Why Excel fails for demand planning (and when it still works)

By Elie Dufeu, CTO & Co-Founder, Metreecs. Published September 27, 2026.

Excel fails for demand planning once the number of products, variants, and locations grows past what one person can maintain by hand: the workbook can no longer hold the detail the forecast needs, refreshes too slowly, and quietly accumulates formula errors. It still works well for small catalogs, a single warehouse, and stable demand.

Every planning team starts in a spreadsheet, and for good reason. Excel is already installed, everyone knows it, and it bends to any process. The trouble arrives gradually, usually around the time the catalog doubles or a second channel opens, and by then the workbook has become the file the whole company depends on and nobody fully trusts. This guide explains why Excel fails for demand planning in structural terms, how to tell whether you have reached that point, and what the move to dedicated software actually involves.

Key Takeaways

• Field audits reviewed by researcher Raymond Panko found errors in 88% of 113 operational spreadsheets, with cell error rates between 0.4% and 6.9%, so a large planning workbook almost certainly contains mistakes nobody has found.
• An Excel worksheet holds 1,048,576 rows (Microsoft). A retailer with 6,000 variants in 35 locations needs about 10.9 million rows for one year of weekly history, roughly ten times the limit, which forces planners to aggregate and lose detail.
• Excel produces one number per cell, so it has no native way to express forecast uncertainty, which is what safety stock and reorder point calculations depend on.
• Gartner advises planning leaders to measure whether teams “reduce manual workarounds” and avoid reverting to legacy tools, which suggests spreadsheet habits often outlast software rollouts.
• Excel remains a reasonable choice below a few hundred products, with one stocking location, steady demand, and a single planner.

Can Excel be used for demand planning?

Yes. Excel can be used for demand planning when the catalog is small, demand is stable, and one person owns the file. Built-in functions such as FORECAST.ETS, moving averages, and pivot tables are enough to project next month’s volume for a few hundred products from a single warehouse, and the flexibility is hard to beat.

The limits show up in three places: the level of detail the plan needs, how often it needs refreshing, and how many people touch it. A beauty brand with 150 products sold through its own website and one wholesale account can run a clean, reliable plan in a workbook. The same brand with 900 products, shade variants, 40 retail doors, and a marketplace channel is asking the spreadsheet to do a different job.

Consider Hugo, supply planner at a regional bakery-café franchise with 22 locations. For three years he planned flour, dairy, and packaging orders from one workbook with a tab per site. When the network added eight sites in a single year and introduced a seasonal menu, the file grew to 60 tabs and took 12 minutes to recalculate. A copy-paste slip on one tab doubled the dairy order for four sites during a holiday week, and roughly €9,000 of short-life product was written off before anyone traced the cause.

If your team is at that stage, talk to our team about what a product x location forecast would look like on your data before rebuilding the workbook one more time.

Why Excel fails for demand planning: five structural reasons

Each of these comes from how the tool is built rather than from how the planner uses it, and each gets more expensive as the business grows.

1. Granularity hits a hard ceiling. Demand happens at product, variant, and location level. Excel caps each worksheet at 1,048,576 rows, so detailed weekly history for a mid-sized network does not fit, and planners aggregate to category or region to make it work.
2. Data is always a few days old. Sales, stock, and open orders arrive as exports from the ERP (enterprise resource planning), POS (point of sale), or WMS (warehouse management system). Every refresh is a manual step, so the plan reflects last week’s reality at best.
3. Formulas break silently. A dragged range that stops one row short, a hard-coded value typed over a formula, or a lookup pointing at the wrong column all produce plausible numbers. Nothing flags them.
4. One number hides the risk. A cell holds a single forecast. Real demand has a spread around it, and that spread is what determines how much safety stock each product needs.
5. The file belongs to one person. Workbooks do not handle several planners editing at once. Teams end up with versions named “final_v3_JM”, and the logic lives in the head of whoever built it.

Granularity: the math behind the ceiling

Take a home décor retailer with 1,500 products averaging four variants each (size, color, finish), sold across 35 stores and warehouses. That is 210,000 product-location combinations. One year of weekly sales history at that level is 10.9 million rows, about ten times what a single worksheet can hold, before adding forecasts, stock positions, or open purchase orders.

Planners respond sensibly: they forecast at category or store-cluster level, then split the result down using fixed ratios. The split is where accuracy disappears, because a ratio built on last year’s mix cannot see that a particular color is selling faster in coastal stores this season. Category-level forecasting produces category-level errors, and no amount of effort inside the workbook changes that.

Error propagation in large workbooks

Raymond Panko’s review of spreadsheet field audits at the University of Hawaii found errors in 88% of the 113 operational spreadsheets examined across seven studies. Where cell error rates were measured, they ranged from 0.4% to 6.9% of formula cells. In a planning workbook with tens of thousands of formulas, even the low end means dozens of wrong cells.

The practical issue for demand planning is that these errors compound. A wrong seasonality index feeds the forecast, the forecast feeds the order quantity, and the order quantity becomes a purchase order with a supplier. One mistake we repeatedly see is a lookup table that was extended for new products but not for new locations, so every newly opened store silently inherits zero demand for part of the range.

The missing uncertainty model

Alongside “how many will sell?”, a good plan needs an answer to “how wrong could that number be?” That answer drives safety stock and reorder points for each product, and it differs from item to item: a steady replenishment product and a fast-moving launch can share the same average forecast with very different risk.

Excel can calculate a standard deviation, but it has no built-in way to learn how forecast error behaves per product and location, update it weekly, and feed it into order decisions. Most spreadsheet plans therefore apply one safety stock rule across the whole range, which overstocks predictable products and understocks volatile ones.

Signs you’ve outgrown Excel for demand planning

The breaking point usually shows up as a pattern of small workarounds that together cost more than anyone budgets for.

• Recalculation takes minutes, and planners avoid touching certain tabs.
• Forecasts are made at category level and split down by fixed percentages.
• The weekly refresh takes a day or more of exporting, pasting, and checking.
• Only one person fully understands the file, and holidays are a risk.
• Stockouts and overstock happen at the same time in different locations for the same product.
• Nobody measures forecast accuracy consistently, because the history of past forecasts was overwritten.

The missing accuracy record matters more than it looks. Without a record of what was forecast and what actually sold, a team cannot calculate MAPE (mean absolute percentage error) or its volume-weighted version, WMAPE, and so cannot tell whether the plan is improving. The real cost of forecast error tends to stay invisible until it appears as markdowns, write-offs, or lost sales.

The planning directors we work with often underestimate how much of their team’s week goes into maintaining the file rather than making decisions with it. Gartner’s September 2026 research on planning automation makes a similar point, arguing that planners need to move away from being “data administrators” and toward orchestrating the plan with the rest of the business.

Nadia, replenishment manager at a pet-supplies chain with 48 stores, recognized four of these signs at once last spring. Her team forecast dog food and litter at region level, split the numbers to stores by last year’s share, and spent every Monday rebuilding the file. When the chain reviewed the previous quarter, the same 30 fast-moving products had been out of stock in city stores while suburban stores held more than eight weeks of cover. The regional totals had been close to actual sales all along, which is why the problem took a full quarter to surface.

Excel vs demand planning software

The comparison below is vendor-neutral. It describes what dedicated demand planning software generally does differently, not any one product.

Excel wins on cost and flexibility, and that advantage is real for small teams. Software wins as soon as the plan needs detail, frequency, or more than one contributor. For most retailers and distributors, that crossover sits somewhere between a few hundred and a few thousand products, depending on how many locations and channels are involved.

How to move demand planning off Excel without disrupting operations

The hard part of switching tools is keeping the business running while the new plan earns trust. Across Metreecs’ work deploying AI-powered demand forecasting with retailers and distributors, the teams that transition smoothly treat the spreadsheet as a benchmark for a few weeks instead of switching it off on day one.

1. Audit what the workbook actually does. List every calculation, override, and manual rule. Many of them encode real business knowledge (minimum display quantities, supplier constraints) that the new system must reproduce.
2. Clean the history once. Pull two to three years of sales at product and location level, and flag stockout periods so the model does not read them as low demand.
3. Run both in parallel for four to eight weeks. Compare forecast accuracy by product and location on the same periods. Keep the spreadsheet’s numbers visible so planners can see where the new forecast differs and why.
4. Move decisions over by category. Start with stable replenishment products, then seasonal and new ones.
5. Keep Excel for what it does well. Ad hoc analysis, one-off scenarios, and presentations still belong in a spreadsheet, while the weekly plan moves to the new system.

Sara, demand planner at a cosmetics distributor supplying 300 pharmacies, followed this approach last year. She ran her Excel plan and a product x pharmacy forecast side by side for six weeks across 1,100 products. The parallel run showed that her category-level split was overestimating demand for low-volume shades in rural pharmacies, and moving to the detailed forecast cut the distributor’s slow-moving stock by 17% over the following two quarters.

If you are evaluating tools, ask each vendor three things: at what level they forecast by default, how they model forecast uncertainty, and how they handle products with no sales history. The answers quickly separate real planning software from a spreadsheet with a nicer interface. The bigger shift for the team is the move from static reporting to forward-looking planning, where planners spend their week on exceptions instead of on the file.

Frequently asked questions

Is Excel good enough for demand planning? For a small catalog with one stocking location and stable demand, often yes. It becomes unreliable once you need forecasts at product, variant, and location level, weekly refreshes, or several planners working on the same plan.

What are the main limitations of Excel for inventory planning? The main limits are the row ceiling that forces aggregation, manual data refreshes, silent formula errors, the lack of a forecast uncertainty model for safety stock, and poor multi-user collaboration. Each gets worse as products and locations grow.

When should I replace Excel with demand planning software? Common triggers are recalculations that take minutes, forecasts made at category level and split by ratio, simultaneous stockouts and overstock of the same product in different locations, and a refresh process that takes more than a day each week.

How accurate is forecasting in Excel? Excel’s statistical functions can be accurate for stable, high-volume products at aggregate level. Accuracy drops for low-volume variants, seasonal items, and new products, and most spreadsheet setups do not track MAPE or WMAPE, so teams rarely know their true error.

What is the best alternative to Excel for demand planning? Dedicated demand planning software that forecasts at product x location level, connects to your ERP and POS automatically, and turns forecast ranges into safety stock and order suggestions. For some businesses, forecasting delivered as a managed service is a lighter option than a full software rollout.

Conclusion

Excel fails for demand planning once a plan needs product and location detail, frequent refreshes, and several contributors, because the spreadsheet then forces aggregation and hides errors. Measure where your workbook stands against the warning signs above, and run any replacement in parallel before trusting it.

To see what product x location forecasting would look like on your own sales history, book a working session with the Metreecs team, or explore how AI demand planning handles the detail a spreadsheet can’t.

Sources

• Raymond R. Panko, “What We Know About Spreadsheet Errors,” University of Hawaii: errors found in 88% of 113 audited operational spreadsheets; cell error rates of 0.4% to 6.9%. http://panko.shidler.hawaii.edu/SSR/Mypapers/whatknow.htm
• Gartner, “Gartner Predicts Only 5% of Organizations Will Make At Least 10% of Supply Chain Planning Decisions Autonomously by 2030,” September 24, 2026: guidance on reducing manual workarounds and shifting planners from data administration to plan orchestration. https://www.gartner.com/en/newsroom/press-releases/2026-09-24-gartner-predicts-only-5-percent-of-organizations-will-make-at-least-10-percent-of-supply-chain-planning-decisions-autonomously-by-2030
• Microsoft, “Excel specifications and limits”: 1,048,576 rows by 16,384 columns per worksheet. https://support.microsoft.com/en-us/office/excel-specifications-and-limits-1672b34d-7043-467e-8e27-269d656771c3