What is your take on this topic?
Adobe can optimize RLSA Bids only if RLSA audiences are setup on an adgroup level. In my eyes this makes the data too granular so you end up with very few data per adgroup per RLSA List. When testing it, it looks like Adobe makes very few adjustments - apparently lacking data. However, right RLSA Adjustment are crutial for success.
The upside of manuel adjustment bidding for RLSA is, that you can set those in AdWords on a campaign level, which gives you enough data to make a solid decision. This however leads to massive inaccuracies in Adobes Cost Predictions, as RSLA clicks take up a large share in my campaigns and Adobe therefore underestimates the actual spend as it does not take into account the higher cpc resulting froim my manual RLSA up-bidding.
So I am puzzled here. What is your approach to this
Hope you're doing well!
Please let me know if you have any further questions.
Thanks Shivani for your help,
I am still a bit puzzled about this.
You say that RLSA bid-Adjustment (when made manually "outside" of AMO) do not affect cost accurancy because forecast is calculated on bid-unit level.
However, what I have expereinced is that in Portfolios where clicks that highly affected by bid-modifiers, the costs accuracy in total as well as for single bid units is very inaccurate - meaning actual costs are significantly higher than forecasted.
For example, a bid unit ist set to a CPC of 1€.
RLSA Modifer (manually): +100%
A click of a customer that has this criteria ends up with a CPC of 3,46€.
So, in a campaign that expecially has a large portion of RLSA Traffic, but also is pushed by other modifers (that are bid manually and/or cannot be bid by Adcloud) the actual cpc is way higher than the cpc set for the keyword.
When I have a look at the historical forecast of the bid units I do not see that adcloud takes this into account and does not "learn" that costs end up higher.
That's fine, adding bid modifiers certainly boost the cpcs for the respective entities to the specific limit & that is what they meant for!
Considering the case you've mentioned above, it doesn't looks like the manual versus auto RLSA scenario.
However, you should segregate the RLSA and non RLSA campaigns in separate portfolios - this would help you to view the segmented performance and you can easily depict the type of campaigns generating inaccuracies.
Moreover, if you are facing inaccuracies in the predicted & actual cost, it must be reflecting in your portfolio spotlight under model accuracy for the selected date range. To investigate the root cause of this issue, you should further drill down to the most affected bid unit & then look into the device, match type or keyword causing this inflation in the cost.
This should definitely help you to rectify the inaccuracies. Do let me know if you still face any issues!
Thank you Shivani,
I dont see how segregation in different portfolios can be an option, as all campaigns have RLSA audiences.
I observe the cost model inaccuracies on a bid unit level. And I come to the same result: For bid units, that have a large share of RLSA exposure (and therefore significantly higher effektive CPCs than maxCPC) and/or are effected by bid-adjustement for demographic or time, it results in predicting effective CPC that are way too low.
So to me it looks like if bid-modifier are set manually, AdClouds is unable to take this into consideration when predicting actual costs. Otherwise the system would have learned that, lets says a max-cpc bid of 1€ leads to a cpc of 1,50 €. But i do not see this reflected in the predictions.
If your portfolio has only RLSA campaigns, would you mind explaining how do you identify which bid units have a large share of RLSA exposure? The reason I am asking this is because, RLSA is a targeting method wherein all entities falling under a particular RLSA target would have equal exposure and applying bid modifiers boost/reduce any particular target(including the associated entities).
However if you're facing CPC inflation issues, you can always restrict it to a minimum & maximum limit using 'constraints'.
I am looking at the segmentation for audience on an adgroup level and see there how many clicks are associated with RLSA audience bids. As the adgoup setup is rather granular, you can tell that the keyword/bid-unit is affected accordingly.