This article has been authored and contributed by Manu Malhotra, Advertising Cloud Consultant - EMEA at Adobe.
Need help on Adobe Advertising Cloud? Join the Community
Adobe Advertising Cloud (previously Adobe Media Optimizer, or AMO) technology predicts performance based on data models it builds over a period of time. However, at times, it is very challenging to get the models accurate. When models go off because of various factors changing in the eco system, it results in uncertain performance. Hence, it is very important to keep a close eye on model accuracy.
Model accuracy shows how precise the cost and revenue models are which are being used to optimize bids. In simple terms, model accuracy is a comparison of predicted performance and the actual performance. You can check model accuracy by clicking on Portfolio Cards > View. Day on day break down of accuracies is listed in the table.
The model accuracy reports for multiple portfolios can also be easily accessed in Advertising Cloud by clicking on Reports > Create Report > Model Accuracy > Forecast Accuracy
If model accuracy is off, it simply means Advertising Cloud doesn’t have as much information as is required and hence the bidding on the bid units is not resulting in expected impactful returns.
Too many changes - Following a large scale account audit, check if you have changed a large number of landing pages / ad copies. In such a case it might result in cost model inaccuracy. It should take some time for quality score to build again.
Nature of the business – Check if there is a case of seasonality or a big event such as Olympics or change in competition such as a new entrant or aggressive strategy by competition. This can bring in cost model inaccuracy. Though Advertising Cloud will adjust and adapt to changes but it may take time. In such cases try reducing the cost half-life to make Advertising Cloud take more consideration of recent change in eco system.
New Additions – If new campaigns and keywords are getting added, this will impact cost models. In such cases, you might like to add the new campaigns in active state initially in a separate portfolio. Else, wait for some days and Advertising Cloud cost models will stabilize.
Significant changes in settings – Any changes in campaign settings such as geo targeting or match type strategy at search engine level can result in cost model inaccuracies too. You can either wait for next sync cycle for Advertising Cloud to fetch the changes from search engine or you can make changes directly from Advertising Cloud so that the technology is very much aware of the changes.
Significant changes in budget - This is a tricky scenario, try sticking to small and incremental changes in budgets. However, if you do have to make drastic changes because of business requirements, be ready for cost model inaccuracies. To handle it, try reducing the half-life.
Changes beyond your control - Bing updates its search algorithm, google decides to do away with right side ads, what do you do? Try reducing the half-life and technology should well adapt swiftly to the change.
Special promotions: The rate of conversions will see a jump and models might take some time to adapt and learn from the changes. It is suggested to reduce the half-life for models to adapt and learn quickly.
Market competition: In case of competition getting aggressive or new entrants bidding high results may differ from forecasted. In such a case resort to reducing revenue half-life for models to catch up.
Lag in revenue numbers from feeds: This can happen for operational as well as other business reasons. In such a case try increasing revenue half-life to accommodate the lag in reporting revenue.
Apart from these scenarios, at times there are pixel tracking codes issues where in pixels aren’t firing for some reason on the web properties. It is best recommended to highlight such an issue to Advertising Cloud and the client team to troubleshoot it.