All articles
8 min read

XYZ Analysis: Why Cut-Offs Differ and How to Set Yours

XYZ analysis sorts stock items by how steady their demand is, using the coefficient of variation. Published cut-offs differ, so test your own on twelve months of sales before trusting a class.

By Praveen Nune, Co-Founder & CEO · Updated 17 Nov 2026

A warehouse manager in India stands beside shelving and looks at three pallets holding steady, uneven and sparse amounts of stock.

Key takeaways

  • XYZ analysis ranks each item by demand steadiness: standard deviation divided by average monthly demand. Low is X, middle is Y, high is Z.
  • No cut-off is standard: published guides put the Z boundary at a CV of 0.25, 0.5 or 1.0.
  • Always run your item list through two rule sets and count how many items change class. If many do, the cut-off is doing the work, not the data.
  • CV (coefficient of variation) misleads on seasonal, intermittent and new items. Separate those before you classify the rest.
In this article
  1. 1What is XYZ analysis and how do you calculate it?
  2. 2Which CV cut-offs should you use for X, Y and Z?
  3. 3How much does the cut-off change an item's class?
  4. 4Where does CV mislead, and what do you do about it?
  5. 5How do you run XYZ analysis on your own items?
  6. 6What should you do with each class?
  7. 7How does Arka Inventory help once items are classed?
  8. 8Frequently asked questions

The same item can be an X, a Y or a Z depending on which guide you follow. A part with a coefficient of variation of 0.36 is a Y under one published rule set, an X under another and a Z under a third. The class is a product of the cut-off you chose, not a fact about the part.

That matters because XYZ analysis is used to decide how much safety stock each item gets and which items are bought only against orders. If many items slide between classes because of a threshold copied from the web, the stock policy moves with them.

The worked example below shows it with four demand series, in which the 0.36 item is a Z under the tightest rule and an X under the widest. This page gives you the calculation, a worksheet that shows how sensitive your own classes are to the cut-off, and a table of what to do with each class. All you need is a spreadsheet with twelve months of sales per item.

What is XYZ analysis and how do you calculate it?

XYZ analysis classifies inventory items by how much their demand varies. X items have steady turnover, Z items vary the most, and Y sits between (AICPA & CIMA, 2 August 2018). The measure is the coefficient of variation (CV): the standard deviation of monthly demand divided by its mean.

The standard deviation is the typical distance of a month's demand from the average. Dividing by the average makes a fast item and a slow item comparable.

Worked on a steady fastener that sold 100, 102, 98, 101, 99, 103, 97, 100, 102, 98, 101 and 99 units over twelve months:

  • Mean = 100 units a month.
  • Sample standard deviation = 1.86.
  • CV = 1.86 ÷ 100 = 0.02, or 2%.

In Excel the formula is =STDEV.S(B2:M2)/AVERAGE(B2:M2). Using STDEV.P (population standard deviation) instead lowers CV by about 3% to 5% on twelve data points. That only changes a class for an item sitting right on a cut-off. Pick one function and use it throughout.

XYZ analysis ignores what an item is worth. For value-based classes, use our ABC analysis page.

Which CV cut-offs should you use for X, Y and Z?

There is no standard set of cut-offs: published guides use three families of boundaries.

FamilyXYZGuides using it
TightCV up to 10%over 10% to 25%over 25%
Midbelow 25%25% to 50%above 50%

Kladana; Cleverence is close, with Z above 0.50

Widebelow 0.50.5 to below 1.01.0 and above

Cleverence suggests a 12-month lookback (Cleverence) and SCM Dojo 12 to 24 months (SCM Dojo). The Z boundary is 25%, 50% or 100% (1.0), so the tightest rule is a factor of four stricter than the widest.

Our recommendation: treat the cut-off as a setting you choose, not a rule you look up. Choose it by the test in the worksheet below. Aim for a Z class small enough that someone can review it item by item.

How much does the cut-off change an item's class?

In our four illustrative series, two of four items (Festive and Noisy) change class depending on the cut-off family, and only Steady and Spare part stay put. Any item whose CV falls between the tight and wide boundaries can change class, which is our assumption for many real item ranges; check yours with the worksheet below. We built four demand series (illustrative, not client data) and classified each under the three families, using sample standard deviation over twelve months.

  • Steady: 100, 102, 98, 101, 99, 103, 97, 100, 102, 98, 101, 99
  • Festive: 40, 40, 45, 60, 90, 140, 160, 120, 70, 50, 42, 40
  • Noisy: 80, 130, 60, 120, 70, 140, 50, 110, 90, 150, 55, 100
  • Spare part: 0, 0, 40, 0, 0, 0, 35, 0, 0, 0, 0, 45
ItemMean a monthCVTightMidWide
Steady1000.02XXX
Festive74.750.57ZZY
Noisy96.250.36ZYX
Spare part101.82ZZZ

Two things stand out. The noisy item, which swings month to month with no seasonal curve, is an X, a Y or a Z depending on the family. And the festive item repeats the same curve every year, so it is easy to plan, yet it still scores Z or Y.

The window matters as well. The spare part's CV is 2.45 on its first six months and 1.57 on its last six, because a single order moves a mean this small.

Where does CV mislead, and what do you do about it?

CV misleads on seasonal items, intermittent items and new items, because it measures spread, not how forecastable the spread is. Forecasting researcher Nikolaos Kourentzes puts it directly: "The coefficient of variation is a scaled version of the standard deviation of the historical sales. This tells us nothing about the easiness to forecast sales or not." A strongly seasonal item with a repeatable curve looks harder than a random one, as the festive row above shows.

Intermittent items are those with many zero months. A tiny average makes CV huge, so the spare part is a Z regardless of the family. Forecasting researchers separate these with two measures: the average gap between demands (average demand interval, ADI) and the squared coefficient of variation of the demand sizes (CV²), using cut-offs of 1.32 and 0.49. Those cut-offs were derived only to choose between two forecasting methods, not to set stock policy (Svetunkov, OpenForecast, 4 June 2025). Use them as a prompt to handle these items separately, not as stocking rules.

Our recommendation for each trap:

  • Seasonal items: with 24 months of data, compare each month with the same month a year earlier. If the gaps are small, treat the item as plannable whatever its CV.
  • Intermittent items: list items with three or more zero months out of twelve separately, and review them by order history, not CV.
  • New items: leave anything under twelve months of history unclassified.
  • Stock-out months: exclude them. Sales were zero because there was nothing to sell, not because nobody wanted it.
A shopkeeper studies a stockroom shelf holding a tall stack of festive goods, one lone spare part box and a few new sealed cartons.
Seasonal, rarely sold and brand-new items each look erratic for different reasons, which is why one CV number can mislead.

How do you run XYZ analysis on your own items?

Run it as a worksheet of eight tests, each with a clear pass or fail. Use one row per item and twelve columns of monthly demand (units shipped or consumed).

An inventory planner works through monthly sales figures on a laptop at a desk with a notebook and a cup of tea.
Working through twelve months of sales item by item is a desk job you can do yourself before trusting any class.
  1. 1
    Complete history.
    Pass: at least twelve usable months after stock-out months are removed; otherwise the item stays unclassified.
  2. 2
    Stock-outs marked.
    Pass: every month with zero stock is flagged and excluded. Fail: a stock-out month sits in the data as a zero.
  3. 3
    CV computed.

    Pass: mean, STDEV.S and CV in three columns, with a fourth counting zero months.

  4. 4
    Intermittent split.
    Pass: items with three or more zero months are listed apart. This threshold is our rule of thumb.
  5. 5
    Two rule sets.

    Classify the rest under the mid family and the wide family, for example with =IF(N2<0.25,"X",IF(N2<0.5,"Y","Z")) for the mid family, where N2 holds the CV (use 0.5 and 1.0 for the wide family). Count items whose class differs. Pass: no more than 20% of items change class between the mid and wide sets (our recommendation). Fail: look at those items by hand.

  6. 6
    Z share.
    Pass: Z is under 20% of items (our recommendation). Fail: loosen the cut-off.
  7. 7
    Seasonal check.
    For items with 24 months, compare each month with the same month last year. Pass: each month is within 20% of the same month last year (our recommendation).
  8. 8
    Re-run every quarter.
    Pass: the list of items that moved class since the last quarter is recorded.

What should you do with each class?

Give X items a lean buffer, Y items a seasonal buffer, and Z items an order-based or reviewed policy. The table is our recommendation, built from standard practice, not a fixed rule. Combine it with value classes from the ABC analysis page.

ClassTypical policyReview
XReorder point with a small safety-stock buffer; automate the reorderMonthly exception check
YReorder point plus a buffer sized for the peak; raise before the seasonBefore and after each season
Z, high valueBuy or make against orders, or hold a small agreed bufferItem by item
Z, low valueHold a cheap buffer or stop stockingQuarterly sweep
IntermittentReview order history; keep only what a stock-out would costItem by item

Safety stock sized from lead-time and demand variability is covered in our demand planning and inventory forecasting methods pages. For movement-based classes see FSN analysis.

How does Arka Inventory help once items are classed?

Arka Inventory, built natively on Salesforce, turns each class into a stock rule: reorder alerts for X items, forecast-led purchasing for Y, and a dashboard of your Z items. We are trusted by more than 30 paying B2B customers. The classification itself is a spreadsheet job, as the worksheet above shows.

Our forecasting and purchase automation links sales, stock levels and forecasts so purchases are planned against what you actually sell. Our inventory alerts flag items approaching or falling below minimum stock levels, which suits your X items. Our custom reports and dashboards let you focus on the data most relevant to your operation, so you can keep your Z list in front of the planner.

You can try all features free for 15 days, with no credit card. Onboarding typically takes a minimum of two weeks, depending on data readiness. As of November 2026, plans start at $199 a month, paid yearly (see pricing); there is no rupee price list, so contact our sales team for rupee quotes.

Talk to our team

Leave your details and we'll get back to you.

Frequently asked questions

ABC ranks items by value, so it shows where money is tied up. XYZ ranks items by how steady demand is, so it shows how hard they are to stock. Combining them gives a nine-cell grid.
Choose the cut-off by running the worksheet above, not by copying a guide. The right boundary is the one that leaves Z a small share of items that someone can review by hand.

Yes. Put twelve monthly demand figures in a row and use =STDEV.S(B2:M2)/AVERAGE(B2:M2) to get CV. Then classify with an IF formula against your chosen cut-offs. Handle zero-demand and new items separately.

Use at least twelve months, and 24 if the item is seasonal, so one full season is covered; shorter histories give unstable classes.
No. FSN groups items by movement: fast, slow or non-moving. VED groups them by criticality: vital, essential or desirable. XYZ groups them by how steady demand is.

Sources

  1. 1.Cost transformation: XYZ inventory management, AICPA & CIMA, 2 August 2018
  2. 2.ABC-XYZ analysis for forecasting, Nikolaos Kourentzes, 15 October 2016
  3. 3.SBC is not for you, Ivan Svetunkov, OpenForecast, 4 June 2025
  4. 4.ABC XYZ analysis, Interlake Mecalux
  5. 5.XYZ analysis in inventory management, Kladana
  6. 6.XYZ inventory analysis, Cleverence
  7. 7.ABC XYZ analysis guide, SCM Dojo
  8. 8.XYZ demand classification, MetricGate
  9. 9.Worked table: our own calculation on four illustrative twelve-month series, using sample standard deviation.
  10. 10.Arka Inventory forecasting and purchase automation

Keep reading

Arka Inventory.

© 2026 Arka Inventory

Powered by PageLens.ai

Get in touch — we'd love to help.

Request a Demo