How To Evaluate An IDP Partner?

Infrrd
6 min readSep 7, 2022

--

Choosing a new technology partner is like making a move in a chess match. A wrong move from your end gives your opponent a competitive edge. While automation is inevitable in today’s world, picking the wrong partner for automating business processes can be a detrimental move for your business.

In our recent conversations, we have seen many companies choose Infrrd once they have become frustrated with their vendors after the first year of partnership. This is not about bragging but rather a true statement based on feedback from these new clients. We tried to understand why this happened and assumed it was due to a lack of effort invested in the evaluation of very critical aspects of the platform. This resulted in choosing an IDP partner that could deliver short-term benefits without the vision of how their solution would perform in the long-term.

How can companies make an informed decision when choosing an Intelligent Document Processing (IDP) partner and avoid wasted efforts while accelerating digital transformation and fostering business growth?

Let us share some insights with you.

Where can things go wrong?

Some probable causes for choosing a wrong IDP partner may include:

Choosing what you believe to be the ‘best’

Yes, you read that right! When you make a choice based on only their brand reputation, things may go wrong. For example, Ferrari may be known for building the best cars in the world but when buying a bicycle for everyday use, would you go for a Ferrari bicycle? This is why it is important to check which company or product can work best for your business. Do not make decisions with only part of the picture, spend time determining the most applicable features for meeting your specific needs. Brand names, reviews, funding, and testimonials may be limited to one area of expertise but somebody who does a great job in one area may not have similar expertise in other areas.

For example, perhaps some IDP solutions work best with invoice processing but may not be very efficient with mortgage documents, and vice versa. Or maybe some have expertise in handling extreme variations but may not be great with complex data extraction from a table or handwritten text. Find the solution best suited to meet your unique requirements.

This is a fundamental mistake many of us make. Don’t choose a vendor purely on budget. It can cost you a lot more later when you start using the product. If an IDP solution solves the problems of today and can make day-to-day operations more efficient with time, it is likely to save you more business dollars in the long run. If you find such an IDP partner, increase your budget and choose the right solution.

The world is full of sales and marketing magicians who have immense talent in presenting their products. One of the key aspects of automation products is that they initiate or start operations with a lot of force but many of them lose their momentum midstream. If an IDP solution uses AI technologies and machine learning-based algorithms, it is important to evaluate — in great detail — the quality of data processing, which is primarily based on proprietary ML algorithms, and assess how effectively the system matures as the data extraction progresses.

How to identify your IDP challenge

Before starting to evaluate IDP vendors, it is best to first identify an IDP solution that works for your document types because different solutions match different needs.

IDP extraction requirements generally fall under one of these categories depending on the document type:

These are the documents that contain a fixed format. This is primarily found in forms, payment slips or utility bills from a provider. As long as you deal with just one provider, you’re dealing with structured documents. The data in these documents have fixed locations — e.g., the date will always be located in one place, the name of the person will occupy a fixed location, etc.

These are documents that contain a fixed set of data but may not be presented in a fixed format. In some documents, the date appears in the top right corner, in another variation, it is at the center of the document. And, in yet another, it might be in the bottom left corner. Another added complication is that the same data is qualified by different names. In one variation, a field may be called ‘Purchase Order Number’, in another, a ‘PO Number’. A few others may call it “PO #”, “PO No.”, or “Order Number.” It isn’t always presented in the same way.

These documents do not have any fixed layout or fixed data points. These are free-flowing verbose documents similar to this blog post that can have any information presented anywhere or in any format.

If you are dealing with structured documents, which are template-based, an efficient OCR system may be the right choice for you. There are several vendors and solutions available for structured documents that do a pretty good job of data extraction. However, these vendors may not be effective when it comes to semi-structured or unstructured documents. If you are dealing with semi-structured or unstructured documents, you need a template-free, comprehensive, ML- and AI-based IDP platform to manage variation and provide reliable predictability with high accuracy. After you’ve decided to choose an IDP solution for your document extraction needs, it’s time for evaluation.

How to evaluate an IDP solution

As an example, let’s assume you are dealing with semi-structured or unstructured documents and deemed it necessary to choose the best IDP vendor for you. There are many vendors that make big claims to guide you to make the right decision. With a lot of information, how do you run an effective evaluation?

The following process may help:

Figure out your key complexity:

Do not try to evaluate everything but rather focus on the most complex extraction needed for your documents. For example, if you are dealing with invoices which have tabular information, you can check how efficiently and effectively the IDP solution extracts tables. If an IDP solution seamlessly extracts complex data, such as tabular information, relying on such a system is a safe bet. Other complex areas may vary based on your business, but can include handwritten text or extreme variations in your semi-structured or unstructured data.

Take a demo with the top players

There are hundreds of IDP vendors in the market. Do some research and shortlist the top three vendors, based on your bandwidth. Ask for a demo with each vendor and be sure that you ask them to show how the complex data you identified, such as tables or handwritten text, is extracted or handled. A good IDP system fit is expected to show 80% to 90% accuracy right out-of-the-box for most of the variations. Also, you may consider checking how the system handles at least seven to ten different types of variation.

Try Before Buy

After the initial evaluation, choose the best — or at the most — the top two solutions and use their trial versions if available to then conduct a proof of concept (POC) with the vendor. A POC is usually a paid service offered by IDP vendors, which primarily means training the IDP solution to match your business or extraction requirement. Ensure that all your critical and complex variations are discussed and tested thoroughly before signing the purchase contract. Be cautious of an elaborate POC. This is another mistake that most businesses make. Focus primarily on your critical requirements to keep the POC shorter, faster, and more effective.

Great start vs. great finish

During the trial, ensure that you do not fall into the “great start” trap. There are several template-based OCR solutions that solve simple problems quickly and provide the impression of a good start but they fail to offer a good finish. Evaluate the extraction capabilities of the IDP solution to get a high-level understanding of how the IDP solution is designed, how AI-based technologies are implemented, and how superior their ML algorithms are.

Finally, before signing any contract, check in with some existing customers to get references about the potential IDP vendor and talk with them and gain insights into their experience using the solution. This can be the best guide for you if you ask salient and pointed questions.

Choose the Right Partner

IDP solutions have never been more essential when processing semi-structured or unstructured data. Selecting the right partner for meeting your business process needs and addressing challenges is the most important step in initiating and ensuring a successful IDP journey. Manual oversight may cost you both time and money so it’s essential to choose the right IDP partner wisely.

Originally published at https://www.infrrd.ai.

--

--

Infrrd
Infrrd

Written by Infrrd

Infrrd has been offering AI as a Service since inception. Their focus is on developing faster Enterprise AI platform using AI, ML & NLP- https://infrrd.ai