November 17, 2026 in supply chain

Automating the Hidden Work Behind Monthly Supply Chain Analytics

SHARE: PRINT ARTICLE:print this page https://doi.org/10.1287/LYTX.2026.03.23

data automation screens

Every month, before a single insight can be shared with leadership, the real work begins quietly, in spreadsheets. Vendor invoices have to be pulled from one system, category codes reconciled against one another, and duplicate line items hunted down by hand. By the time the data is clean enough to analyze, days have already passed. 

As part of vendor spend reporting, the reports are updated monthly, which means repeating the data cleaning process each month with new accounting codes, vendors, and spend. This is a familiar story for analysts across operations research and analytics functions. 

When I started as an analyst out of college, I found these tasks fun, because it was the first time I was working with real-world data, and it gave me room to explore different Excel functions and formulas. 

Gradually, I started realizing the most intellectually demanding part of the job, finding patterns, building forecasts, and advising decision-makers, often get squeezed into whatever time is left after data preparation eats its share. 

I encountered this firsthand as a data analyst supporting procurement and supply chain analytics at a large energy company. I was responsible for delivering monthly spend analytics across vendors and categories, identifying cost variance trends, and delivering recommendations that fed directly into cost optimization decisions. The analysis itself was valuable and visible. But the process behind it was not sustainable. 

The Hidden Cost of Manual Preparation

Enterprise data rarely arrives ready for analysis. In the supply chain world, procurement and spend information live across multiple systems, each with its own structure, naming conventions, and quirks. Producing a single month’s spend report means extracting data from these systems, cleaning inconsistent vendor and category labels, joining tables, and manually checking for errors that may have been introduced along the way.
 
None of this was difficult work, exactly. It was just repetitive and highly susceptible to small human errors that could quietly distort an entire report, and it did many times. A mislabeled category or a missed duplicate could shift a variance calculation just enough to send leadership chasing the wrong trend. And because the process was manual, we had to track down where the mistake was made. 

The deeper cost was in missed opportunities. Every hour spent reconciling spreadsheets was an hour not spent digging into why a variance existed, what it meant for the upcoming quarter, or how to present it in a way that would change a decision. 

Building the Pipeline

I started looking into automation tools. In school, I heard about Alteryx but never paid enough attention to what that tool did. For me, being a good analyst just meant learning mainstream technical/data tools such as Excel, SQL, R, etc. 

I looked into automation because I started making more mistakes over time and my brain did not want to pay attention to the repetitive tasks. It craved new challenges. As I learned more about Alteryx from YouTube, I took that opportunity to start building pipelines to automate some of the repetitive tasks, such as using De duplication, removing null rows, renaming fields, and joining spreadsheets. 

It was a steady effort to identify which parts of the monthly processes were repeatable and could be completely handed over to a workflow automation tool. 

We mapped out the full data journey: where each source system lived, what transformations were applied by hand each month, and where errors tended to creep in. Then we rebuilt that journey as an automated workflow, one that could pull data directly from the source systems, apply consistent cleaning and standardization rules, join datasets on the correct keys, and flag anomalies for review rather than requiring a full manual audit. 

The design choice that mattered most was treating this effort as infrastructure rather than a one-off report. Instead of automating a single spreadsheet, we automated the underlying process, so that any downstream report, whether a dashboard, a forecast model, or an ad hoc analysis, could draw from the same clean, structured dataset. This meant the benefit compounded. 

In addition, we could also build validation checks at each stage, so that data quality issues surfaced early and visibly, rather than being discovered after a dashboard had already gone out to leadership. This shifted error-catching from a manual, end-of-process scramble to an automated, built-in safeguard. 

The Result

The most immediate and measurable outcome was saved time and effort. What once took days of manual preparation each month was reduced by roughly 70%. But the more meaningful shift was in how that reclaimed time got used. 

With data preparation largely automated, I could redirect my energy toward the analytical work that actually required human judgment: building forecasting models to give leadership forward-looking visibility into spend, digging into cost variances rather than simply reporting them, and refining executive-facing dashboards so supply chain KPIs and vendor performance trends were easier to understand at a glance. 

There was also a quieter benefit: consistency. Because the same automated process ran every month, reports became more reliable and easier to compare over time. Leadership could trust that a shift in a metric reflected a real change in the business, not an artifact of a slightly different manual process that month. 

Beyond Procurement Analytics

This experience mirrors a pattern operations research and analytics professionals see across industries. As organizations increasingly look to layer forecasting, optimization, and even AI-driven tools on top of their data, they often discover that the limiting factor isn’t the sophistication of the model, but the reliability of the pipeline feeding it. An organization preparing to deploy an AI agent for self-service analytics, for example, quickly learns that the agent is only as useful as the structured, validated data it can draw from. The unglamorous work of automating data preparation turns out to be a prerequisite for the more advanced analytics work everyone wants to do. 

For analytics teams still relying on manual, spreadsheet-driven processes, the case for automation isn’t just efficiency for its own sake. It’s about reclaiming analysts’ time for the work that genuinely requires human expertise. This improves the reliability of the numbers decision-makers depend on and builds a foundation that can support increasingly sophisticated analytics as an organization matures. 

The tools available for this kind of automation vary widely, and the right choice depends on an organization’s existing systems and technical resources.

 

 

Shirin Sehgal

SHARE:

INFORMS site uses cookies to store information on your computer. Some are essential to make our site work; Others help us improve the user experience. By using this site, you consent to the placement of these cookies. Please read our Privacy Statement to learn more.