Implementation

Ecommerce Demand Forecasting and Reorder Planning

Convert demand history, supplier lead time, and safety stock into controlled purchase decisions with a transparent ecommerce reorder planning method.

Physical demand bars with a reorder threshold, transparent stock buffer, and parcel at a lead-time gate
Physical demand bars with a reorder threshold, transparent stock buffer, and parcel at a lead-time gate

Purchase decisions fail when demand history, supplier lead time, and safety stock live in separate heads. One planner trusts last week sales. Another pads every order. Finance sees cash tied in slow movers while support fields stockout tickets on the items that actually sell. A forecast number alone does not fix that. A controlled method that turns history into an approved purchase quantity does.

This article is a transparent planning method for ecommerce reorder work. It is not a forecasting software list. If you need tool comparisons later, keep them secondary to the math and the approval path. Category context lives in AI inventory management and demand forecasting. Tool shortlists belong in separate buying guides such as AI inventory forecasting tools for Shopify. Here the outcome is controlled purchase decisions you can defend in a weekly ops review.

Decision summary

Treat replenishment as a decision system with named inputs, named owners, and an approval gate before money leaves the account.

Use three external anchors while you plan:

Decision rule for the next buying cycle:

  1. Clean demand history before you average it. Remove stockout days, one time wholesale spikes, and canceled orders that never shipped.
  2. Hold supplier lead time as a controlled input with an owner, not as a memory in chat.
  3. Set safety stock from service level intent and demand variance, then write the formula on the sheet.
  4. Convert baseline demand, lead time, and safety stock into a reorder point and a suggested order quantity.
  5. Route exceptions through approval before the purchase order becomes a commitment.
DecisionChoose this whenAvoid this when
Moving average baselineDemand is stable and the series has enough clean weeksPromotions, launches, or stockouts dominate the window
Seasonal index added to baselineYou have at least one comparable prior season for the SKUYou invent seasonality from a single spike
Higher safety stockStockout cost is clear and lead time is long or unstableCash is tight and the item is slow with cheap recovery
Manual override with expiryA known promo, delay, or launch breaks the baselineOverrides become permanent and erase the method

Scope and definitions

Demand forecasting and reorder planning here means the operating practice that converts cleaned demand history, supplier lead time, and safety stock into a purchase recommendation a human can approve or reject. It covers input controls, baseline method, seasonality and promotion handling, reorder point math, exception workflow, and review cadence.

Out of scope: vendor scorecards, ERP selection, warehouse layout, and pricing strategy. Those topics matter after the buying rule is clear.

Definitions used in this article:

  • Demand history: units sold or consumed over a defined window after cleaning for false signals.
  • Lead time: calendar days from purchase order send to sellable receipt at the planning location.
  • Safety stock: buffer units held against demand and lead time uncertainty.
  • Reorder point: on hand plus inbound threshold that triggers a purchase review.
  • Suggested order quantity: units to buy so coverage reaches a target horizon after the receipt lands.
  • Stockout: an open offer with zero sellable units when demand still exists. Use the shared stockout glossary definition when teams argue about partial location failures.
  • Inventory turnover: how often inventory cycles through sales in a period. See inventory turnover when finance challenges coverage targets.
  • SKU: the sellable stock keeping unit you plan. Use the shared SKU glossary definition when teams mix parents and variants.

This method assumes you can read inventory movement and that purchase orders are the system of record for inbound stock. Shopify inventory reports and Shopify purchase orders documentation define those platform surfaces for Shopify heavy teams. Other platforms still need the same three inputs and the same approval logic.

Forecast input controls

A forecast is only as honest as the inputs. Control five fields before you compute anything.

  1. Planning unit: plan at the sellable SKU, not the parent product, when variants buy and sell separately.
  2. Demand window: pick a window long enough to damp noise and short enough to reflect current assortment. Document the window length next to the owner.
  3. Clean demand: exclude days at zero sellable stock when the offer was live, exclude canceled orders, and flag wholesale or marketplace outliers.
  4. Lead time: store expected days and recent actual days per supplier and location. Update after every late receipt.
  5. Open inbound: subtract committed purchase orders and in transit units so you do not double buy.

Owners matter as much as fields. Merchandising owns promotion calendars. Ops owns lead time and receipts. Finance owns cash caps and coverage targets. Planning owns the formula sheet. If one person owns all four without a review, the sheet becomes private judgment dressed as math.

Pull inventory positions and movement from a report the team can recheck. For Shopify, inventory reports give stock and movement views you can reconcile against the plan. Do not paste a weekly sales export into a forecast and call the inputs controlled. The export is a raw feed. The control sheet is the cleaned series plus lead time plus open inbound.

Automation can help collect these inputs, especially when purchase orders, receipts, and sales sit in different tools. Keep collection automation under the same ownership rules you would use for any ecommerce operations automation work. Automating a dirty input only scales the error.

Baseline demand method

Start with a baseline that a skeptical planner can recompute by hand.

For stable SKUs, use a simple moving average of clean daily or weekly demand. The NIST engineering statistics handbook on process forecasting describes moving averages and related process forecasting ideas for series where recent history is a fair guide to the near future. Use that guidance as method context. Do not treat it as a promise that average demand will hold through a launch or a viral spike.

Illustrative example only: a SKU sells 70 clean units across 10 weeks with no stockout days in the window. Average weekly demand is 7 units. Average daily demand on a 7 day retail week is 1 unit. If your store sells only five days each week, state that calendar on the sheet so lead time coverage stays consistent.

Steps for the baseline:

  1. Export demand for the chosen window at SKU level.
  2. Mark and remove contaminated days.
  3. Compute average demand per day or per week.
  4. Record the sample size and the count of removed days.
  5. Freeze the baseline until the next scheduled refresh.

Reject baselines built on contaminated history. A stockout week that shows zero sales is not true zero demand. A flash sale week is not normal demand. If more than a third of the window is contaminated, widen the window, switch to a comparable prior period, or mark the SKU as judgment only until history recovers.

Trend needs humility. A four week rise can be noise, a promo echo, or real growth. Require a second signal before you raise the baseline permanently: sustained weekly gains after promo removal, consistent sell through across channels, or a merchandising decision that expands distribution. Until then, handle upside with temporary coverage or a time boxed override.

Seasonality and promotions

Seasonality and promotions break naive averages. Separate them from the baseline instead of hoping the average absorbs them.

Seasonality rule:

  • Build a seasonal index only when you have comparable history for the same SKU or a tight analog group.
  • Apply the index to the baseline for the upcoming period.
  • Retire the index when the season ends so it does not linger into quiet months.

Promotion rule:

  • Record planned lift as a separate add on with start date, end date, and owner.
  • Keep the baseline untouched so post promo planning returns to normal demand.
  • After the promo, compare actual lift to planned lift and store the miss for the next event.

Illustrative example only: baseline weekly demand is 7 units. A planned campaign expects a lift of 3 units per week for two weeks. Planning demand for those weeks is 10 units per week. After the campaign, planning demand returns to 7 unless the post campaign review proves a lasting shift.

Do not invent seasonality from one noisy year. Do not store promo lift inside the baseline. Do not let marketing promise lift without writing units and dates on the planning sheet. Unwritten lift becomes either a stockout during the campaign or overstock after it.

New SKUs and one season wonders need an analog method. Pick a comparable SKU with similar price, category, and channel mix. State the analog on the sheet. Replace analog demand with true demand as soon as enough clean weeks exist.

Reorder point and safety stock

Reorder point links demand to time. Safety stock links uncertainty to service intent.

A stockroom operator checks shelf inventory against a clipboard beside cartons and reusable bins.
Check shelf stock and inbound supply before opening a purchase review.

Illustrative example

Turn lead time and safety stock into a reorder trigger

Daily demand × lead time1 × 14 = 14 units
Safety stock7 units
Reorder point21 units
12 on hand + 0 inboundBelow the 21-unit reorder pointReview purchase
35 target units − 12 available23 to orderbefore supplier minimums
Reorder point signals when to review a purchase. Suggested quantity uses the separate target-coverage rule.

Safety stock is a policy choice expressed in units. Higher service intent and higher demand variance raise it. Longer and less stable lead times raise it. Cheap, fast replenishment and low stockout cost lower it. Write the policy in one sentence next to the number so finance and ops argue about intent, not vibes.

Practical controls:

  • Recalculate reorder points on a fixed cadence, not only when someone panics.
  • Refresh lead time from actual receipts, not from the original supplier quote alone.
  • Cap safety stock for slow movers so buffer does not become silent dead stock.
  • Split planning by location when transfer time is material.

When on hand plus inbound falls to or below reorder point, open a purchase review. The review creates or updates a purchase order. For Shopify teams, purchase orders are the buying document that receives inventory into the location. Keep the recommended quantity on the planning sheet and the committed quantity on the purchase order so audit trails stay clear.

If your team evaluates software later, judge tools by whether they expose these inputs and formulas. A buying guide such as Inventory Planner alternatives for ecommerce teams is useful only after this method is written down. Software that hides lead time and safety stock behind a single suggested order trains the team to stop thinking.

Exception and approval workflow

Exceptions are normal. Silent exceptions are how the method dies.

Define exception types before peak season:

  • Demand spike above a set multiple of baseline
  • Supplier delay beyond expected lead time
  • Promo or launch not yet on the calendar
  • Quality hold or returns surge that changes sellable stock
  • Cash or credit cap that blocks the suggested order quantity
  • MOQ that forces a larger buy than the formula suggests

Workflow:

  1. Planner flags the exception with type, SKU, and proposed override quantity.
  2. Owner of the broken input updates lead time, promo lift, or sellable stock when the facts changed.
  3. Approver accepts, reduces, splits, or rejects the buy against cash and service goals.
  4. Purchase order is created or revised with the approved quantity.
  5. Exception gets an expiry date. On expiry, the SKU returns to the standard formula unless renewed.

Approval thresholds keep speed without losing control. Low value routine reorders can auto route after checklist completion. High value buys, new suppliers, and first time seasonal deep buys need a second pair of eyes. Write the thresholds in currency or days of cover so the team does not renegotiate them in chat during a rush.

Refuse permanent overrides. An override without expiry becomes a second catalog of tribal rules. New hires will copy the override and never learn the baseline. Veterans will defend the override long after the promo or delay that justified it.

Implementation checklist

  1. Name the planning owner, lead time owner, promo calendar owner, and purchase approver.
  2. Choose the planning unit rule for parents versus sellable SKUs and publish it.
  3. Build the input control sheet with demand window, clean demand flags, lead time, open inbound, and safety stock policy.
  4. Connect inventory reports so on hand and movement can be rechecked without a private export trail.
  5. Compute baseline demand with a moving average on clean history and record removed days.
  6. Add seasonal indexes and promo lift as separate dated layers, not as hidden baseline edits.
  7. Calculate reorder point and suggested order quantity with the formulas above.
  8. Set exception types, approval thresholds, and expiry rules.
  9. Create or revise purchase orders only from approved quantities and receive against them.
  10. Schedule weekly exception review and monthly forecast error review with named attendees.

Failure modes

These failures repeat. Name them so the team assigns the right fix on first sight.

Contaminated baseline: Stockout days sit inside the average as true zero demand. The forecast understates need. The next cycle stockout looks like bad luck and is actually bad cleaning.

Lead time fiction: Expected lead time stays at the brochure number while actual receipts drift later. Reorder points fire too late. Expedites become normal cost.

Promo baked into baseline: A campaign week permanently raises average demand. Post campaign buys create overstock. Finance blames forecasting. Planning buried the promo in the average.

Safety stock as emotion: Buffer rises after every scare and never falls. Cash locks into slow units while inventory turnover weakens. Service intent was never written, so nobody can reverse the pad.

Inbound blindness: Open purchase orders are ignored. The team double buys. Receiving then creates a surplus that takes months to sell through.

Override forever: A temporary quantity becomes the new normal. The formula sheet is theatre. Only the override list is real.

No receipt discipline: Purchases exist in chat or spreadsheets while the purchase order system stays empty. Inventory reports and planning sheets diverge. Nobody can audit what was supposed to arrive.

Each failure maps to a missing control: cleaning, lead time ownership, promo separation, policy text, inbound visibility, expiry, or system of record. Repair the control. Do not only change one SKU quantity and declare the process healthy.

Measurement and review

Measure the method, not the comfort of the last meeting.

Track:

  • Forecast error on baseline SKUs after cleaning, reviewed monthly
  • Stockout rate on A items and time to recover
  • Days of cover on A items versus policy target
  • Overstock aging on items bought above formula without an open exception
  • Lead time actual versus expected by supplier
  • Exception count, age, and expiry compliance
  • Purchase order receipt variance: ordered versus received sellable units

Review weekly for exceptions and late receipts. Review monthly for baseline refresh, promo lift accuracy, and safety stock policy. Review quarterly for whether coverage targets still match cash and service goals.

Do not invent company results. Report your own measured error, stockout rate, and cover. If a tool vendor shows a case study, treat it as their claim, not your baseline. Illustrative math in this article exists to teach the formulas. It is not a benchmark for your catalog.

When error falls and exception expiry compliance rises, then consider heavier automation or a forecasting app. Method first. Software second. Category reading and tool guides stay useful references, not substitutes for the control sheet.

This week, publish the input control sheet for your fifty priority SKUs. Assign owners for demand cleaning, lead time, promo lift, and purchase approval. Write safety stock policy in one sentence per service tier.

Next, compute baseline demand and reorder points for those SKUs using clean history and current lead times. Route any buy that needs an override through the exception form with an expiry date. Create purchase orders only from approved quantities and receive against them so inventory reports match the plan.

When the method is live, use category pages and tool guides only to decide whether software should accelerate the same inputs. Keep this article as the planning layer those buys must obey.