Find the levers that grow your revenue
Diamond Forecast Agent projects your Shopify sales for the next 52 weeks in one click, and measures which growth levers (discounts, social media and more) are moving it. Update your forecast without complex spreadsheets or dashboards, simply describe your plan in plain English.
A Prophet model, fitted per collection
We fit Meta's open-source Prophet model to each collection's weekly sales history and project 52 weeks ahead. We use Prophet because it is easy to understand, not a black box, and easy to customize. Prophet breaks revenue into parts you can read on their own and adds them together:
Forecast = Trend + Seasonality + Growth Lever Regressors. Collections with fewer than 12 weeks of sales history are excluded, because Prophet can't generate a valuable forecast from so little data.
Trend
g(t)Piecewise linear growth. Prophet places changepoints where the growth rate shifts (the dashed line) so a slow spring doesn't flatten the whole year.
Seasonality
s(t)A day-of-year pattern learned from past years, like a strong start to the year, a summer slowdown and a pickup in the fall.
Regressors
Σ β·x(t)Growth levers enter as extra regressors: discount %, whether social traffic from Instagram, Facebook or YouTube was active and generating orders, plus any custom lever you add.
Confidence interval
Each forecast carries a 80% band built from simulated trend changes and observation noise. It widens the further out you look.
Forecast
Cumulative within each calendar year · revenue in USD · all collections
Forecast 1 is the baseline: no discounts, no social media traffic generating orders. It shows where revenue is heading on trend and seasonality alone.
Forecast 2 adds your planned campaigns: discounts, social media and custom levers (e.g. newsletters).
Weekly lift per lever, per collection
A growth lever is anything you do that might change revenue. Because each lever is a regressor in the model, we can compare the forecast with the lever on and off and read the difference as its weekly revenue lift, with a confidence band around it. Like the trend and seasonality, each lever's effect is estimated from your store's historical data.
Discounts
DefaultDiscount % per collection per week, read from your order history. Lift is measured as e.g. 10% vs 0%.
Social media traffic
DefaultWhether traffic from Instagram, Facebook or YouTube was active and generating orders that week. Lift is active vs inactive.
Custom levers
OptionalNewsletter campaigns, affiliate campaigns, trade shows: any dated activity can become a lever. Ask in chat or attach a schedule.
Reading the chart
The tick is the estimated lift; the bar is its 80% band. A band clear of $0 (blue) is an effective lever. A band that crosses $0 (grey) means the data can't tell. These are associations, not proven cause and effect.
Growth levers
Weekly revenue lift estimates with 80% confidence band · association, not proven cause and effect.
Change the plan in plain English
Tell the agent what you're planning. It turns your message into new forecast parameters, re-runs the forecast with these, and replies with what changed and how the projection moved. Every change is logged and accessible in your session history.
Forecast 2 now includes a 10% discount during Black Friday week. Compared with the baseline Forecast 1, the discount adds an estimated $540 that week (80% band $310–$770).
Discounts
Add or modify discount values and schedules, for all collections or one. Example: “Assume 10% discount the week of Black Friday”
Social media
Add or modify social media conversions by week. Example: “Assume social media traffic is active and driving orders in December”
Custom levers
Add or modify any dated activity, past and future, by describing it or attaching a newsletter, trade-show or campaign schedule (CSV, XLSX or PDF). The agent backfills history so the model can learn the lever's effect. Example: “I've been sending newsletter on the first week of every month for the past year, and will continue until end of year. Include this in the forecast”
Questions or Feedback
Email us at support@diamond-analytics.com