Where the curves come from
The sector estimate runs on assumptions we wrote. Here they are, with the evidence behind each one and the places where there is none.
The short version
If you upload your own launch history, none of this applies to your answer: your launches replace these patterns entirely, and they are better evidence than anything on this page. The sector estimate exists for the case where you have no usable history to hand, and it is built from published research where published research exists, and from our own judgment where it does not.
The single most important thing to know about that research: there is no published study of week-by-week launch trajectories for industrial products sold through distribution. We looked. The diffusion literature is consumer durables and country-level adoption; the operational launch-curve work is electronics and retail. Every sector shape here is an extrapolation from an adjacent domain, and that is a statement about the state of the field rather than about our effort.
The six shapes a launch can take
Every sector in the tool is built from these, and each one names what it rests on. Read the badge as covering the shape: that a launch of this kind rises and then plateaus is research, while how many weeks the ramp takes is almost always ours. Each family says which is which, because a badge on the family that quietly covers the numbers inside it would be the most flattering way to be wrong.
S1. Slow build Published research
A near-linear rise across all 26 weeks with no peak inside them. Week 26 runs about 1.2 to 2 times week 1.
Applies to: Genuinely new technical products that have to be specified in or adopted by installers.
Evidence: Follows from the diffusion literature at its own average parameters: a meta-analysis of 213 Bass-model applications (Sultan, Farley and Lehmann, 1990) puts the adoption peak at about six years, so inside 26 weeks there is no peak to see, only the start of the ramp. Contractor-channel friction is documented separately for heat pumps.
What the badge does not cover: The shape is the research; the steepness is ours. The tool ships a rise of 1.8 times from week 1 to week 26, chosen inside the 1.2 to 2 the literature implies. Nothing picks 1.8 out of that range but us.
S2. Fast riser to a plateau Published research
Ramps over two to six weeks, then flat to week 26 with no decay inside the window.
Applies to: Line extensions, range refreshes and catalogue additions where the demand already exists and transfers from something else.
Evidence: The trapezoid is one of the two best-fitting simple shapes on 170 Dell products covering four million units of B2B customer orders (Hu, Acimovic, Erize, Thomas and Van Mieghem, 2019). That is the strongest published B2B launch-shape evidence we found.
What the badge does not cover: How long the ramp lasts, two to six weeks depending on the sector, and how low it starts. Both are ours.
S3. Pipeline fill Known in the trade, not measured
Weeks 1 to 3 spike as branches and key customers stock up, weeks 4 to 8 dip below the run rate while that stock sells down, then it settles.
Applies to: Anything pushed into a distribution network with an initial stocking programme. Applied only when you tell the tool this order is sell-in.
Evidence: Sell-in is not sell-through, and every distributor knows it. We found no academic quantification of it for distribution. It is included because ignoring it overstates real demand in the early weeks, and the size of the spike and dip are ours.
S4. Fast peak, then decay Published research
Peaks in weeks 2 to 6, then declines steadily.
Applies to: Launches with a marketing push behind them, and some equipment launched at a trade show.
Evidence: The triangle fits a subset of the Dell clusters (Hu et al., 2019) and appears in shape-clustered retail work (Baardman et al., 2018; van Steenbergen and Mes, 2020). Not expected for slow-moving B2B categories.
What the badge does not cover: Which week it peaks in and how far it falls by week 26.
S5. Sparse and lumpy Published research
Not a curve at all. Orders arrive occasionally, in uneven sizes, with many weeks of nothing.
Applies to: The long tail of valves, fittings, components and spare parts, which is most of what technical distribution adds.
Evidence: The classification and its thresholds are standard: average interval above 1.32 weeks and squared order-size variation above 0.49 (Syntetos, Boylan and Croston, 2005), with distributional forecasts beating point forecasts on nine industrial datasets (Willemain, Smart and Schwarz, 2004).
What the badge does not cover: Applying that framework to launch weeks at all: the literature is about established items. And the values inside it, a two-week average interval and an order-size variation of 0.9, are ours, chosen to satisfy both documented thresholds at once.
S6. Season gated Published research
Any of the above multiplied by a seasonal index, where the launch month matters more than the launch curve.
Applies to: HVAC and seasonal equipment, and the outdoor season in building materials.
Evidence: Radas and Shugan (1998) show seasonality is properly modelled as time running faster in high season, applied to the underlying curve, and that the best launch timing depends on the curve's shape. That is the mechanism.
What the badge does not cover: Every monthly number. The research says plainly that real seasonal indices have to come from category history, which we do not have, so the shape of the season is right in kind and ours in size. These are the least defensible numbers in the tool.
What each sector is made of
The shape families each sector is built from, and what the tool assumes about its season. These are the actual settings the simulation uses, read out of the engine rather than retyped, so this table cannot drift away from the code.
| Sector | Shapes | Season |
|---|---|---|
| Technical distribution (valves, fittings, components, spare parts) | S5-intermittent, S2-trapezoid | No seasonal pattern assumed |
| HVAC and seasonal equipment | S6-season-gated, S2-trapezoid, S1-slow-build | Season gated: the launch month moves the answer |
| Building materials and sanitary | S2-trapezoid, S6-season-gated, S5-intermittent | Season gated: the launch month moves the answer |
| Electrical components | S2-trapezoid, S5-intermittent, S1-slow-build | No seasonal pattern assumed |
| Professional equipment | S1-slow-build, S4-triangle | No seasonal pattern assumed |
| Consumer durables import or brand distribution | S4-triangle, S2-trapezoid | No seasonal pattern assumed |
| Private-label ecommerce | S2-trapezoid, S4-triangle | No seasonal pattern assumed |
| Other or not sure | S2-trapezoid | No seasonal pattern assumed |
How wide the range is, and how much to trust it
The width of the band comes from one question, how new the product is to your range, and from one source. A survey of 168 firms reports forecast accuracy by newness class: 63% for a line extension, 47% for something new to the company, 40% for something new to the world (Kahn, 2002). Newness is the strongest documented driver of forecast error, so the tool asks about it directly rather than asking how confident you feel.
Turning an accuracy percentage into a range is our step, not the survey's, and it is the single assumption with the most leverage over every number the tool produces. It is written down as such. Sectors deliberately do not add a second width on top: the sector figures published anywhere were themselves inferred from the same survey, so adding them would count one piece of evidence twice and make every range look wider than the evidence supports.
So read the range as scenarios, not as probabilities. When the tool says P10 and P90 it means the low and high ends of the spread it explored, and we have not checked how often real launches land inside that spread, because doing so needs the backtesting that is not built yet. What the survey supports firmly is the order: a variant of something you already sell is more predictable than something new to you, which is more predictable than something new to the market. The tool asks for that ordering and uses it. The exact width attached to each step is our arithmetic on top of it, and it is the first thing that should be replaced by measurement rather than defended.
What this cannot tell you
- Whether your product is like the sector. These are patterns for a category, not a forecast of your item. The tool says so on the result and it is not false modesty.
- Anything about your customers. A named customer waiting on the launch, a project that will absorb half the order, a competitor discontinuing theirs: all of that beats a sector pattern, and none of it is in here.
- Seasonal magnitudes. The seasonal shapes are right in kind and ours in size. They are the numbers we would replace first given real category history.
The honest summary is that these are informed starting points, better than a single average and much worse than your own launch history. If you have that history, use Company data instead and this page becomes irrelevant to you, which is the intended outcome.