Spend Classification Series: Our Data is Really Messy! (Part 2)

Spend Classification Series – 5-part guide
Previous: Part 1 – Why Does it Really Matter?

Next: Part 3 – Building a Taxonomy People Actually Use

Spend Classification Series – 5-part guide
This Post: Part 1 – Why Care?
Next Up: Part 2 – Our Data Is Too Messy!

You’ve confirmed that spend classification is super important, but there’s a catch: your transactions sit in five siloed systems that barely acknowledge each other. The AP ledger lists “ACME Ltd.”, the P-card portal calls it “Acme Pty Ltd”, and the travel tool logs simply “ACME”. Dates, currencies, even units of measure differ. No wonder the CFO sees a sea of “Miscellaneous”.

In Part 1 we showed why accurate categorisation drives savings and compliance. In this instalment you’ll learn exactly how to integrate and cleanse spend data, standardising suppliers, normalising formats, and funneling every record into a single “crystal-clear” view ready for automated classification and analytics.

Before we can unlock those benefits, we need to solve a critical upstream problem, messy, inconsistent spend data.

What Qualifies as Spend Data in Procurement Analytics?

So, what actually qualifies as “spend data”? It goes far beyond your ERP system. Spend data includes any record that tracks an outgoing payment tied to a good or service, regardless of format, source system, or department.

Spend data is any record that shows money leaving the organisation in exchange for goods or services. That means more than just the ERP:

• ERP & Finance Systems – invoices, purchase orders, GL codes.
• P-Card Platforms – high-volume, low-value swipes for office supplies or ad-hoc buys.
• Expense & Travel Tools – flights, hotels, mileage claims, meals.
• Contract & Facilities Systems – work-orders, scheduled maintenance, utilities.
• Ad-hoc Spreadsheets and CSV Dumps – anything Finance or a business unit tracks offline.

Leaving even one feed out skews analysis; hidden tail spend can bury duplicate suppliers, missed volume breaks, or compliance risk. Partial visibility equals partial savings and patchy governance. The first step toward accurate procurement spend analysis is pulling every source, structured or unstructured, into one consolidated, cleansed dataset.

Now that we’ve defined what spend data includes, the next challenge is stitching it all together into a single source of truth.

How to Integrate Spend Data: A DIY Mapping Example

Integrating five messy systems into one clean spend analysis environment may sound overwhelming, but with a structured approach, it’s entirely doable. Here’s how to break it down into five clear steps.

Pulling five unruly systems into one clean spend-data environment can feel like trying to herd cats, but it doesn’t have to be. If you break it down into five disciplined moves, the chaos starts to flow a little smoother.

First up is source mapping. This means running a full audit of where your “money out” data actually lives. Don’t just stop at the ERP, check P-card exports, expense APIs, facilities work-order files, even those sneaky ad-hoc Excel trackers. For each source, log who owns it, how often it’s updated, and what fields it contains (plus any known data quality issues). A handy checklist can help you catch easy-to-miss sources like freight portals or SaaS auto-renewals.

Next comes consolidation. The goal here is to funnel all those feeds into a single, queryable environment, what some call a “spend cube.” You might use a cloud data lake, a BI warehouse, or a purpose-built spend analytics tool. Where you can, automate the data pulls using APIs; otherwise, set up regular CSV drops from stubborn legacy systems. Use an ETL tool like Fivetran, Matillion, or even Power Automate to tag every record with standard markers: source system, business unit, load date. This step gets everything into one place, no more switching screens to find answers.

Then there’s normalisation, standardising currencies, dates, and units. Convert all amounts into your reporting currency using transaction-date FX rates. Clean up date formats using Power Query or Python. And don’t let inconsistent units trip you up: map “LTR”, “Litres”, and “L” into a single standard so your dashboards don’t throw false variances. Getting this right makes the data analysis-ready.

Once your data’s normalised, tackle supplier harmonisation. You’ll want to merge different spellings, aliases, or subsidiaries into one canonical vendor record. In Australia, the ABN is your friend; elsewhere, use VAT IDs, DUNS numbers, or local registration codes. Fuzzy-matching logic helps here, think “Dell Inc.” and “Dell Technologies” being treated as the same supplier if the match confidence is above 90%. This step powers cleaner supplier reporting and strengthens your case when negotiating volume discounts.

Finally, don’t forget gap-spotting. Once everything’s loaded, reconcile your total spend against what Finance has in the books, quarter by quarter. If things don’t match, you might have missed a feed (like a petty-cash account or a newly acquired entity) or overlooked a messy tail of unclassified spend. Flag any records with blank supplier fields or vague descriptions for follow-up. This last step keeps your “miscellaneous” bucket from quietly creeping back in.

Prefer the Fast-Track? Let Purchasing Index Handle the Heavy Lifting

You can run the five-step roadmap in-house, but many teams decide their time is better spent acting on insights, not wrestling with file formats. Here’s the shortcut PI offers and why clients pick it.

You gain rapid spend visibility, proven procurement data cleansing solutions, and the confidence that your classification engine is fed by robust, secure pipelines, without burning procurement bandwidth on data plumbing.

Purchasing Index automates secure extraction, cleansing, enrichment, and supplier-name standardisation, so your transactions flow straight into live classification dashboards. Skip the manual grunt work and start analysing tomorrow, not next quarter.

Explore the solution and book a 15-minute walkthrough

In Part 3, “Building a Taxonomy People Actually Use”, we’ll tackle the big design choice: stick with out-of-the-box codes like UNSPSC or craft a purpose-built spend taxonomy? You’ll see how the MECE rule keeps categories sharp, why three levels is the “Goldilocks” depth, and how to secure stakeholder buy-in before the first line of data is classified.