Syntha Lab RU

Trends and the range forecast: from a signal to the trend share in the matrix, the life window, the test batch and the chase order

How to turn a trend from inspiration into a manageable purchasing decision: how to read a signal (rising, falling, flat), how to count the trend share in the matrix and set a ceiling, how to estimate the trend life window, release a test batch and decide whether to chase. The document answers the question of how much budget and how many SKUs to give to a trend and at what moment to add volume so as not to buy a lot at the peak and not to be left with stock when the trend has already gone.

Commerce and buying Brand and retail Verified: 2026-11-15

Open interactively →

Why this document

Trend items give a noticeable part of the interest in a collection and at the same time the riskiest stock: a trend lives for weeks while production takes months. So the decision is built not on faith in a silhouette but on three numbers: the share of the budget, the life window and the result of the test. The buyer, the category manager, the brand’s product manager and the CFO read the document when drawing up the matrix and the purchase budget. It gives an order: a signal, a hypothesis, a share ceiling, a test, a chase or a stop, and a reconciliation with the budget and the season.

At a glance

What is this document and why is it needed?
How to turn a trend from inspiration into a manageable purchasing decision: how to read a signal (rising, falling, flat), how to count the trend share in the matrix and set a ceiling, how to estimate the trend life window, release a test batch and decide whether to chase. The document answers the question of how much budget and how many SKUs to give to a trend and at what moment to add volume so as not to buy a lot at the peak and not to be left with stock when the trend has already gone.
Who owns the document and how often is it rebuilt?
The product manager and the category manager keep the hypotheses and the tests, the buyer or the planner counts the share and the chase, the CFO agrees the ceiling; reviewed every season and by the signal during the season.
What mistakes are the most common?
Brand: the whole trend batch is ordered at once at the peak of the signal → at weak demand a lot of stock remains → release a test batch and chase by the observed rate. Brand: the share ceiling is not set and trend SKUs are added during the assembly of the matrix → the trend takes more of the budget than planned → fix the ceiling by budget and SKUs before the matrix is assembled. Retailer: a trend is taken from a brand without a right to reorder and a delivery lead time → at good demand a chase is impossible → fix the reorder in the contract.

For whom

Brand
A brand develops trend items itself and produces them: it needs a ceiling for the share in the matrix, a window allowing for the production lead time, a test batch in a small run and a fast reorder from the factory.
Retail
A retailer chooses trends from the collections of brands and suppliers: it needs a hypothesis, a small trial purchase, the right to reorder from the brand and control of the trend share in the category budget and in stock.

What it consists of

A trend as a purchasing decision

  • The life of a trend passes through stages: emergence (a narrow circle notices it), growth (brands and marketplaces repeat it), peak (it is mass and the price presses), decline (stock and repeats remain). The goal is not to guess a trend but to choose a share of the budget and the moment of entry at which an error does not hit the result of the season.
  • A trend is measured in a silhouette, a color, a material, a detail and a theme; each of these levels lives at a different speed: color and detail change faster than a silhouette. For the decision write down what exactly is part of the trend so as to count trend SKUs in the same way from season to season.
  • Brand: turn a trend into a numbered hypothesis: what garment, for whom, in what window, what test batch and what test threshold; a hypothesis without a threshold remains an opinion.
  • Retailer: the buyer records a trend as a hypothesis by category and supplier: which items are taken on trial, for how many weeks and by what threshold a reorder is awaited.

Signals and their check

  • The signals are queries in search and on marketplaces, mentions in social networks, sales by keywords on marketplaces, shows and exhibitions, and also your own data: views, adds to cart, a waiting list and test sales. Every source is noisy and may be distorted by inflation, so a decision is not taken from a single source.
  • The seasonal growth of queries for a product is mistaken for a trend: in autumn the queries for jackets grow for any silhouette. Compare with the same period last year and with the category as a whole; a trend is seen as growth above the seasonal one.
  • The signal in Lab is read in two ways: by the growth rate per week (the slope of the logarithm of the index) and by comparing the average of the last weeks with the previous ones. The threshold “rising / falling” is set by your policy; below the threshold the signal is read as “flat”.
  • Brand: keep the signal series of a hypothesis by week and update it before the chase decision; if the index falls chase only at a very high test result.
  • Retailer: compare the signal with the sales by keywords on marketplaces and with the stock of competitors and suppliers: if their stock grows, the trend is saturating.

The trend share in the matrix and the ceiling

  • The share is counted on four bases: the number of SKUs, the budget, units and revenue. For the decision the two main ones are the share of SKUs and the share of the budget, because they show how much focus and money is given to the trend. If the share of the budget is above the share of SKUs, the trend has been given more depth than the rest.
  • There is no universal share, the ceiling is set by the brand policy. A practical way to choose it: draw up a table of past seasons (the trend share of the budget, sold at the full price, stock at the end of the season, margin) and find the share above which the trend dragged the result down.
  • Brand: set the ceiling separately for the budget and for the number of SKUs and fix it before the matrix is assembled; agree an excess as an exception with the CFO.
  • Retailer: count the trend share inside a category and by stores: in different formats and regions a trend lives differently, so one ceiling for the chain is too coarse.

The life window, the test batch and the chase

  • The trend life window is your estimate of the number of weeks in which it sells at the full price from the decision date. The time to the shelf is subtracted from it and the selling window after arrival remains; from it the test weeks and the reorder lead time are subtracted and the chase window remains.
  • The test batch is small so that an error costs little but sufficient to see the rate in a few weeks: sold in the test equals the rate multiplied by the weeks and divided by the size of the test batch. The chase is made if the sold share is not below the threshold, there is a window and the signal does not fall.
  • The chase volume is counted by the observed rate and the weeks of the window with fading: the rate is multiplied by the sum of a geometric series by the fade. It is the upper bound of demand at the given fade and not a guarantee of sales.
  • Brand: agree a fast reorder with the factory in advance (the lead time and the minimum lot), otherwise the chase window is zero; put the materials for the chase into the swatch and trim card as a reserve.
  • Retailer: fix in the contract with the brand the right to reorder and the delivery lead time; without them a test on a trial batch gives no way to react to demand.

The range forecast and control

  • The share of a silhouette, a color or a category in the next season’s matrix is built from the share in the sales of past seasons (in units, adjusted for availability) plus a managed adjustment by the signal. The size of the adjustment is limited by a maximum shift set by policy so that one season does not rebuild the whole matrix.
  • The quality of work with trends is checked by the share of trend SKUs that passed the test threshold, the share sold at the full price for trend and basic SKUs and the trend share in stock at the end of the season; compare seasons with each other.
  • Brand: after the season go through every hypothesis: what the signal gave, what the test showed, whether it was chased and how it ended; record the conclusions in a register of hypotheses so as not to repeat errors.
  • Retailer: compare the trend share in stock with the plan for price reductions and clearance (see the document on the season and price): trend stock loses value faster than basic stock.

Where the data comes from

Data sourceFields
Signals of a hypothesisa series of the index by week (queries, mentions, sales by keywords), the comparison base of last year
The matrix and the budgetthe number of SKUs, the budget, the trend SKUs and their budget, the SKU roles
Production and logisticsthe time from the decision to the shelf, the minimum lot, the reorder lead time, the possibility of a fast repeat
Sales and stock of past seasonsthe share sold at the full price, stock at the end of the season, the margin on trend SKUs

How to build it step by step

  1. Step 1: Brand: formulate the trend hypothesis: the garment, the audience, the window, the test batch and the test threshold.
  2. Step 2: Collect a signal series by week from several sources and compare with the same period last year to remove seasonality.
  3. Step 3: Count the growth rate of the signal and the comparison of the last weeks with the previous ones; classify the signal as “rising”, “falling” or “flat” by the threshold.
  4. Step 4: Count the share of trend SKUs and of the budget and compare with the ceiling; in case of an excess cut the number of SKUs or the depth.
  5. Step 5: Retailer: agree with the brand a trial purchase, the right to reorder and the delivery times before placing the order.
  6. Step 6: Estimate the life window and count the selling window after arrival and the chase window; if there is no chase, release only the test batch.
  7. Step 7: Release the test batch and after the test weeks count the sold share in the test; compare with the threshold.
  8. Step 8: Decide on the chase: the window above zero, the test not below the threshold, the signal not falling; count the volume and the cost of the chase and the trend share of the budget after it.
  9. Step 9: Brand: by the end of the season review the hypothesis and enter the conclusions in the register; revise the ceiling and the life window for the next season.

Formulas

MeasureFormulasNote
Trend sharesShare of SKUs = trend ÷ all; share of the budget = the trend budget ÷ the whole budgetThe ceiling is set by policy; compare both shares.
WindowsWindow after arrival = life − time to the shelf; chase window = window after arrival − test weeks − reorder lead timeThe trend life is an estimate and not a fact.
Sold in the test and the chaseSold = rate × weeks ÷ test batch (not above 100%); chase = rate × (1 − (1 − fade)^window) ÷ fadeAt a zero fade the chase equals the rate multiplied by the window.
SignalGrowth per week = exp(the slope of ln index by week) − 1The logarithm makes the rate independent of the level of the index.

Formulas in notation

Chase window
W_c = T_life − T_shelf − t_test − T_reorder
T_life: the trend life from the decision; T_shelf: the time to the shelf; t_test: the test weeks; T_reorder: the reorder lead time; if W_c ≤ 0 there is no chase.
Chase volume
Q = r · (1 − (1 − d)^W_c) / d
r: the sales rate per SKU per week; d: the weekly fade of the rate; W_c: the chase window; at d = 0 the volume equals r · W_c.
Signal rate
g = exp(β) − 1, β = Σ(t − t̄)(ln y − mean) / Σ(t − t̄)²
y: the signal index by week; β: the slope of the linear regression of the logarithm of the index on the weeks t; g: growth per week.

Worked example

An illustrative test of a trend on 10 SKUs: what happens if the real rate differs from the expected one (the expected one is 14 units a week per SKU)

Real rate, units/weekSold in the test, %Stock when the whole batch is ordered at once, units per SKUStock with a test and a chase, units per SKULost demand with a test and a chase, units per SKU
7.03584024
14.070000
21.01000030

Illustrative data: a test batch of 60 units, 3 test weeks, a chase window of 9 weeks, a fade of 4% a week, a threshold of 60%. Ordering the whole batch at once (168 units per SKU) at weak demand leaves 84 units per SKU (about RUB 1.6m at cost for 10 SKUs); the test stops the chase and limits losses at the price of a little lost demand.

Figures

The signal index by week and the growth line
Line chart: Index: 100 → 148; Growth line: 100 → 146100110120130140150W1W2W3W4W5W6W7W8W9W10Index, W1: 100Index, W2: 106Index, W3: 111Index, W4: 109Index, W5: 120Index, W6: 127Index, W7: 131Index, W8: 128Index, W9: 142Index, W10: 148Growth line, W1: 100Growth line, W2: 104Growth line, W3: 109Growth line, W4: 113Growth line, W5: 118Growth line, W6: 123Growth line, W7: 129Growth line, W8: 134Growth line, W9: 140Growth line, W10: 146rel. units
  • Index
  • Growth line
Data as a table
MeasureIndexGrowth line
W1100100
W2106104
W3111109
W4109113
W5120118
W6127123
W7131129
W8128134
W9142140
W10148146

The signal index by week and the growth line. The index grows by about 4% a week; the last four weeks are 18% above the previous four. The signal is “rising” at a threshold of 3% a week.

Stock per SKU at weak, expected and strong demand
Bar chart: Demand ×0.5 84, Demand ×1 0, Demand ×1.5 0020406080100Demand ×0.5Demand ×0.5The whole batch at once: 8484Test and chase: 00Demand ×1Demand ×1The whole batch at once: 00Test and chase: 00Demand ×1.5Demand ×1.5The whole batch at once: 00Test and chase: 00units per SKU
  • The whole batch at once
  • Test and chase
Data as a table
MeasureThe whole batch at onceTest and chase
Demand ×0.5840
Demand ×100
Demand ×1.500

Stock per SKU at weak, expected and strong demand. A test with a threshold removes the stock at weak demand; at strong demand it forgoes part of the sales and the chase by the observed rate returns it.

The trend life window: development, test, reorder and the sale of the chase. Of the 26 weeks of life 10 go on the way to the shelf, 3 on the test and 4 on the reorder; 9 weeks remain for selling the chase.

StageStart, weeksLength, weeks
Trend life from the decision026
Development and delivery to the shelf010
Test on the shelf103
Reorder134
Sale of the chase179

Specifics by category

Apparel
Apparel: a trend is more often in the silhouette, color and material; the time to the shelf is several months, so the chase window is short and the main tool is a test with a small batch and a fast repeat at a factory with a short lead time.
Footwear
Footwear: the lead time for developing a last and for production is longer, a trend passes through the last, the heel and the material; the chase window is almost zero, so the trend share is limited more strictly and the test is built on a small series and a pre-order.
Accessories
Accessories: a trend changes quickly (the shape of a bag, hardware, jewelry), the time to the shelf is shorter, the batches are small; a test and a chase work best and the stock is easier to dispose of, so the share ceiling can be set higher if the stock is easy to sell off.
Kids
Children’s: a trend is milder and demand is tied to the season, school and the size run; a child’s growth makes the stock of a size irrecoverable, so the test is built by sizes and the trend share in the basic sizes is kept lower.
Men
Men’s: trends are slower and evolve inside the basics (fit, length, material); the share of trend SKUs is usually lower, the life window longer and the signal is more reliably read by repeat purchases than by queries. The test decision here is more often taken on basic items with a small fashion detail than on separate fashion garments.
Women
Women’s: trends are faster and wider (silhouette, color, print, detail), so the risk of stock is higher too; set the share ceiling by budget and control it by weeks, and check the signal on several sources at once.

Common mistakes

  • Brand: the whole trend batch is ordered at once at the peak of the signal → at weak demand a lot of stock remains → release a test batch and chase by the observed rate.
  • Brand: the share ceiling is not set and trend SKUs are added during the assembly of the matrix → the trend takes more of the budget than planned → fix the ceiling by budget and SKUs before the matrix is assembled.
  • Retailer: a trend is taken from a brand without a right to reorder and a delivery lead time → at good demand a chase is impossible → fix the reorder in the contract.
  • Retailer: the seasonal growth of queries is taken for a trend → the demand of a category is overestimated → compare with the same period last year and with the category.
  • The life window is treated as a fact and not revised → a trend that has already gone is chased → update the estimate by the signal series and the test sales.

Russian specifics

  • Brand: the signal series is built from aggregated data (queries, mentions, sales by keywords), the customers’ personal data are not needed for it; if the competitors’ data are collected from marketplaces automatically, check the marketplace terms and the law, unverified.
  • Retailer: the right to reorder, the delivery lead time and the terms of returning the remainder of a test batch are fixed in the supply contract; check the wording and the tax consequences of a return with a lawyer and an accountant, unverified.
  • For garments imported into the EAEU the time to the shelf includes certification and marking: count them in the time to the shelf and not after; check the requirements against the current acts.

Open interactively →