Every planner knows this moment. A new style hits the line. A replacement item takes over for something discontinued. A slow-mover barely has enough sales data to trust. The forecast doesn't know what to do with any of it.
Left alone, the system flatlines the number, defaults to a generic category average, or leaves you to eyeball it based on gut feel and a spreadsheet of "stuff that seemed kind of similar." So you do what planners have always done. You dig through history for an item that looks and sold like the new one. You pull its sales curve into a side calculation. You hand-adjust the new item's forecast to roughly match it.
That workaround has a name: analogous forecasting, or like-for-like forecasting. It's one of several ways to improve accuracy on items with little history, and if you want the full set, start with our guide to using AI to improve forecast accuracy for new and slow-turning products. This post goes deep on just the like-for-like method: what it is, why the manual version breaks down, and how to make it repeatable.
Forecasting New products Is a Process Problem
Planners already know how to solve this problem. The issue is that the solution is manual, it lives nowhere the system can see, and it has to be rebuilt from scratch every time a new or replacement choice comes up. Multiply that across a full assortment of new styles each season, and it becomes a recurring tax on time that never shows up as a line item anywhere.
Manual Like-For-Like Forecasting Is Costly
Even if it's hard to see on a P&L, the cost is real. Manually identifying an analog item and hand-adjusting a forecast is work that gets redone every season, for every low-history SKU, with no system memory of the comparison you already worked out.
Without a structured link to a comparable item's demand pattern, new and replacement products often default to flat or category-average forecasts that don't reflect how a similar product actually sold.
That gap between forecast and reality shows up later as stockouts on the winners and markdowns on the misses. A wrong initial buy on a new item is expensive twice over: once if it under-forecasts a hit, again if it over-forecasts a dud.
New and replacement items carry outsized risk precisely because they have no historical safety net to catch a bad guess.
When Analog Logic Lives is Institutional Knowledge
There's also a quieter cost: consistency. When the "similar item" logic lives in one planner's head or in an offline spreadsheet, it doesn't transfer.
Different planners make different judgment calls about what counts as similar enough, so forecasts for comparable products end up inconsistent across categories or teams, and nobody can audit why.
How to Make Like-For-Like Forecasting Repeatable
More data won’t fix the issue. You can't manufacture sales history for a product that hasn't shipped yet. The fix is a structured way to borrow relevant history from somewhere else when your own item doesn't have any.
Formalize the Comparison
Instead of an ad hoc mental note that a new style is "kind of like" one from last year, explicitly identify and record which existing item's demand pattern should stand in for the missing history. That single step turns a judgment call into something repeatable.
Feed the Forecast Engine the Right Input
The forecast for the new item should be generated directly from the linked item's actual historical demand curve, not a flat default and not a category average. Instead of guessing at a number, you're borrowing a real pattern from something comparable.
Keep the Two Records Independent
The comparison should inform the forecast without altering either item's actual sales history. You want to borrow the pattern, not merge the data. That distinction matters later, when you're trying to trust your historical data for other purposes.
Build in Flexibility to Revisit the Call
The best analog you picked early on might turn out to be wrong, or a better comparison might emerge once the product is in market. The process should make it easy to swap the reference item or drop the comparison entirely as real sales data starts coming in, without leaving the forecast permanently anchored to a stale link.
Make the Logic Visible and Repeatable
Whatever comparison a planner uses should be captured somewhere the whole team, and next season's planner, can see and reuse. It shouldn't live in one person's head or one person's spreadsheet.
How Toolio's Like-For-Like Forecast works
This is exactly the workflow Toolio's like for-like forecast is built around. Link a new, replacement, or low-history choice to a similar one, and the system generates an AI Forecast from that choice's actual sales pattern, without touching either item's historical data, and with the flexibility to relink or unlink whenever the comparison no longer fits.
