applies what it just learned by retrieving the right addresses from the rest of the documents in this collection. Your page will begin to fill up with data.
Some cells may initially be blank, or have a colored flag displayed on the left-hand side of the cell. This flag color is a visual confidence indicator of Impira AutoML’s prediction: Green meaning confident and red meaning not confident.
Because Impira AutoML is a continuously learning system, you can boost these confidence levels by confirming correct information or correcting misidentified information. These affirmations increase Impira AutoML’s accuracy by teaching it what you want it to do. The more it learns, the more you’ll see red flags turning green, indicating high confidence.
Double-click a red-flagged cell and take a look to see if Impira AutoML identified the right information you want. Click the check mark on the right-hand side if it’s correct.
Sometimes, Impira AutoML will misidentify a piece of information (e.g., accidentally highlight the wrong address). Easily correct this by clicking the X in the right-hand sidebar and use your mouse to redraw a box around the correct address.
As you continue to confirm or correct predictions across all your collections, Impira AutoML will automatically update each of their custom machine learning models, making them more and more useful for you.
Repeat Step 4 to keep extracting different information from your documents (e.g., P.O. Date, Total Amount Due). Every time you do this, your table will grow and display the information you need to see.
You’re at the finish line. You’ve pulled out all the data you need and you’ve made all the confirmations and corrections to see all red flags turn green, indicating high confidence. Export and download this table as a CSV file that you can open in Excel or Google Sheets.
Click the Download button in the top-right corner and select All files records (CSV).