A collection roadmap by gates: hypotheses, the pilot batch, stop criteria and decisions (stage-gate)
How to lead the models of a collection from an idea to scale in the logic of product management: to formulate a hypothesis, to put a gate after every stage with criteria set before it begins, to check a model with a pilot batch and to stop a weak one while it is cheap. The document shows the funnel of ideas, the price of a hit, the gates’ saving against “all ideas go to the end” and the stop rule of the pilot batch, and links the roadmap with the calendar, purchasing and tests.
Why this document
Most costly mistakes of a collection are made at the end of the path: a sample and a pilot batch cost tens of times more than a sketch, and it is “a pity” to stop a model into which you have already invested. Without gates decisions are taken by mood and weak models reach purchasing. The collection head, the product manager, the category manager and the founder read this document when planning a season and a new line. It gives an order: a hypothesis, the gate criteria, the pilot batch, the stop rule, the decision. The “Collection gates” calculation counts the funnel and the economics, while the document on launching a label is responsible for the readiness of tasks.
At a glance
- What is this document and why is it needed?
- How to lead the models of a collection from an idea to scale in the logic of product management: to formulate a hypothesis, to put a gate after every stage with criteria set before it begins, to check a model with a pilot batch and to stop a weak one while it is cheap. The document shows the funnel of ideas, the price of a hit, the gates’ saving against “all ideas go to the end” and the stop rule of the pilot batch, and links the roadmap with the calendar, purchasing and tests.
- Who owns the document and how often is it rebuilt?
- The collection head or the product manager leads the roadmap, the category manager and the CFO approve the criteria and budgets of the gates, the buyer confirms the scale; the criteria are written before a stage, the decisions are recorded in a gate log each season.
- What mistakes are the most common?
- Brand: sets the stop threshold after sales start → it is moved to fit the result → record the threshold in units before the first sale and do not change it. Retail: lets weak models through the gates “out of friendship” → the funnel stops saving → write the criteria before the stage and check against them aloud. Brand: counts the price of a hit only by the pilot’s spending → understates it many times over → include the rejected models: the price of a hit is all the funnel’s spending divided by hits.
For whom
- Brand
- A brand leads models from an idea through a tech pack, a sample and a pilot batch to a scale purchase: the gates decide which models enter the capsule and which stay a sketch; the sample budget, the limit of simultaneous pilots and design rights matter.
- Retail
- A retailer with a private label or exclusives leads the same gates on a shorter path: choosing the supplier’s models, a pilot batch in a few points or on a platform and a re-order by the fact; the pilot points, the control stores and the possibility of returning the remainder matter.
What it consists of
A hypothesis, gates and decisions
- A collection hypothesis is a testable statement: who the buyer is, which garment, at what price and how many we sell by which week; “it will sell well” is not a hypothesis but a hope.
- A gate is a decision point after a stage: go on, rework or stop. The criteria are written before the stage and not moved after; a gate without a set threshold turns into a conversation.
- The stages: an idea and a hypothesis, a sketch and a tech pack, a sample and a fit test, a pilot batch and sales, the scale. The cost grows by an order of magnitude at each, so gates are put before every costly step.
- A pilot batch is an experiment and not a small order: the term, channels, price, the control week and the sell-through threshold are recorded in advance, and what is done on crossing the threshold up and down.
- The decisions by the results: scale (re-order and purchase), rework (price, card, fit), pause (waiting for data) or stop (selling the remainder and closing the model). A stop is not a failure but a result of selection.
How gates work for a brand and a retailer
- Brand: put gates by the development stages and write the criteria for each: at the sketch conformity with the brief and the price, at the sample fit and cost within tolerance, at the pilot batch the sales threshold; the log of decisions is kept in a PLM or a table.
- Retail: gates are shorter: choosing from the supplier’s catalog, a pilot at a few points or on a platform, a re-order by the fact; the threshold is set in units by the control week and the control points are compared with the rest.
The link with the calendar and purchasing
- The gates are tied to the dates of the collection calendar: the terms of samples, pilots and scale run back from the date on the shelf; a late gate shifts the whole path, so every gate has a term of decision.
- The scale goes through the purchasing budget and the sales plan: a hit enters purchasing under the budget, and the number of pilots is limited by a limit per season so as not to blur money and attention.
- The readiness of the label launch tasks (an entity, the brand, marking, channels) is counted separately by launch stages; here a different question is decided: which models are worth ordering at all. Do not mix these two lists.
Where the data comes from
| Data source | Fields |
|---|---|
| Past seasons | how many ideas reached a tech pack, a sample, a pilot and a hit; the stage cost per model; the profit of a hit |
| The pilot batch | units, the unit cost, channels, the control week, the sell-through threshold, the recovery of leftovers |
| Calendar and budget | the gate dates, the limit of pilots per season, the sample budget, the scale batch and the unsold share in a failure |
How to build it step by step
- Step 1: Brand: write the hypothesis of each model in the form “to whom, what, at what price, how many we sell by week W”; a model without a number in the hypothesis does not enter the funnel.
- Step 2: Retail: collect the pass share by stage from past seasons and the stage cost per model; for a new brand take cautious values and mark them “expectation”.
- Step 3: Brand: run the gates calculation and look at how many hits the funnel gives, what one hit costs and how much the gates save against “all ideas to the end”.
- Step 4: Retail: write the criteria of each gate before the stage: what must be true to go on (fit in tolerance, cost not above, the target by price); the head signs the criteria.
- Step 5: Brand: set the pilot batch, the control week and the sell-through threshold in units; compare the threshold with the model’s expected sales, it must not be obviously unattainable or trivial.
- Step 6: Retail: run the pilot at a few points or in a channel and do not change the price and the card before the control week: otherwise it cannot be understood what worked.
- Step 7: Brand: at the control week compare the fact with the threshold: above the threshold go to scale, below stop or rework once; record the decision in the log with a number.
- Step 8: Retail: at the end of the season recount the pass shares and the stage costs by the fact and update the criteria: the funnel learns from its own data, otherwise it stays an assumption.
Formulas
| Measure | Formulas | Note |
|---|---|---|
| The funnel by stage | Nᵢ = N₀ × Π(pⱼ), j < i; hits = N₀ × Π(pⱼ) | N₀: ideas at the entry, pⱼ: the pass share of gate j; Nᵢ: models that entered stage i. |
| Spending and the price of a hit | Σ Nᵢ × cᵢ; the price of a hit = Σ Nᵢ × cᵢ ÷ hits | cᵢ: the stage cost per model; the rejected models are in the price of a hit too. |
| The gates’ saving | N₀ × Σ cᵢ − Σ Nᵢ × cᵢ | The spending “all ideas go to the end” minus the spending with gates. |
| The pilot stop rule | threshold = pilot batch × s; the loss on stopping = batch × (1 − s) × cost × (1 − r) | s: the sell-through threshold at the control week, r: the share of cost recovered from the leftovers. |
Formulas in notation
- The funnel and hits
N_i = N_0 · Π p_j, H = N_0 · Π p_j
N_0: ideas at the entry, p_j: the pass shares of the gates before stage i (all gates for hits), H: hits.- The price of a hit and the gates’ saving
P_H = Σ N_i·c_i / H, S = N_0·Σ c_i − Σ N_i·c_i
c_i: the stage cost per model; P_H: the price of a hit including rejected models; S: the gates’ saving against “all ideas to the end”.
Worked example
20 ideas, four gates, the profit of a hit RUB 2.2m; a pilot batch of 150 units, a threshold of 20% by week 3; all values conditional
| Stage | Models at the entry | Cost per model, RUB | Stage spend, RUB | Share passing, % | Reached the end of the stage |
|---|---|---|---|---|---|
| 1. Idea and hypothesis: brief, references, target calculation | 20.0 | 4,000 | 80,000 | 60 | 12.0 |
| 2. Sketch and tech pack | 12.0 | 25,000 | 300,000 | 50 | 6.0 |
| 3. Sample and fit test | 6.0 | 55,000 | 330,000 | 60 | 3.6 |
| 4. Pilot batch and sales | 3.6 | 257,500 | 927,000 | 35 | 1.3 |
About 1.3 models out of 20 reach a hit (6.3% of ideas); RUB 1,637,000 is spent in all, the price of a hit is about RUB 1.3m at a profit of a hit of RUB 2.2m, the expected profit of the funnel is RUB 1,135,000. Without gates the spending would be RUB 6,830,000: the gates save RUB 5,193,000. Stopping a pilot batch at the threshold costs RUB 87,000 against RUB 1,305,000 on a failed scale without a pilot. The numbers are illustrative.
Figures
Gates and decisions on a model’s path. After each stage the gate gives three outcomes: go on, rework once or stop; the last gate decides the fate of the scale.
- Idea and hypothesis → Sketch and tech pack
- Sketch and tech pack → Sample and fit; Stop or rework
- Sample and fit → Pilot batch; Stop or rework
- Pilot batch → Scale; Stop or rework
- Scale
- Stop or rework
Data as a table
| Measure | Models |
|---|---|
| 1. Idea and hypothesis: brief, references, target calculation | 20 |
| 2. Sketch and tech pack | 12 |
| 3. Sample and fit test | 6 |
| 4. Pilot batch and sales | 3.6 |
| Hits | 1.3 |
How many models enter each stage. The funnel narrows from 20 ideas to 1.3 hits: the most expensive stage gets 3.6 models instead of 20.
- With gates
- Without gates
Data as a table
| Measure | With gates | Without gates |
|---|---|---|
| 1. Idea and hypothesis: brief, references, target calculation | 80 | 80 |
| 2. Sketch and tech pack | 300 | 500 |
| 3. Sample and fit test | 330 | 1,100 |
| 4. Pilot batch and sales | 927 | 5,150 |
Spending by stage: with and without gates. Gates cut the spending at the costly last stages: RUB 927 thousand instead of RUB 5.15m on the pilot batch.
Specifics by category
- Apparel
- Apparel: the sample and the fit test give the main rejections, the pilot batch goes in the popular sizes; seasonal capsules need hard gate dates, otherwise the pilot cannot be done before the season.
- Footwear
- Footwear: the last, samples and tooling cost more, the development cycle is longer, the pilot batch is larger and the size run wider; there are fewer gates but each costs more, the pass shares are higher because weak models are screened out earlier.
- Accessories
- Accessories: samples are cheaper and a pilot can be a small series; the funnel is wider and faster, hits are determined by the speed of sales and the gates are set by the number of units sold at the control week.
- Kids
- Children’s: safety and certification add mandatory gates before sales, the pilot goes in popular sizes with a margin for the child’s growth; the pass shares are lower because of the requirements, the terms are longer.
- Men
- Men’s: a basic line, many repeats; the funnel is shorter, the gates more often check the differences from last season and the cost, and new models are allowed in a limited number.
- Women
- Women’s: more ideas, faster fashion, lower pass shares; a wide funnel, cheap early stages and a hard pilot threshold are needed, otherwise the season slides into discounts.
Common mistakes
- Brand: sets the stop threshold after sales start → it is moved to fit the result → record the threshold in units before the first sale and do not change it.
- Retail: lets weak models through the gates “out of friendship” → the funnel stops saving → write the criteria before the stage and check against them aloud.
- Brand: counts the price of a hit only by the pilot’s spending → understates it many times over → include the rejected models: the price of a hit is all the funnel’s spending divided by hits.
- Retail: scales on the first days of sales → one spike is taken for demand → wait for the control week and compare with the threshold.
- Brand: takes the pass shares “from the ceiling” → the funnel misleads → count from past seasons and hold the values as a range.
Russian specifics
- Brand: the pilot batch is sold to buyers, so the same requirements apply to it: conformity confirmation under the technical regulation, marking and labels before sale; do not plan a pilot without documents. Rule: CU Commission decision 876 of 09.12.2011: TR CU 017/2011 on safety of light-industry products (Customs Union Commission Decision of 09.12.2011 No. 876).
- Retail: selling the pilot batch on a platform or in a store goes by the rules of consumer rights protection: returns and exchange, a reliable card and price (law of 07.02.1992 No. 2300-1); include returns in the calculation of the threshold. Rule: Law 2300-1 of 07.02.1992: Law of 07.02.1992 No. 2300-1 on consumer rights protection.
- Brand: secure the rights to a model’s design and trademark before the sample and the pilot: with a contractor factory and a designer the rights may remain with them under the contract; check the terms with a lawyer.
- Retail: the pass shares, the sell-through thresholds, the stage costs and the profit of a hit have no norm and are counted from your data; in Lab they are illustrative and marked as an expectation.