Season results: brand, category, SKU
From gross sales to contribution; sell-through by delivery; bestsellers, slow sellers and no-sale SKUs; the contribution of categories to growth; new items and carryover; brand history by season.
About the topic
Season results exist to understand what happened to the money and to decide what to do next season. Read them in steps, like a staircase down: gross sales at list price, minus discounts, minus returns, gives net sales; minus cost of goods, gives gross margin; minus write-offs and selling costs, gives contribution. Example: gross RUB 10.0 million at list price, discounts 2.0 million, returns 0.5 million: net RUB 7.5 million. Cost of goods sold RUB 3.0 million: gross margin 4.5 million, which is 60% of net. Write-offs of 0.3 million and selling costs of 2.4 million leave contribution of RUB 1.8 million. You can see where each rouble went and that discounts ate a fifth of list price.
Next the season is split by deliveries. If of 1,000 units delivered at the start 650 were sold by day sixty, sell-through (the share sold of what was delivered) is 65%. If 480 of them sold at full price, full-price sell-through is 48%, and the gap shows how much sold at a discount. The higher the full-price share, the healthier the season. Sales speed and time to clearance suggest when to start discounts.
Then the season is split by SKU: bestsellers, middle, low sellers and SKUs with no sales. Usually a small share of SKUs gives a large share of sales: for example 20% give 55%. That is not bad by itself but tells where there will be shortage and where leftovers. Bestsellers must be replenished in time, low sellers not repeated next season, and no-sales SKUs reviewed: is the problem price, size or display.
Last comes comparison. New arrivals versus carryover (items that carry on from season to season) and the brand’s history across seasons: where growth, discounts and margin are going. Growth through discounts and growth through full price are different stories.
Start with the net staircase, then sell-through by delivery, then SKUs. Definitions of figures (gross, net, low seller) may differ in your company, and thresholds are set by the user: there is no single standard. Traps: comparing seasons of different length, mixing gross and net, judging by total without margin.
Key ideas
The staircase from gross to contribution
Every point of discount costs more than it seems. With gross of RUB 8.0 million, discounts of 1.6 and returns of 0.4 give net 6.0; cost of 2.4 leaves margin of 3.6 million, and costs with write-offs of 2.4 leave contribution of RUB 1.2 million. One extra point of discount is RUB 80,000, which is 6.7% of contribution: contribution is thinner than margin.
Full-price sell-through is more honest than total
Watch the full-price sell-through curve at checkpoint days. For example 25% on day 30, 40% on day 60 and 46% on day 90: the monthly gain fell from 15 to 6 points. Full-price demand is used up and the remaining stock simply waits for a markdown. That is the signal to start the clearance, not to wait for the end of the season.
SKUs are split into action classes
Concentration makes a single SKU vulnerable. If a bestseller gives 12% of season sales over 13 weeks and arrives three weeks late, the season loses about 12% x 3 / 13, roughly 2.8% of sales, because of it alone. Such a SKU needs reserve capacity at the factory and a buffer in the schedule in advance.
New arrivals and carryover sell differently
Carryover sells season after season: it is the base and stable margin. New arrivals give growth and risk. Compare return per SKU and sell-through of both groups to decide where to widen assortment.
Season history shows direction
Growth of 15% with discounts that rose from 12% to 20% of list price is growth bought with margin. A single year-on-year comparison does not show what growth was bought with. Look at three or four seasons in a row.
The lag of the naive baseline: the longer it is, the easier it is to “win”
The naive baseline takes the actual of period t − lag. The same AI forecast gives an improvement of 15.8% at lag 1, 56.2% at lag 2, 68.2% at lag 3 and 75.0% at lag 4: the naive error grows from 14.4% to 47.0% because the seasonal curve moves away from the old value. This is not a win of the method but a weak baseline. For season history take the strongest simple baseline (last week or the same period of the last season) and state the lag next to the result.
Formula: Naive forecast of period t = actual of period t − lag
What to do: Always state the lag of the baseline and take the strongest of the simple options.
Think first: The AI “improves” by 56% against a baseline with lag 2. Is that a strong result?
Answer: No: at lag 1 the improvement is 15.8%. The big number is given by a weak baseline, not by a strong method.
How to work through it
- Build the staircase: gross, discounts, returns, net, cost, margin, costs, contribution.
- Calculate sell-through for each delivery: total and at full price, and time to clearance.
- Split SKUs into bestsellers, middle, low sellers and no sales by thresholds you set.
- Identify causes for no-sales items: price, size, display, late delivery.
- Compare categories with last year on contribution to growth and on margin.
- Compare new arrivals and carryover by return per SKU, sell-through and margin.
- Write a decision for each class for next season: replenish, repeat, drop, review.
Where to start
First “Season result: from gross sales to contribution”, then “SKU classification” and “Sell-through by delivery”. For year on year: “Categories” and “Brand history”. Calculations of the topic in order: “Season result: from gross sales to contribution”, “Season stages: BS, IS, AS, SO”, “Sell-through by delivery: speed and time to sell out”, “SKU classification: ABC and sell-through, bestsellers, slow sellers, no-sale”, “Categories: contribution to growth and margin, this year versus last”, “New and carryover: share, return per SKU, sell-through and margin”, “Brand history by season: growth, discounts, margin and sell-through”, “Season matrix: product season × sales month and stock against the window”, “Checking an AI forecast against a naive baseline: WAPE, bias, improvement, share of periods and the gain net of costs”.
Calculations in this topic
Season result: from gross sales to contribution
Lays out the season result as a ladder: gross sales at list price, discounts, returns, net sales, cost of goods, gross margin, write-offs, selling costs. Shows how many margin points discounts ate. Fits a brand, a category and a retailer; check the names of the rungs against your own reporting.
Where the season’s gross sales went and how much margin discounts ate.
Season stages: BS, IS, AS, SO
Splits a season into four stages (before the sale, in the sale, after it, the final clearance): sales, share of the stage, discount depth, gross margin, stock at the end of the stage and cover. Shows how much the season earned at full price, how much it gave away in discounts and where the leftover remained.
How much the season earned at full price, how much it gave away in discounts and at which stage the leftover remained.
Sell-through by delivery: speed and time to sell out
Compares the deliveries (drops) of a season: what share of each delivery is sold and what share at full price, at what speed it sells and in how many weeks it will finish at the current pace. Shows which delivery has become heavy.
Which of the season’s deliveries sells slowly and when it will finish.
SKU classification: ABC and sell-through, bestsellers, slow sellers, no-sale
Ranks SKUs by net sales, assigns class A, B or C by the cumulative share and compares it with sell-through. A bestseller is class A with a high sell-through, a slow seller is one with a low sell-through, “no-sale” is a SKU with no sales. You set the thresholds; the calculation shows the quartiles of your data as a hint.
Which SKUs are bestsellers, which are slow sellers and which do not sell at all, and how much stock lies in them.
Categories: contribution to growth and margin, this year versus last
Shows which categories raised or dropped sales and how many points of growth each gave, and splits the change in the total margin into a mix effect (category shares shifted) and a rate effect (margin within categories changed).
Which categories gave growth and which lost margin and what changed the total margin.
New and carryover: share, return per SKU, sell-through and margin
Compares groups of SKUs of a season, for example new items, carryover (carried over from past seasons) and basics: the share in the range and in sales, sales per SKU, sell-through and margin. Shows which group gives a return above its share in the range.
Which group of SKUs (new, carryover, basics) returns more than its share in the range.
Brand history by season: growth, discounts, margin and sell-through
Builds a history report for a brand or category: net sales, discount depth, gross margin and sell-through by season, and compares each season with the same season a year earlier and the last year with the previous one.
Where the brand is heading from season to season: growth, discounts, margin and sell-through year on year.
Season matrix: product season × sales month and stock against the window
From 12 months of sales in four groups (SS, AW, leftovers of past seasons, core) calculates the share of seasonal goods and leftovers by month and calendar season, how much of its sales each season makes inside its own window and whether the window is enough to sell the stock.
What part of sales comes from seasonal goods, leftovers and core by month and whether stock fits in the season window.
Checking an AI forecast against a naive baseline: WAPE, bias, improvement, share of periods and the gain net of costs
From a series of actual sales and an AI forecast for the same past periods it builds a naive forecast (the value of the period t − lag: last week or the same period of the last season) and compares with it. It counts the WAPE and the bias of the AI forecast and of the naive one, the improvement against the baseline, the share of periods where the AI is more accurate, whether the improvement threshold is passed, the saved error in units and the gain in money net of the cost of the tool. It answers whether the AI forecast deserves more trust than a simple rule. The forecast must have been made before the fact and not fitted; the default data are illustrative.
Whether the AI forecast is more accurate than the simple rule “as in the last period”, by how much, in how many periods and whether it pays off after the cost of the tool. WAPE and bias are counted by the definitions of the “Forecast accuracy by week” calculation (fcdash).
Self-check questions
- Are gross, net and season boundaries defined consistently?
- What share of list price did discounts and returns eat?
- What is full-price sell-through for each delivery?
- What share of SKUs gives most of the sales?
- Which SKUs had no sales and why?
- Which category gave growth and which lost margin?
- Where are discounts and margin heading across recent seasons?
- What do we do differently next season, and which numbers show it?
Knowledge check
Multiple-choice questions. The correct answer and the explanation follow each question.
Gross sales of the season at list price are RUB 50 million, discounts RUB 10 million, returns RUB 2.5 million, the cost of the units sold RUB 15 million. What is the gross margin and how many pp did discounts take?
- A. 60.0%; discounts took 8.4 pp
- B. 45.0%; discounts took 8.4 pp
- C. 60.0%; discounts took 20 pp
- D. 70.0%; discounts took 10 pp
Correct answer: A. 60.0%; discounts took 8.4 pp. Net sales are 50 − 10 − 2.5 = RUB 37.5 million, the gross margin is (37.5 − 15) / 37.5 = 60.0%. Without discounts the margin of the same units is (37.5 + 10 − 15) / (37.5 + 10) = 68.4%: discounts took 8.4 pp 45% appears if the margin of RUB 22.5 million is divided by gross RUB 50 million; 20 pp is the depth of the discount in gross sales (10 / 50), not a loss of margin; 70% is the margin on gross sales (35 / 50).
A delivery of 2,000 units, 600 sold in 4 weeks. In how many weeks will it be gone at the current pace?
- A. 4.0 weeks
- B. 13.3 weeks
- C. 9.3 weeks
- D. 2.3 weeks
Correct answer: C. 9.3 weeks. The pace is 600 / 4 = 150 units a week, the remainder is 2,000 − 600 = 1,400, the time to the end is 1,400 / 150 = 9.3 weeks. 13.3 is the whole delivery / pace (2,000 / 150), ignoring what is already sold; 2.3 is the remainder / sold (1,400 / 600), without conversion to weeks; 4.0 is the time the delivery has already been on sale. Sell-through is now 600 / 2,000 = 30%, the speed is 7.5% a week.
At which lag of the naive baseline is the AI improvement over it the largest among 1, 2, 3 and 4?
- A. lag 1
- B. lag 2
- C. lag 3
- D. lag 4
Correct answer: D. lag 4. The improvement grows with lag: 16%, 56%, 68%, 75%. But it is a weakening baseline, not a method. Covered in: The lag of the naive baseline: the longer it is, the easier it is to “win”.
The naive baseline error grew from 14.4% at lag 1 to 27.9% at lag 2. Why?
- A. an error in the data
- B. a value from two weeks ago is outdated: the season curve changes
- C. the AI got worse
- D. the price of error grew
Correct answer: B. a value from two weeks ago is outdated: the season curve changes. The series is seasonal: the older the value, the further it is from the current one. So the lag must be named next to the result. Covered in: The lag of the naive baseline: the longer it is, the easier it is to “win”.
Which baseline to take for an honest check by season history if the improvement is 15.8% at lag 1 and 68.2% at lag 3?
- A. lag 3: the improvement is higher
- B. lag 1: the strongest simple baseline
- C. the average of the lags
- D. any: there is no difference
Correct answer: B. lag 1: the strongest simple baseline. A method must beat the strongest simple baseline. Choosing a weak baseline overstates the result and does not show whether the method is needed. Covered in: The lag of the naive baseline: the longer it is, the easier it is to “win”.
Common mistakes
- Mixing gross and net. Consequence: margin and growth are calculated on different bases. Avoid: label the base at every step.
- Looking at overall sell-through. Consequence: discounted sales hide a weak season. Avoid: calculate full-price too.
- Comparing seasons of different length and mix. Consequence: false trends. Avoid: compare comparable windows and assortment.
- Judging by total without margin. Consequence: a growing category can lose money. Avoid: look at contribution and margin.
- Repeating low sellers next season. Consequence: the same aged leftovers. Avoid: decide by the cause of weak sales.
Russia and the EAEU
In Russia season results are easiest to read in roubles without VAT: the general rate from 2026 is 22%, so revenue of RUB 122 with VAT is RUB 100 net. For the children’s category with confirmed conditions the 10% rate applies. Gross at list price and net by receipts differ in accounts by discounts and returns, so label the base in the report.
Stock at season end is measured in accounting under FSBU 5/2019 on inventories: if net realizable value is below cost, a provision for impairment must be created. This ties to discount decisions: better to plan the write-down in advance than to recognize it unexpectedly. Check the tax consequences of markdowns and write-offs with your accountant.
Marking affects the season: stock without a code cannot be sold after the set deadlines, so for marked goods the clearance date is limited. The Russian sales calendar is set by seasonal peaks: 23 February, 8 March, 1 September, New Year and marketplace sales; compare seasons on comparable windows. For marketplace sales distinguish ordered from completed purchases: returns after delivery change the net total.
The method of season review is the same. Calendar, taxes and currency differ. VAT rates by country are on the page “EAEU and CIS countries” (indicative as of 14.11.2026, check with an accountant). Seasons are shifted by climate and holidays: Nauryz (21 to 23 March) in Kazakhstan and Kyrgyzstan, Muslim holidays on the lunar calendar, local school dates. Results in different currencies cannot be added without a rate, and the rate fluctuates: calculate separately in local currency and then consolidate. Rules for valuing stock and provisions are local: indicative as of 14.11.2026, check the text of the act.
Categories and audiences
- Apparel
- If it is apparel: the season is rigid, stock ages fast, so full-price sell-through and time to clearance are the main metrics. Classify SKUs early: low sellers show by week four. Deepen bestsellers and replenish the size run by sales.
- Footwear
- On footwear: the season is longer, a pair is tied to a size, stock breaks along the size run. A low seller is often simply extreme sizes. Look at sell-through by size, not just by model, and for how many weeks a full run holds.
- Accessories
- In accessories: less seasonality, high carryover, so analysis goes by return per SKU and stock turnover. Gift peaks (February, March, December) define the season. Small no-sales SKUs occupy space: clean the assortment.
- Kids
- For children’s wear: the season is tied to school and cold weather, deliveries must arrive before the peak. Size run by height, a leftover size loses value after a season. Sell-through by delivery matters: a late delivery sells at a discount.
- Men
- If it is the men’s range: more carryover (basics), a lower share of new arrivals, so growth comes from categories and price. Sales are weaker, base stock carries to the next season. Look at margin by category: suits, outerwear, shoes.
- Women
- In the women’s line there are more new arrivals and each has a shorter season window, so full-price sell-through is better counted in weeks since a model’s launch. Record which models have missed plan by week three: the markdown decision on them is made earlier than for the season as a whole. Compare the full-price share across seasons so that net sales do not grow merely through discounts.
Questions this topic answers
- Where did gross sales go and how much margin did discounts eat?
- Which deliveries and SKUs sell slowly?
- Which categories gave growth and which lost margin?
- Do new items and carryover return more or less than their share in the range?
- Where is the brand heading from season to season?
Key measures and terms
- Sell-through (share sold)
- Full-price share
- ABC analysis
- Maintained margin (MMU)
- GMROI
- Cost of goods sold (COGS)
- Season result ladder
- Sell-through by delivery and speed
- SKU classification: bestseller, lowseller, no-sale
- Mix effect and rate effect
- Return index of a group of SKUs
- Growth against the same season
- Season codes and collection windows
- Carryover report
- Category report
Study first
What next
Related: “Assortment and sizes”, “Collection and range”, “Markdown and exit price”.
Documents on this topic
- Plan versus actual and profit factor analysis (bridge)
- Accessories: bags, leather goods, belts, headwear, scarves, jewelry
- Kids: height sizing, back-to-school, TR CU 007/2011 and 10% VAT
- Niches: sport and outdoor, underwear, outerwear, knitwear, bridal, school uniform, workwear
- Accessories line economics: bags, jewelry, hats and belts
- Demand forecasting: smoothing, seasonal indices, analogs and error control
- Markdown mathematics: steps, leftovers, timing and liquidation
- Holt-Winters, seasonal decomposition and Croston: forecasting with trend, season and intermittent demand
- 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
- AI in fashion and analytics: where to apply it, where not, how to check the result, which data are needed and which errors occur. A memo and a checklist