eProcurement Toolkit

You are here: sourcing, stop 1/7

The lineStrategic sourcing

What is tail spend analytics, and how do you run a tail spend analysis?

Short answer

Tail spend analytics is the work of finding, measuring and classifying the many small purchases in your spending data, so you know what the tail holds before you act on it.

Cite this answereProcurement Toolkit, “What is tail spend analytics, and how do you run a tail spend analysis?”, updated 15 Nov 2026, eprocurementtoolkit.org/strategic-sourcing/tail-spend-analytics

Key takeaways

  • Tail spend analytics turns raw purchasing data into a measured picture of the tail: how many suppliers, how much money, how many transactions, and what was bought through which channel.
  • Extract twelve months from every channel, not only invoices: purchase orders, invoices with no order, card statements and expense claims. The tail hides most in cards and expenses.
  • Clean before you count. Merging duplicate supplier records into one parent company often moves money out of the tail, because one large supplier was split across several names.
  • Classify each line twice: what was bought (a category) and how it was bought (contract, catalogue, card, no order). Rules handle known suppliers; machine learning handles the free-text descriptions that rules miss, with a person reviewing low-confidence lines.
  • A few measures show the shape of the tail: share of suppliers, share of spend, average transaction value, one-time suppliers, spend with no order, and suppliers per category.

What is tail spend analytics?

Tail spend analytics, often called tail spend analysis, is the analysis of purchasing data to find, measure and classify tail spend: the many small, often one-off purchases outside procurement’s contracts. It answers three questions: how big the tail is, what it is made of, and through which channels it is bought. The answers decide where action pays off.

The CIPS glossary describes tail spend as the bottom 20% of an organisation’s spend, which usually accounts for 80% of its suppliers. That rule of thumb is a starting point, not a result. Only your own data tells you whether your tail is 5% of spend spread over 3,000 suppliers, or 15% spread over 800. What tail spend is, with examples and types, is explained in what is tail spend.

Analytics is the first part of the work, not the whole of it. Tail spend management covers how procurement controls the tail once it is known, and reduce tail spend gives the step-by-step plan. This guide stays on the data: what to extract, how to prepare and classify it, and what to measure. Gartner lists analytics and reporting among the mandatory features of tail spend solutions, next to spend visibility and the automation of buying steps.

Which data do you need for tail spend analytics?

You need twelve months of data from every route by which money leaves the organisation, plus the supplier records and the list of contracts. A tail built from invoices alone misses most card and expense purchases, which are often the smallest and most frequent ones.

Data sources for tail spend analytics, what each adds and the usual gap
SourceWhat it addsUsual gap
Accounts payable invoicesSupplier, amount, date, general ledger account, cost centre; the base of most analysesShort or empty line descriptions; one supplier under several vendor records
Purchase order linesWhat was ordered, quantity, unit price, requester, link to a contract or catalogueMissing for invoices that arrive with no order
Purchasing card statementsMerchant, amount, date, cardholder and a merchant category codeMerchants are not in the vendor master; no item detail unless the merchant sends line-level data
Expense claimsSmall purchases paid by employees and claimed backFree-text descriptions; often kept in a separate tool
Vendor master fileSupplier names, tax numbers, addresses, bank details, statusDuplicates, subsidiaries under different names, dormant records
Contract listWhich suppliers and items are covered by an agreementKept in a spreadsheet or a separate system; no supplier identifiers

Invoices give the size of the tail; card, expense and contract data tell you what it is made of and whether it was bought outside an agreement.

Card data needs a word of explanation. A merchant category code (MCC) is a four-digit code that identifies the type of business a merchant runs; the codes are set out in the ISO 18245 standard, and the merchant chooses its code with its bank. The MCC tells you that a purchase was made at an office supply store, but not what was bought. It is a first classification, not a full one.

How do you prepare the data for a tail spend analysis?

You prepare the data by merging supplier records into one parent company, matching card merchants to known suppliers, converting currencies and marking the channel of each line. Without this step, the same supplier counts several times and the tail looks larger and more fragmented than it is.

  1. Extract twelve months of lines from each source in the table above, with the same date range.
  2. Normalise supplier names: “ACME Ltd”, “Acme Limited” and “ACME UK” become one parent, ACME. Use the tax number or company registration number first, then the name and address.
  3. Match card merchants and expense payees to existing suppliers where they are the same company; keep the others as separate suppliers.
  4. Convert all amounts to one currency at a stated rate, and remove internal transfers, taxes paid to the state and refunds.
  5. Mark the channel of every line: purchase order against a contract, purchase order without a contract, invoice with no order, card, expense.
  6. Link each line to a contract when the supplier or item is covered by one.

The worked example below shows why the order matters. A fictional company spends €62 million a year. It counts as tail every supplier with less than €25,000 of annual spend, a rule explained in tail spend management. It runs the count once on the raw extract and once after preparation.

Worked example (fictional): the tail before and after data preparation, threshold €25,000 per supplier per year, total spend €62 million
MeasureRaw extractAfter preparation
Supplier entries (vendor records plus card merchants)2,550 (2,140 + 410)2,000 (1,630 + 370)
Suppliers in the tail2,2801,690
Suppliers above the threshold270310
Tail spend (€ million)7.95.6
Tail share of total spend12.7%9.0%

Cleaning removed 590 apparent tail suppliers and €2.3 million of apparent tail spend: most of it belonged to 40 large suppliers recorded under several names.

The figures check: 2,550 − 2,280 = 270 and 2,000 − 1,690 = 310 suppliers above the threshold; 7.9 ÷ 62 = 12.7% and 5.6 ÷ 62 = 9.0%; 2,280 − 1,690 = 590 suppliers and 7.9 − 5.6 = €2.3 million left the tail. One office supplier, for example, appeared under six names, each below €25,000, but €70,000 in total. That supplier needs a contract review, not tail treatment.

Preparation also finds dormant records. An internal audit of the City of Madison found that 9,927 of its 19,629 active vendors had no recorded transactions for over four years, and that inactive vendors could still be used for invoices and payments. A supplier count taken from the vendor master, rather than from transactions, would have counted them all.

How do you classify tail spend?

You classify tail spend on two axes: what was bought, and how it was bought. The first gives categories, such as lab consumables or local repairs. The second gives channels, such as contract, catalogue, card or invoice with no order. A tail line is only useful for decisions when it has both.

What was bought
A category from a taxonomy. It can be your own category tree or a standard such as UNSPSC, which codes products and services on four levels: segment, family, class and commodity. Keep the taxonomy two or three levels deep for the tail; nobody acts on a commodity code bought twice a year.
How it was bought
The channel and whether a contract covered it. US federal buying used a similar idea: OMB memorandum M-19-13 sorted spend into tiers, from Tier 0, unaligned spending bought in a decentralised way, up to spend managed through government-wide solutions.
Who bought it
The cost centre, site or requester. It shows whether the tail is spread across the organisation or comes from a few departments.

Rules classify the lines you can predict. A supplier that sells only one thing (a courier, a hotel, a staffing agency) gets one category for all its lines. A general ledger account such as “training” gives a first category. Card merchant codes give another. Rules are cheap and easy to check, but they fail on suppliers that sell many things, such as a distributor or an online marketplace, and on free-text lines like “parts for repair”.

Machine learning (ML) classifies the rest. A model reads the line description, supplier, amount and account, and proposes a category with a confidence score. It learns from lines that people have already classified. Gartner defines spend analytics solutions as tools for the cleansing, normalisation, enrichment and visualisation of spend data, and some products listed in that market describe themselves as AI-powered. Tail spend ML is useful because the tail has the most lines and the poorest descriptions, which is where manual work costs most.

Worked example (fictional): how the 14,800 tail lines of the company above were classified
MethodLinesShare of tail linesChecked by a person?
Rule on supplier, account or card merchant code6,20041.9%A sample of each rule, once
Model, high confidence5,90039.9%A random sample each quarter
Model, low confidence2,70018.2%Every line, or grouped by supplier
Total14,800100.0%

Rules and a model together left 2,700 lines for people to review, instead of 14,800; the reviewed lines then train the next run of the model.

The totals check: 6,200 + 5,900 + 2,700 = 14,800 lines. The shares depend on your data; a company with good order descriptions will need fewer reviews. Whatever the tool, keep a record of which method classified each line, so that a wrong category can be traced and fixed.

Which measures show the shape of the tail?

Six measures show the shape of the tail: its share of suppliers, its share of spend, the average value of a tail transaction, the share of one-time suppliers, the share of tail spend with no order, and the number of suppliers per category. Calculate them on the prepared data, with the same threshold each time.

Worked example (fictional): diagnostic measures of the tail after preparation (12 months, threshold €25,000 per supplier)
MeasureCalculationResultWhat it tells you
Tail share of suppliers1,690 tail suppliers ÷ 2,000 suppliers84.5%How much of the supplier base is in the tail
Tail share of spend€5.6 million ÷ €62 million9.0%How much money is at stake
Average tail transaction€5.6 million ÷ 14,800 linesabout €378Whether a card or catalogue would cost less to run than an order and an invoice
One-time suppliers720 suppliers paid once ÷ 1,690 tail suppliers42.6%How often people find a new supplier for each need
Tail spend with no order€3.1 million on cards, expenses and invoices with no order ÷ €5.6 million55.4%How much of the tail never passed an approval before the commitment
Fragmented categoriesCategories with 10 or more tail suppliers: 14 categories, 610 suppliers ÷ 1,69036.1% of tail suppliersWhere the same need is bought from many suppliers

In this example the tail is small in money (9%) but large in suppliers (84.5%), mostly bought with no order (55.4%), and partly concentrated in 14 categories: a channel problem first, a price problem second.

These are diagnostic measures: they describe the tail at one point in time. The indicators that track a tail spend programme over time, such as tail suppliers, new one-time suppliers and spend through channels, are set out in tail spend management. Use the same definitions for both, so that the starting point and the progress can be compared.

Which analyses find the patterns in tail spend?

Five analyses find most of the patterns: a supplier ranking curve, a category-by-supplier grid, a search for the same item at different prices, a review of repeated card purchases, and a check for split purchases below approval limits. Each one points to a kind of action, which the reduce tail spend guide then turns into a plan.

  • Supplier ranking curve: rank suppliers by spend and plot the cumulative share. The point where the curve flattens shows where your tail starts, and how long it is.
  • Category-by-supplier grid: count tail suppliers in each category and site. A category with 40 suppliers across six sites, such as safety gloves, is a candidate for one contract.
  • Same item, different prices: compare unit prices for the same description or part number across suppliers and sites. It needs order lines or line-level card data.
  • Repeated card purchases: list cardholders who pay the same merchant every month. A repeated purchase often belongs on a purchase order with a contracted supplier.
  • Split purchases: look for several payments to one supplier on the same day or week, each just below a card or approval limit. GSA SmartPay training for US federal cardholders says it is never permissible to divide purchases on purpose to stay under card limits, and EU public buyers may not split a contract to avoid the thresholds of Directive 2014/24/EU.

A public audit shows the card review at work. In 2019, the internal audit team of the University of Texas MD Anderson Cancer Center tested card activity of about $395,400 for 26 of 552 cardholders, over February and March 2019. Its analysis found repetitive purchases with ten suppliers, totalling about $114,000 in the two months, although many of those suppliers were already set up in the ERP. The auditors recommended that the card programme administrator periodically analyse card transactions to find where purchase orders would be more appropriate.

How does tail spend analytics work in higher education?

Higher education tail spend comes mostly from decentralised buying: departments, laboratories and research groups buy what they need, often by card, from suppliers chosen by the person who needs the item. The analytics are the same as elsewhere, but two dimensions matter more: who bought it, and with what money.

  • Add the department and the funding source to every line. A research grant often has its own rules on what may be bought and how it is recorded, so the same purchase may be fine on one budget and not on another.
  • Look at card data first. In a university, many lab consumables, small equipment and conference fees are paid by card, and they do not appear in the purchase order data.
  • Check for the same need across departments. Two laboratories buying the same pipette tips from different distributors is a typical fragmented category.

US universities that receive federal grants have a rule that makes this analysis useful. Under 2 CFR 200.318, recipients of federal awards (other than States, which follow their own procedures) must keep documented procurement procedures, and those procedures must avoid buying unnecessary or duplicative items. A category-by-department grid of tail spend is a practical way to show that duplicates are being looked for. The MD Anderson audit described above is one published example of the card review in a University of Texas System institution.

Which tools do you need for tail spend analytics?

A first tail spend analysis can be done in a spreadsheet or a database with twelve months of lines. Software becomes worth it when you need to refresh the analysis every quarter, classify tens of thousands of lines, or combine data from several ERPs and card programmes.

Options for tail spend analytics and when each fits
OptionFits whenLimit
Spreadsheet or desktop databaseOne-off analysis, one ERP, up to tens of thousands of linesHard to repeat each quarter; classification by hand
Business intelligence tool on a data warehouseData already in a warehouse; an analyst can maintain the modelSupplier normalisation and classification must be built
Spend analysis softwareSeveral source systems; regular refresh; machine learning classification wantedCost; the taxonomy and rules still need your input
Analytics inside procure-to-pay or tail spend toolsMost tail spend already passes through that toolSees only the spend that passes through it

Start with a spreadsheet to learn what your tail holds; choose software only when the refresh and the classification become the bottleneck.

How to choose a spend analysis tool, with tests on your own data, is set out in spend analysis software. The general method of a spend analysis, including the spend cube, is in spend analysis. A tail spend dashboard needs only a few views: the six measures above, the fragmented categories, the repeated card purchases, and new suppliers created in the last quarter.

Questions

Is tail spend analytics the same as tail spend analysis?

The two terms are used for the same work: finding and measuring the tail in purchasing data. Some writers use “analysis” for a one-off study and “analytics” for a repeated, often automated, process with dashboards. The method is the same: extract, prepare, classify and measure.

Can you do tail spend analytics in Excel?

Yes, for a first analysis. Twelve months of invoice, order and card lines from one ERP usually fit in a spreadsheet. Excel becomes hard to use when you must merge several systems, classify free-text lines in volume, or repeat the work every quarter with the same rules.

What does machine learning add to tail spend analytics?

Machine learning classifies lines that rules cannot, such as free-text descriptions from distributors and marketplaces. It proposes a category with a confidence score, and people review the low-confidence lines. It saves most time on the tail, which has the most lines and the poorest data, but it still needs a taxonomy and checked examples to learn from.

How often should tail spend analytics be refreshed?

Quarterly is a common rhythm: often enough to see new suppliers and card patterns, and rare enough for changes to show. Keep the same threshold and taxonomy at each refresh, or the figures cannot be compared.

Which threshold defines the tail in the analysis?

Your own written rule, for example every supplier below a set amount of annual spend, or every supplier after a set share of cumulative spend. The choice of rule is explained in the tail spend management guide; what matters for the analytics is that the rule stays the same from one year to the next.

New guides on the strategic sourcing line

One short email when a guide on this line is added or revised. Nothing else.

Coming soon: sign-up opens with the first email.

Seb Troz

Source-to-pay consultant · written and maintained this guide

Seb Troz is a source-to-pay consultant who implements Ivalua for buying organisations. He writes the guides on eProcurement Toolkit from that hands-on work.

Disclosure: the author works as an Ivalua consultant. Pages that compare software say so next to the comparison.