How AI Will Help Procurement Overcome the Historical Flaws of Spend Analytics

spend analytics devrim_pinar/Adobe Stock

Spend analytics technology is not new to procurement. Yet a significant number of practitioners report feeling overwhelmed by the amount of data they have to handle, let alone knowing how to make sense of it all in a meaningful way. How can this be when the technology to solve such problems has existed for several decades?

The answer lies in the historical flaws of the spend analytics market. Up until now, the dirty secret is that analytics technology has actually performed worse than humans at classifying spend data. But with advances in machine learning, particularly the power of deep learning, the gaps in analytics offerings are beginning to narrow.

To see how, here are the top three historical flaws in the spend analysis market — and how artificial intelligence will help procurement move beyond them.

Less Accurate than a Human

While previous-generation spend analysis solutions offered procurement organizations much needed automation, one of the fundamental flaws of these solutions was that they were also far less accurate at classifying data than the average human.

Old approaches relied on statistical classification or pattern matching to categorize spend data, for the simple reason that these techniques were the ones most readily available during the prior spend analytics era. But whereas a team of spend analysts or consultants could manually classify data to a standard of 98% accuracy, solutions that rely on statistical mapping can only offer, at best, 80% accuracy.

Why the gap? Because statistical and pattern-based techniques have inherent weaknesses that require human intervention to overcome.

While the system is capable when handing data it is familiar with, previous-generation solutions often fail to classify tail spend correctly, as well as on seldom-used suppliers, products and services that are new to their engines. They also are often stumped by new languages and geographies, an important consideration as supply chains become increasingly complex and global.

New machine learning-based platforms present a compelling solution to these weaknesses. By automating analytical model building, machine learning uses algorithms to learn from data, allowing platforms to continuously improve themselves. That means as providers have trained their algorithms on reams of spend data, the system has exponentially improved and already learned how to deal with atypical spend scenarios.

The end result: Today’s leading AI-based spend analysis platforms can cleanse and classify spend data at levels of 98% accuracy — the same as a team of human analysts.

Speed for the Modern Era 

Beyond getting your money’s worth for a solution that actually works, AI-powered spend analysis platforms help procurement groups save on an even more valuable resource: their time.

Because of their historical inaccuracy, solutions that relied on statistical mapping still required plenty of human work in addition to the automation they promised. When an analysis yielded erroneous results and bad mappings from poorly classified data, users would have to define special overrides — new mapping rules — and rerun them after every data load and classification to produce usable results.

Rather than spending valuable time training their already expensive spend analysis system — and on a regular basis in order to keep up with batched processing of data — machine learning-based solutions allow for a much faster classification process, in addition to the time saved not having to write new rules and rerun an analysis.

What’s more, even in the infrequent scenario where the modern system produces erroneous results (that final 2%), the deep learning algorithms can learn and adapt, tuning its knowledge base for self-improvement rather than relying on regular human intervention.

Bottom line: machine learning-based solutions allow for much faster and more accurate classification, freeing procurement resources from data wrangling and remediation so they can instead focus on more valuable projects.

Beyond the Basics

Gaining time to focus on adding value for stakeholders is obviously helpful, but the best-case scenario is that your spend analysis system can help you deliver additional value as well, rather than limit your insights and recommendations to basic spend analysis processes.

As we discussed in Part 1 of this series, leading procurement organizations expect to use spend analytics to enable a true supply analytics approach. This strategic effort is based on a real-time analysis of both transactional data and new information sources that broaden and enhance procurement’s analysis of supply management throughout the enterprise.

Put simply, the past approaches to spend analysis relied on regular processing of historical data and kept analytics siloed from other areas of procurement. Machine learning algorithms, however, can be applied in real time to various procurement scenarios, offering predictive insights about purchasing decisions before money is spent.

In addition, machine learning-based approaches are made for the cloud, which enables the inclusion of data beyond basic transactional sources. The inclusion of customer data from the cloud in peer benchmarking systems, for example, allows procurement to analyze not just what and how it spent but how well it spent compared with other companies. While previous solutions offered price recommendations based on historical averages, such approaches were far from accurate.

The Future is Bright

While statistical mapping and pattern matching helped procurement take spend analysis into the digital world by replacing manual analysis, the technological limitations of such approaches prevented organizations from realizing a true supply analytics strategy. The rise of artificial intelligence and machine learning in the enterprise have finally changed this, allowing spend analytics to overcome its historical flaws and take this capability into the future.

Just what does that future look like? Stay tuned for the final installment of this series, in which we explore how AI will enable a community intelligence approach to spend analysis, from competitive peer benchmarking to highly granular analysis down to the subcategory and product level.

Share on Procurious

Discuss this:

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.