What the two curves show

Historical Model

The Historical Model shows how the campaign responded to different levels of average daily spend during the periods when it was active.

It already reflects the mix of weekdays, months, holidays, promotions, and other conditions that occurred while the campaign was running. It is not a “seasonality off” curve.

Seasonal Forecast

The Seasonal Forecast shows how that campaign’s response may shift across the upcoming 28-day calendar.

The switch reveals a forward-looking curve that has already been calculated. It does not simply turn one seasonal factor on while leaving everything else unchanged.

What the forecast considers

Depending on the available data and selected models, the forward curve can reflect:

• recurring weekly, monthly, and yearly demand patterns;
• holidays and other calendar events;
• the periods in which the campaign historically ran;
• whether modeled performance has recently strengthened, weakened, or remained stable;
• diminishing returns as spend increases;
• delayed impact from earlier marketing spend;
• supported halo or cross-channel effects;
• alignment with historical modeled attribution; and
• contributions across selected outcomes such as ecommerce, Amazon, or TikTok Shop.

Why campaigns from the same company can look different

Campaigns do not all translate customer demand into outcomes in the same way.

Two campaigns may have different forecasts because they:

• ran during different periods;
• belong to different channels or tactics;
• have different amounts of supporting history;
• reach diminishing returns at different spend levels;
• have different delayed or halo effects; or
• contribute differently across selected sales outcomes.

Campaign-level differences can therefore be reasonable even when the campaigns share the same company-level demand patterns.

How to read the comparison

• Forward curve above historical: the model expects a stronger campaign response at that spend level during the upcoming window.
• Forward curve below historical: the model expects a weaker response than the campaign’s historical active-period baseline.
• Curves close together: the model expects the upcoming response to be broadly similar to the contexts in which the campaign historically operated.

The distance between the curves should not be translated directly into a “seasonal percentage.” It represents the combined forward-looking effect of upcoming demand and learned campaign behavior.

How marketers should use it

Use the forecast as a planning signal to:

• anticipate periods of potentially stronger or weaker campaign response;
• explore expected outcomes at different average daily spend levels;
• identify campaigns that deserve closer budget review; and
• support planning conversations alongside marginal efficiency and historical confidence.

Also consider promotions, inventory, creative changes, campaign objectives, and other business knowledge that may not yet be reflected in the model.

What it is not

The Seasonal Forecast is not:

• a guarantee of future performance;
• a forecast of total company revenue;
• proof that seasonality alone caused the difference between the curves;
• an automatic instruction to increase or decrease spend; or
• a replacement for marketer judgment.

Recommended CSM talk track

The Historical Model shows how the campaign responded during the periods when it was active. The Seasonal Forecast shows how that response may shift over the next 28 days. Upcoming seasonal demand is an important input, but the forecast also reflects what the model learned about this campaign’s recent direction, diminishing returns, delayed effects, and contribution across selected sales outcomes. That is why two campaigns for the same client can have different forward curves.

When a customer asks, “Is this just seasonality?”

No. The upcoming calendar is what makes the curve forward-looking, but the model keeps the campaign’s learned response behavior, carryover, halo effects, calibration, and selected outcome contributions. The result is the campaign’s complete modeled response for the upcoming window.

When the lines are close

The model expects this campaign’s upcoming response to be similar to the contexts in which it historically operated. It does not mean the company has no seasonality.

When the lines are far apart

The model expects a materially different response in the upcoming window. Seasonality may contribute, but the difference can also reflect recent modeled direction, campaign history, diminishing returns, delayed effects, and the selected outcome mix.

Language guide

Say

• “Forward-looking campaign response”
• “Expected response over the next 28 days”
• “Upcoming seasonal demand and learned campaign behavior”
• “Seasonality is one of several inputs”
• “Directional planning signal”
• “Historical active-period baseline”

Avoid

• “This campaign has 50% more seasonality”
• “Seasonality will increase revenue by 50%”
• “The historical line has seasonality turned off”
• “The model guarantees this return”
• “The forecast says to increase budget”

Key takeaway

The Seasonal Forecast uses the upcoming calendar together with the complete campaign model. It is a campaign-specific planning signal, not an isolated measure of seasonality.