Traditionally, TV advertising has been largely guesswork – we all know it works, but proving its impact is highly complex. You’ve also been told that knowing the actual impact of your TV ad campaign is far more difficult than knowing the impact of any digital campaign.
Measuring the performance of digital works because the advertiser can easily create a “baseline” through continuous A/B testing. At TVSquared, we approach TV the same way – continuously measuring against a baseline – which allows us to create an environment where we can calculate expected vs. observed response at the minute level. This approach allows us to literally see the amount of traffic that TV drives.
So How Do We Do It?
Essentially, the baseline process is about noise cancellation.
Imagine you throw a pebble into a lake. The closer you are to looking at where the pebble hit the water, the bigger the impact will be to you. The TV spot airing is the pebble hitting the water, and the ripples it causes are the calculated audience response.
If we know how still the lake is naturally, when the TV airing happens, we can go: “Ah, this is what we expected to see – but this is what we actually observed.” So, when there are other factors causing response spikes, you want to factor them out to know the true impact of the TV ad.
As with any impact evaluation, first, you need to understand the volume of website, call center or app traffic you would have experienced without the TV advertising. We call that the baseline. This is specifically non-TV driven activity a brand would have experienced if it did not advertise on TV.
- We dynamically model the amount of traffic generated over time (not using historical data)
- We discount non-TV inspired traffic
- We can now see the impact of the TV campaign
In other words, we use a sophisticated statistical approach called “frequency decomposition” to accurately calculate the baseline every minute of every day – which generates a continuous, dynamic baseline profile. Our approach means the baseline will automatically (and continuously) adjust for different volumes of traffic, discounting the impact of natural fluctuations throughout the day, including seasonality from day to day. It is unaffected by clashing spots (ads airing at the same time on different channels), periods of extreme change (i.e. late night and early morning) and variation in baseline levels.
Finally, it’s not reliant on historical data. Each day’s baseline calculation is started afresh using only data from that day, which means it’s more effective than historical modeling. Let me explain that now.
Why Historical “Look-Back” Methods Don’t Work
Many baseline approaches use data from just before each spot airs, and they use this to produce an average activity level for the period prior to and after the airing. This approach is fundamentally flawed for three reasons:
Why Our Approach is The Industry’s Gold-Standard
By calculating the baseline independently for every possible permutation of type of customer-response method, type of customer action and country or DMA (Designated Marketing Area), our approach gives the most accurate insights into the impact of TV.
Contact us to learn more about TVSquared and our baseline.