The weekly buying tape: weeks of supply, clean rate of sale, safety stock and the season reforecast
The document shows how to get from the sales forecast, the stock and the orders in transit a schedule of receipts and order dates by week, how to clean the rate of sale of stock-outs and promotions, how to count safety stock from the forecast error, how to recalculate the season by the actual and how to take seasonality indices. The decision it supports: how much to receive each week, when to order and what stock to hold.
Why this document
A seasonal buying budget answers how much can be bought, but does not say which week to receive what and when to place the order. Because of this goods arrive in a lump or late, and stock either sits or runs out in the peak. The weekly tape links the forecast, the stock, the orders in transit and the lead time. Its inputs need cleaning: a rate of sale distorted by stock-outs and promotions and safety stock counted from the wrong error give wrong receipts. The document is read by the buyer, the planner and the finance director before placing the orders of the season.
At a glance
- What is this document and why is it needed?
- The document shows how to get from the sales forecast, the stock and the orders in transit a schedule of receipts and order dates by week, how to clean the rate of sale of stock-outs and promotions, how to count safety stock from the forecast error, how to recalculate the season by the actual and how to take seasonality indices. The decision it supports: how much to receive each week, when to order and what stock to hold.
- Who owns the document and how often is it rebuilt?
- Buying and planning: the planner builds the tape and recalculates the season, the buyer places the orders, the finance director compares the sums with the budget remainder; recounted once in two to four weeks.
- What mistakes are the most common?
- One target W for the whole year → it falls short in the peak and stock sits outside it → set the cover by week. Not subtracting orders in transit → the receipt is doubled and the stock is overstated → subtract them by week of receipt. Taking raw sales as the rate → stock-outs understate demand, promotions overstate → clean by in-stock days.
For whom
- Brand
- A brand orders from the factory months before sales: the lead time and prepayment make the forecast error the main risk, so count safety stock from the error and the tape by factory batches with the multiple and the minimum batch.
- Retail
- A retailer reorders from suppliers more often and over shorter terms: build the tape by the warehouse and stores, take in-stock days from store balances and subtract the orders in transit by week of receipt.
What it consists of
The receipts tape from the target cover
- The target at the end of the week equals the sum of the sales forecast of the next W weeks; beyond the forecast horizon the average of the last three weeks is used. The stock before receipt equals the opening stock plus the receipt in transit minus the week sales; the needed receipt equals the difference between the target and the stock but not below zero.
- The order week equals the receipt week minus the lead time. If it is earlier than the first week of the tape, the receipt is already too late to order: speeding up, another source or accepting the loss is needed. This is the most useful signal of the tape, so look at it first.
- On illustrative data (sales 120–150 a week, W = 3, lead time 1 week, stock 450, 100 in transit) 920 units are received over eight weeks, the first receipt is needed in week 2, so the order is in week 1, the end stock is 390 units.
- The target cover changes by season: in the peak (11.11, Black Friday, 8 March, 1 September) it is higher, outside the peak lower. The pack multiple and the minimum batch raise the receipt and the end stock above the target; this is the cost of the supplier constraints.
The clean rate of sale
- Raw weekly sales mix demand with stock-outs and promotions. The weekly rate equals sales divided by in-stock days and multiplied by seven; weeks with a promotion and with availability below the threshold are excluded, the base rate is the average of the remaining weeks.
- On six illustrative weeks the base rate is 12.0 units a week against the raw average of 11.8, the promotion uplift is 1.67, a stock of 96 units lasts 8.0 weeks, a variation of 18% gives class Y.
- Take in-stock days from the stock balances of the store or the platform warehouse, not from the item card. The XYZ class by the variation of weeks is editable and not a norm; for a seasonal item the variation is inflated by the seasonal course, so first divide sales by the seasonality index.
Safety stock from the forecast error
- The classic stock from the demand scatter answers how demand fluctuates. At a long lead time it matters more how the forecast errs. The weekly error is taken as 1.25 × WAPE × average demand and accumulates over the horizon of T weeks with the correlation of adjacent weeks: the root of T + T(T − 1)ρ.
- At demand of 100 units a week, WAPE 35%, a horizon of 6 weeks, a correlation of 0.3 and a service level of 95% safety stock is 279 units (2.8 weeks); at independent errors 176 units, at fully linked 432 units.
- The assumptions of the model (a normal error, one correlation over the horizon) are approximate: check them on your history and take WAPE for your item class. A better forecast directly cuts the stock.
The season reforecast and seasonality indices
- The weight of the actual equals the weeks divided by the sum of the weeks and the half-threshold; the coefficient equals one minus the weight plus the weight times the ratio of the actual to the plan to date. The forecast of the rest equals the plan of the rest times the coefficient, the season forecast equals the actual plus the forecast of the rest.
- At a plan of 10,000, a plan to date of 4,000, an actual of 3,000 and 8 weeks the weight is 0.57, the coefficient 0.86, the season forecast 8,142.9, which is −18.6% to the plan.
- The index of a month is the average ratio of the sales of the month to the monthly average of its year; on two illustrative years the maximum is 1.88 in December, the minimum 0.65 in February, the span 123% of the average. A span below 15% means there is no seasonality.
Where the data comes from
| Data source | Fields |
|---|---|
| Forecast and plan | the sales forecast by week, the season plan and the plan to date, sales by month over 2–3 years of one measure, the annual plan |
| Stock and orders | available stock, orders in transit by week of receipt, in-stock days by week, lead time, the multiple and the minimum batch |
| Forecast quality | WAPE for the item class on the weekly horizon, the correlation of errors of adjacent weeks, the target service level |
How to build it step by step
- Step 1: Clean the rate of sale: exclude weeks with a promotion and low availability.
- Step 2: Take seasonality indices over two or three years without an anomalous year.
- Step 3: Build the weekly forecast from the base rate, the indices and the event adjustments.
- Step 4: Set the target cover by week, higher in the peak, and the lead time by channel.
- Step 5: Count safety stock from the forecast error and add it to the target.
- Step 6: Build the receipts tape and find the receipts that are already too late to order.
- Step 7: Compare the sums of receipts with the budget remainder and send the orders with the weeks of receipt.
- Step 8: Once in two to four weeks recalculate the season by the actual and update the tape.
Formulas
| Measure | Formulas | Note |
|---|---|---|
| Receipt of the week | R(t) = max(0; sum of the forecast of weeks t+1…t+W − (stock + in transit − sales)) | Then up to the pack and not below the minimum batch. |
| Safety stock | z × 1.25 × WAPE × demand × √(T + T(T − 1)ρ) | Bounds: ρ = 0 and ρ = 1. |
| The season reforecast coefficient | κ = (1 − w) + w × actual ÷ plan to date, w = weeks ÷ (weeks + k) | Season forecast = actual + plan of the rest × κ. |
Formulas in notation
- Accumulation of the error over the term
\sigma_T = \sigma \sqrt{T + T(T-1)\rho}
σ: the weekly error, T: lead time plus review, ρ: the correlation of adjacent weeks.
Worked example
The tape on illustrative data: sales 120–150 a week, cover 3 weeks, lead time 1 week, stock 450, 100 in transit
| Week | Forecast | Receipt | End stock |
|---|---|---|---|
| 1 | 120 | 0 | 430 |
| 2 | 130 | 140 | 440 |
| 3 | 140 | 140 | 440 |
| 4 | 150 | 130 | 420 |
| 5 | 150 | 120 | 390 |
| 6 | 140 | 130 | 380 |
| 7 | 130 | 130 | 380 |
| 8 | 120 | 130 | 390 |
920 units in total worth RUB 1.4 million at RUB 1,500; the first order in week 1. All numbers are illustrative.
Figures
- End stock
- Target
Data as a table
| Measure | End stock | Target |
|---|---|---|
| 1 | 430 | 420 |
| 2 | 440 | 440 |
| 3 | 440 | 440 |
| 4 | 420 | 420 |
| 5 | 390 | 390 |
| 6 | 380 | 380 |
| 7 | 380 | 380 |
| 8 | 390 | 390 |
End-of-week stock and the target. The stock stays near the target: the receipt pulls it up every week.
Data as a table
| Measure | Receipt |
|---|---|
| 1 | 0 |
| 2 | 140 |
| 3 | 140 |
| 4 | 130 |
| 5 | 120 |
| 6 | 130 |
| 7 | 130 |
| 8 | 130 |
Receipt by week. The first week has no receipt: stock and the pipeline are enough.
Specifics by category
- Apparel
- Apparel: peaks by seasons and sizes; build the tape by groups with the same lead time and split sizes by the true curve.
- Footwear
- Footwear: the size grid is stiffer, short delivery of extreme sizes costs more, the safety stock for them is higher.
- Accessories
- Accessories: less seasonality, more carried-over styles; a tape for a longer cover is enough.
- Kids
- Kids: the school season gives a sharp peak, the target cover there is noticeably higher.
- Men
- Men: a steady core range, class X by variation is more frequent.
- Women
- Women: more novelties without history, take the rate from analogs and count safety stock carefully.
Common mistakes
- One target W for the whole year → it falls short in the peak and stock sits outside it → set the cover by week.
- Not subtracting orders in transit → the receipt is doubled and the stock is overstated → subtract them by week of receipt.
- Taking raw sales as the rate → stock-outs understate demand, promotions overstate → clean by in-stock days.
- Counting safety stock from the demand scatter → at a long lead time the stock is understated → count it from the forecast error.
- Extrapolating the pace of the first week → the plan “jumps” → give the actual a weight by the number of weeks.
- Not excluding an anomalous year from the indices → the seasonal shape is distorted → exclude the whole year.
Russian specifics
- The method is managerial and not regulated; prices in the calculations are excl. VAT. Lead times and EDI documents for imports should be checked with a customs broker and an accountant.