We were able to measure how TV advertising affects on-site performance by modelling visits and conversions. We built and validated a statistical model, capable of predicting visits to the client site with an accuracy of over 92%. The model we employed uses TV impacts as a regressor variable, which means we can interpret the fitted coefficients to measure the effect of TV advertising on traffic.
We were also able to model on-site conversions to a similar degree of accuracy. This meant we could successfully determine an estimated conversion rate for the traffic induced by TV advertising, and ultimately deliver a CPA.
It was crucial to consider the lag effect of TV. We measured this effect, which we found remained statistically significant up to -4 days.
A big push in TV advertising would continue to affect site visits well beyond the airdate. The effect was measured to decay in line with expectation.
The seasonality of both site visits and TV advertising impacts were highly correlated, making it very difficult to isolate the effect of TV. Using some statistical techniques, we were able to achieve this isolation.
Forward3D’s channel-neutral data science team conducted the analysis using statistical computing language, R. The analysis required the use of sophisticated time-series techniques and methodologies. We used 3 years of historical performance data segmented by day, with TV impact data by day and region for the same period.
TV advertising is measured to have driven 37.9% of all visits to site over the past 3 years. Conversion rate of these visitors was measured to be 5.14%.
Over the same date range, TV advertising delivered 406,561 conversions at a CPA of £19.46, which is slightly more expensive than generic PPC, but at a hugely increased volume.