Streams progressively — you don’t need to wait for the whole file. You can also download it to listen offline.
Every forecast starts with a question you can’t answer from a spreadsheet: how much will people want this? For an established SKU, history gets you most of the way. For anything new, history doesn’t exist yet — and that’s where most inventory mistakes are made.
This episode looks at the two very different tools available for that problem, and why sellers who rely on only one of them keep getting caught out.
In this episode
- 00:00Why new products break traditional forecasting
- 04:30What AI can infer from weak and indirect signals
- 09:15Where models confidently get it wrong
- 13:40What customer interviews reveal that data cannot
- 17:20Combining both into a demand range, not a number
- 20:05Turning that range into a purchase order
Chapter times are approximate.
Why this matters for inventory
Every other piece in this section deals with forecasting once you have sales history to work from. This one deals with the harder case — the launch, the new variation, the move into a new marketplace — where the forecast has to come from somewhere other than the past.
Forecasting starts by identifying uncertainty, not eliminating it. A model gives you a starting point. A conversation tells you which of its assumptions are wrong.
Related reading
If the episode leaves you wanting the mechanics written down:
- The reorder point formula — what to do once you do have a demand estimate.
- Asking better questions — the argument that forecasting is a decision problem, not a maths problem.
- Why the classic formula is wrong — including demand scenarios instead of single numbers.