How Smart Model reads your sales history
Smart Model starts by looking at your last 12 months of sales (this lookback period can be adjusted per store). For every SKU, in every warehouse, it calculates your average daily sales and how that demand splits across the different SKUs within the same product.
If a SKU hasn't been around for the full lookback window, Smart Model only counts the days it was actually available to sell — so a recent launch isn't penalized for "missing" sales on days it didn't exist yet.
Example: a SKU launched 30 days ago that sold 60 units is treated as selling 2 units a day, not 60 units spread thinly across a full year.
Learn hands-on steps of setting up a plan for a newly launched product on Shopify
How seasonality is detected
Most businesses have a rhythm to their sales - holidays, summer/winter swings, a spike around Black Friday.
Smart Model looks for these patterns automatically, in two layers.
Store-wide seasonality
This looks at your entire store's sales across the year to find the big, repeating patterns, and applies an adjustment to your forecast for the months that need it.
Example: if November typically brings in 40% more revenue than an average month, forecasts for November get scaled up to reflect that.
Category-level seasonality
If your store sells a mix of products that behave very differently across the year, think swimwear versus winter coats, a single store-wide pattern can flatten out real differences.
For stores with at least 2 years of sales history, Smart Model can detect these different rhythms and group similar categories together (swimwear and sunscreen might peak together in summer, coats and boots in winter), giving each group its own seasonal pattern instead of one blanket curve for everything.
This only kicks in when the pattern in your data is clear enough to be reliable.
If it isn't, Smart Model falls back to the simpler store-wide seasonality instead, so your forecast is never built on a shaky signal.
This is also part of why the confidence tag on your plan needs 2+ years of history to reach "High."
How your growth target shapes the forecast
Alongside your sales history, you tell Prediko your expected revenue for the next 12 months - your Revene Expectation.
Smart Model treats your historical patterns as the starting shape of the forecast, then scales that shape to reach your target.
If you're planning for growth, that increase is applied gradually, month over month, rather than dropped in all at once. If you're planning for a decline, Smart Model applies a simple, even scaling instead, so the forecast doesn't produce an unrealistic cliff.
You can set and edit this number yourself at any time.
Smart Model never changes it for you. See how to edit a Smart Model?
Blending in this month's actual sales
For the month you're currently in, Smart Model doesn't forecast over days that have already happened.
Whatever you've actually sold so far this month is used as-is, and only the remaining days of the month are forecast.
This keeps the current month grounded in what's actually happening in your store, rather than a prediction made at the start of the month.
Choosing how sales are spread across the year
Once your 12-month total is set, you can choose how granular the distribution of that total should be.
Option | What it does | Best for |
Whole Year | One consistent pattern applied across all 12 months | Steady businesses without strong month-to-month swings |
Quarterly | A different pattern for each of Q1, Q2, Q3, Q4 | Businesses with clear seasonal quarters |
Monthly | A unique distribution for each individual month | Catalogs with specific month-level rhythms, like a known slow January or a mid-year product launch |
You can change this at the store level in your plan settings.
When there isn't enough history yet
Smart Model needs a meaningful amount of sales history to detect patterns reliably.
If your store doesn't have enough yet, Prediko automatically falls back to the Simple Model which is a simpler, average-based forecast with no seasonality or growth curve applied, scaled to match your expected revenue.
What happens when you Refresh Plan
When you refresh your Smart Model forecast (Type of refresh is Manual):
Historical data is frozen - Everything up to today stays exactly as it happened
Only future days are recalculated, based on the latest sales data available
Draft and Archived products are excluded from the regeneration
The updated forecast is checked against your expected revenue target, so the total still adds up
If you only regenerate specific SKUs, the rest of your plan is left untouched
FAQs
Why does my forecast show sales on days I historically had none?
Smart Model predicts expected demand, not an exact repeat of history.
It's averaging patterns across weeks and months, and factoring in your growth target, so a day that happened to have zero sales last year (maybe due to a stockout, a paused campaign, or plain randomness) can still show forecasted demand if the underlying pattern supports it.
What happens to my forecast if I deactivate a warehouse?
Prediko redistributes it automatically. For SKUs that also exist in other warehouses, that demand is added proportionally to your active locations. For SKUs that only existed in the deactivated warehouse, they're reassigned based on your overall sales split. Historical sales already recorded in that warehouse stay exactly as they were.
How do I add category-level seasonality to my plan?
It's on by default, and Smart Model only applies it where your data actually supports it. It's worth keeping on if you have 2+ years of history and genuinely different-behaving product categories. If your catalog is fairly uniform, or you're still building up sales history, it won't make much difference either way — Smart Model will simply lean on store-wide seasonality until there's enough signal to do more.