Getting Started with TV & Radio

We recently brought in Mark Zamuner from TwoNil for Office Hours on “getting started with TV & Radio.” His presentation and notes from the session are embedded below. Mark, founder of TwoNil, was formerly VP of Customer Acquisition at eHarmony, and his agency now runs video and audio campaigns for companies like Uber, Zillow, and TripAdvisor.

Fundamentally, TwoNil’s approach to getting started with TV & radio is to run a campaign with a single piece of creative for ~6 weeks or more, and to use the campaign to determine one thing: whether these channel will work for you. If the channel works, you can then scale your spend, optimize and A/B test creative, etc. In Mark’s view, anything within a 2x striking distance of a profitable campaign means the channel can work.

Rough cost guidelines:
- TV campaign: $500K to $1M, which includes ~$150K set aside for developing solid creative. A local ad buy run ~$350K and a national buy is ~$750K. Recommended campaign length of 4-6 weeks.
- Radio: $150K to $500K (with a lower cost of creative.) Recommended campaign length is 10 weeks.

To determine whether to test radio/TV, you’ll obviously want to make sure that -- with reasonable assumptions -- you can get a good return. An example of Mark’s model is in the presentation below. The critical metric here is response rate: a benchmark for TV/audio response rate is 0.25% (below which you’re doing something wrong) to 0.75%. 1.0% is best in class. That, paired with your standard funnel metrics, will tell you whether TV/radio can theoretically work.



Getting Started with Radio & TV from FirstMark Capital

If you decide to take the leap, Mark suggests you’ll need to prepare two key things:

- The internal team that will own the entire end to end test. You may, for example, want a designated engineer or data scientist who can help get and process data in real time, particularly given the complexity around attribution.
- A multivariate model that will serve as the baseline during this test. Specifically, you will need a model that -- given assumptions around Facebook spend $X, search spend $Y, etc. -- can predict how many visits and conversions you’d expect.

Now that you’re ready to get started, you need to think about creative:

- Budget about 3-4 weeks for radio creative and 6-12 weeks for TV creative. There are plenty of agencies who can help with creative, and you can also produce it in house. Mark specifically offered No Slate/Committee LA, Stun Creative, and World War Seven as references.
- While creative development is the toughest piece of the puzzle, it’s also more art than science; as a result, you shouldn’t drive yourself crazy seeking perfection for the first campaign.
- Based on a study of clients, clear, repeated CTAs are critical to success (e.g., “visit mysite.com now”) and discounting has no effect on response rate.

With the creative in hand, you’re ready to make the buy. Agencies can help with this process, but you’re equally able to go direct. A few notes on targeting:
- National is the most cost efficient (often 4-8x less on a CPM basis), but the minimum spend is much higher
- Local is more expensive, but enables more discreet A/B testing. Although if you go local, Mark recommends bundling a few markets to prevent outliers (e.g., bad weather in a particular region drives response rates down.)
- Targeting is generally based on very general demographic research (e.g., “X% of our viewers are age 18-25”.) Buys can be done on a channel level, or against specific shows; some shows, unsurprisingly, have premium prices. The general rule, however, is to look for channels that have the highest density of your target audience.

With an ad running, attribution and ROI calculation is the final piece of the puzzle:

- TV spots are encoded with digital watermarks that enable tracking services to provide daily feeds of specific spot airing times.
- Once you know when your spot aired -- and with your existing baseline model -- you simply look at actions in the five-minute window after a spot airs. Any change from baseline can be attributed to your campaign.
- The key metric to calculate is response rate; that’s the single metric that will determine whether these channels work for you.

Finally, a few general tips:
- Response actually continue to rise with multiple ad exposures (a large part of why you should run a campaign for multiple weeks.) 
- Response rates generally follow an S curve - slow at first, picking up steam with multiple impressions, then tailing off as the campaign hits saturation.
- One attribution tool that Zamuner dismissed is promo codes. Only a small percentage of the audience uses the code, making them unreliable.