Use retargeting from Google Analytics 360 segments to drive first time orders from new customers. CPA goal of $350 per order from a new user, with credit given only to true last click via the client’s own attribution model built in GA360.
Integrating GA360 segments to DBM, Sensitive category distinction by DBM, Driving last-click conversions given the client’s restrictive attribution model; CPA tracking in DBM (via DCM conversion tag) attributed credit to all view and click conversions
GA segments took 2-3 days to cookie sync with DBM, so scaling of campaign was slower than expected. We implemented some prospecting targeting on day 2 to increase spend (leveraging 3rd party data segments and some highly relevant keywords). This was turned off on day 3 after client feedback; day 2 ended up having the highest CPA (all conversions) of the campaign at $175. Once we were just serving towards the full retargeting audience, CPA lowered to ~$50 range over days 4-6.
Initial segments built out of GA focused on site categories (ex: Shop Plans, Success Stories, View Menu), along with targeting all visitors and new visitors for scale. All line items blocked visitors that had already purchased, signed into an existing account, or viewing career pages. Recency bidding was introduced on day 7, which increased bids and frequency for the most recent visitors and decreased over time. Line items were created to target all visitors + food categories, all visitors + top performing states (using historic GA revenue data). Campaign was also initially set to only target desktop, as conversion rate for mobile was low.
Initial Optimization Results
Site categories performed slightly above the campaign average and were on the entirety of the campaign. Recency bidding line items were generally successful, the 30min, 6hour and 24hour segments in particular. Layering food categories and top historic states did not produce above average CPAs and were eventually paused.
DBM flagged the creative for sensitive category – weight loss on day 7. This was noticed by the Specialist during the morning campaign check and Google team was immediately alerted. The campaign basically didn’t serve day 7, as review took 36 hours to complete on Google end internally; campaign restarted in the afternoon of day 8.
CPA bidding, with a target of $40, was added on day 13 to all ad groups as we had over 200 total conversion points for the system to optimize to. We did test the all visitors line item with CPA bidding on day 9 (Thanksgiving), which lowered that line items CPA to $13 over the long weekend. Mobile was added on day 14 in an attempt to increase clicks to the site. Mobile specific ad groups were duplicated on day 15 to control bidding, which was turned from CPA to CPM.
Additional Optimization Results
Overall, the campaign’s best day was Day 14, when all ad groups were on CPA bidding and had mobile and desktop targeted within the same line items. Overall CPA that day was $27, but still didn’t produce any last click conversion in the GA attribution that client was using to judge success.
We did continually decrease CPA, while scaling against the retargeting audience. Campaign was actively managed but failed to produce the end results desired. Over 70% of click conversions were attributed to other advertising efforts.
For heavy direct-response campaigns, a much different set of assumptions need to be taken during initial campaign setup versus branding. We could have honed in more quickly on strategies, like recency targeting, which would increase frequency towards the most engaged audience, especially given the last-click goals. Not experimenting with mobile was probably a mistaken, even with the historic evidence of a lower conversion rate; we should have been more focused on cost-per analysis than rate analysis. An appropriate mobile bidding strategy, coupled with cross-device retargeting that the DBM/GA integration provides, would have given us more click opportunities to convert. Applying prospecting on day 2, in an effort to scale, was a mistake. We didn’t make use of auto-optimization tools like CPA bidding nearly fast enough either. Overall, we panicked at the beginning and end of the campaign, and should have been more patient with both the slow start and the optimization tactics that were showing traction once we were scaling.