Forecast of a new item from analogs
A new model has no history, so its sales are estimated from your past analogs, adjusted for price, store reach and seasonal level. Gives expected weekly sales, a range from the spread of the analogs and pieces over the horizon.
At a glance
- What question does it answer?
- How much we will sell of a new item and in what range from analogs.
- What data is needed?
- The price and reach of the new item, elasticity, seasonal level, horizon.
- How to read the result?
- Three analogs (22, 18 and 30 pieces a week) are adjusted to a new item at a price of 4,800 and a reach of 30 points with an elasticity of 1.2. Expected is 24.0 pieces a week, an 80% range from 20.0 to 28.0. Over 12 weeks that is 288 pieces (240 to 337). Example on the default values: Expected weekly sales — 24 pcs; Weekly range: lower bound (80%) — 20 pcs; Weekly range: upper bound (80%) — 28 pcs; Pieces over the horizon: expected — 288.
Why and how to reason
A new model has no history, but similar models do. The idea: take the weekly sales of analogs and bring them to the conditions of the new item by three adjustments: price, reach and season. Price: demand falls as price rises by elasticity, adjustment (new price ÷ analog price) to the power −1.2. Reach: more stores, more sales, the adjustment is proportional to the number of stores. Season: a multiplier of the season level. Trench: 22 × (4,800 ÷ 5,200)^−1.2 × (30 ÷ 30) = 24.2. Jacket: 18 × 0.93 × 1.25 = 20.8. Coat: 30 × 1.20 × 0.75 = 27.1. The mean of the three analogs is 24.0 units a week; the spread of analogs gives a range of 20.0–28.0 (80%), and over a 12-week horizon about 288 units (240–337). This is not an exact forecast but a sensible guide for a first purchase.
When to use it
Use it when buying a new model or capsule with 2–5 close analogs by type, price and audience. Assumptions: analogs are honestly chosen and elasticity is known at least roughly. It does not fit a fundamentally new category (no analogs) nor fashion hits with viral demand. If the price of the new item differs from an analog by more than 30% the tool warns: the price adjustment is inaccurate. With one analog no range is calculated.
How to read the result
For a first purchase take the lower end of the range (20 a week, about 240 over 12 weeks) and reorder on actual sales. A wide range means the analogs differ: reconsider their set. Check the result against reality: in which stores the item will be placed and in which weeks. If availability for sales matters, look at the upper end but take care not to create leftovers. Decision: place the order in two steps and check sales in the first 2 weeks. If the spread of analogs is large, add more analogs or refine their choice: the closer an analog by type, price and audience, the narrower the range and the safer the first purchase.
Common mistakes
Picking hits as analogs instead of typical models; ignoring that analogs sold under different display; entering elasticity with the wrong sign; ignoring season. Do not trust the mean if there are two analogs: the range cannot be estimated. Do not include as analogs items sold on promotion or in another season without adjustment.
What data is needed?
- The price and reach of the new item, elasticity, seasonal level, horizon.
- For each analog: weekly sales, price and reach.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| New item price excl. VAT | ₽ | 4,800 | Price excl. VAT: the general rate is 22% from 01.01.2026 (17.12.2025). The rate follows the shipment date; confirm your regime and rate with an accountant. |
| New item reach: stores or points | points | 30 | Where the model will be on sale. For an online store set a notional reach in the same units as for the analogs. |
| Price elasticity (absolute) | × | 1.2 | By how many percent demand falls when the price rises by 1%. An example: replace with an estimate from your data; 0: price has no effect. Promotion elasticity is not the elasticity after a permanent price rise: a temporary discount overstates it in absolute value, take a more cautious figure. |
| Seasonal level of the new item to the analogs | × | 1 | 1: the same seasonal level the analogs had in their window; 1.2: 20% higher. Empty: 1. |
| Forecast horizon | weeks | 12 | How many weeks of sales to count. |
How it is calculated
| Step | Formula |
|---|---|
| Adjustments | analogue sales × (new price ÷ analogue price)^(−elasticity) × (new reach ÷ analogue reach) × seasonal level |
| Forecast | average of the adjusted analogue values |
| Range (80%) | forecast ± 1.28 × spread (sample standard deviation) of the adjusted values |
| Over the horizon | weekly × number of weeks |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Expected weekly sales
- 24 pcs
- Weekly range: lower bound (80%)
- 20 pcs
- Weekly range: upper bound (80%)
- 28 pcs
- Pieces over the horizon: expected
- 288
- Pieces over the horizon: from and to
- 240 – 337
- Analogues in the calculation
- 3
Analogues and adjustments
| Analogue | Weekly sales | Price factor | Reach factor | Adjusted to the new item |
|---|---|---|---|---|
| Last season trench | 22 | 1.1 | 1 | 24.2 |
| Basic jacket | 18 | 0.93 | 1.25 | 20.8 |
| Light coat | 30 | 1.2 | 0.75 | 27.1 |
An analogue forecast gives an order of magnitude, not an exact number. Use analogues only from your own sales; refine the forecast with the first weeks of sales.
Nuances and limits
- Take analogs only from your own sales and from the same category and season.
- Elasticity is your assumption: check it against data and, with a price difference over 30%, take other analogs.
- A range from the spread of three analogs is a guide, not a statistical interval.