Cracking The Code: An Approach To Offline Marketing Attribution Will Flaherty VP, Growth Marketing | @flahertyiv
Cracking The Code: An Approach To Offline Marketing Attribution
Will Flaherty VP, Growth Marketing | @flahertyiv
Discover events Pick a perfect seat Buy with two taps Enter with the app
SeatGeek is a mobile-centric ticket marketplace
Offline marketing offers immense promise…..
Tremendous scale combined with unique media arbitrage opportunities1
2 Offline provides an opportunity to reach consumers who can’t be reached via digital
3 Offline mediums offer a broader canvas to communicate USPs or tell a brand story
Channel Example: Direct Mail
…yet also presents numerous challenges, all tied back to measurability
Offline media is difficult to track — and nearly impossible to do so at the granularity offered on digital platforms1
2 Additionally, offline media buys generally require larger time and monetary commitments than online campaigns
3 Poor campaign insights + long feedback loop + $$$ = elevated risk of failure
There’s no “Silver Bullet” for offline attribution
We utilize many methods in concert with one another to measure and gauge offline performance, as each have distinct strengths and weaknesses
We look at lift from offline advertising in markets we advertise. However, this rarely provides a clean view because of seasonality, team performance, and company growth. This method is also ill-suited to measure the long-term effects of offline marketing.
Market lift analysis2
We run brand awareness surveys (i.e. which of the following ticketing companies are you familiar with?). Results help us understand awareness driven by geographically targeted offline campaigns, but they can be expensive and can struggle to provide great granularity into return on ad spend.
Brand awareness1
Unique promocodes can be used by customers at account creation and checkout in order to convey an offer and also collect highly granular data on campaign response. However, many users neglect to enter their promo codes or forget them, meaning that they frequently underreport actual performance.
Promocodes3
Users are asked after making a purchase how they heard about SeatGeek either via e-mail or modal on site or in app, with follow up questions designed to collect more granular attribution data. Not as granular as promocodes, and ROAS calculations rely on user behavior assumptions that may be hard to validate.
Post-transactional survey4
Promo code redemption• A few of our creative units contained promocodes, but
redemption rates of the associated code were low, implying only modest Return on Ad Spend (ROAS)
Our NYC subway campaigns illustrated the limitations of many of these methods…
Overview• The NYC subway has become an advertising
destination for many startups. The scale is enormous; over six million people ride the subway daily.
• NYC Subway offers a unique opportunity to run multiple creatives in a takeover of a single car
Brand awareness• We saw an uptick in New York City aided brand awareness,
but our survey sample size was too small and our questions too generic to attribute performance to the subway campaign
Market lift• With many other marketing activities ongoing in New
York, It was difficult to identify any conclusive lift that we could attribute to the Subway campaign
…but also underscored the immense utility of a post-transaction survey
Post-transactional survey• We launched a post-transaction survey in preparation
for the New York City subway campaign
• Every first-time NYC buyer received an email asking where they heard about SeatGeek
• We began the test a week before the subway campaign launched to measure how many users incorrectly attributed the subway to their purchase. Only 1% of people chose an invalid source
Subway results• 24% of all new users responded to our post-
transactional email, with 20% crediting the subway ads for how they heard about SeatGeek
• We then projected ROAS based on survey results, with the resulting estimated return much higher than implied by promocode redemptions
Pre-Subway
Post-Subway
25% 50% 75% 100%
TV Online Radio / Podcast Friend Mail Subway
20%
1%
Survey Results Before and After Campaign Launch
ROAS Calculation Methodology
The scale-up adjustment yielded an estimated
Deep Dive: How we use post-transactional surveys to calculate subway return
3 Calculate ROAS using scaled salesAfter calculating the scaled up revenue, we divided that figure by the cost of subway media to yield our final ROAS estimate
Determine scale-up factorJust under 25% of all new NYC purchasers responded to the emailed survey. So we applied a scale up factor of 4.21 to survey-attributable subway sales to estimate those numbers for all new NYC purchasers.
2
ID survey-attributable subway salesWe first identified the user IDs of all new users who attributed Subway as how they heard about SeatGeek, and then pulled all purchases made by those users during the period
1
Best practices for post-transactional surveys
Administer survey soon after a “key event” occurs in your service’s user experience1
Design survey to make data input of the key channel-level questions as seamless as possible2
Ensure that you can tie survey results to trackable actions or other metadata (e.g., location) to allow for deeper analysis later down the line
3
Attribution survey in DoorDash Mobile App
Be aware of scale-up assumptions — unlikely that all your users behave exactly as the pool of surveyed users do4
Optimize for high response rates to minimize margin of error in analysis when scaling up results5