A synergy exists when the joint effect of two activities differs from the sum of their separate effects.
A synergy exists when the joint effect of two activities differs from the sum of their separate effects. More conversions among people exposed to both is not enough: they may already have had greater intent. The minimum design is a factorial experiment with four comparable groups: no treatment, A only, B only and A+B. Interaction then becomes a calculated term with its interval and cost.
Decision and method.
Fund a combination for synergy only when the interaction is identified separately from own effects and remains economically positive after added costs. State why A should modify B's response—priming, proof, repetition, reach or recall—and define order, delay and population before testing. Size control, A, B and A+B over the same period with verifiable exposure rules; avoid contamination and measure adherence. Estimate A, B, total and interaction effects; report levels, differences, uncertainty and negative guardrail effects. Value incremental conversions, deduct activity and coordination costs, then test whether a simpler sequence retains most of the effect.
Worked example.
Four comparable groups produce 1,000 control conversions, 1,120 with paid alone, 1,070 with owned alone and 1,260 with both. Contribution per conversion is €45; paid costs €4,000, owned €1,000. Paid effect is 120, owned 70, joint effect 260; interaction is 260 − 120 − 70 = 70 conversions, or €3,150 contribution. The joint plan produces 260 × 45 − 5,000 = €6,700 net. Retain it only if the interaction interval excludes a material loss, and report the 70 separately rather than crediting both activities.
Checks and limits.
Pre-specify mechanism, order and interaction window; all four cells have a counterfactual and controlled exposure; calculate interaction separately from own effects; interaction contribution covers creation and coordination. Interaction depends on spend, creative and sequence. Contamination weakens identification. A two-activity test does not describe every interaction in a full portfolio.
Resources and sources.
Download the factorial dataset. Related: assign media roles, separate paid, owned and earned, measure incrementality. Sources: Naik and Raman (2003); Montgomery (2017); Lewis and Rao (2015).