Crucial Role of Cost Models: Managing Frequently Changing Budget

Managing a budget is no less than an art. It is a mixture of skill, patience and little creativity with an effective technology.

As managing budget is just not a one shot deal and requires an active monitoring and managing skills as well, technologically strong platform with its guidance can help drive the management in right direction. This is exactly where Adobe Media Optimizer (AMO) and its technology comes into picture and helps in doing the job at its best. Beauty of AMO’s features is that it guides the management of budget in such a way that the conversions/objective would not get disturbed and actually maximized. 

AMO provides various options that can be used in managing day to day spend targets at portfolio level. Such as:

• Daily spend target
• Spend strategy
• Campaign caps and multiplier
• Auto adjust campaign budget limits
• Mobile bid adjustments
• ConstraintsLearning budgets
• Models – Cost Model and Revenue Models and their half life’s
• Campaign bids

In an account where the budget amount varies frequently every month and the history of traffic and conversions inflow has been strong, the ‘Cost Model’ feature of AMO helps a lot in effective budget management. Particularly the case when the budget has been reduced suddenly by big percentage and the history of previous month has been strong. The unexpected pulling interferes with AMO’s gathered information and takes time to get under control. This is where cost models provide the detail information at bid unit level. To recap, the bid unit in AMO at search level is actually a ‘keyword + match type’.

Modeling can be understood as forecasting (i.e. predicting performance). 

Based on gathered information, the AMO comes up with forecasts of how the bid unit can be expected to perform and based on the objective, it will optimize the bid for each bid unit, such that the objective is maximized¹. 

• Advertisers run their campaigns on search engine.
• Campaign’s click and cost data info. gets collected.
• Conversions data is also captured from client website in AMO.
• AMO then processes the data captured to come up with analysis, forecast in the form of models.
• To create models, AMO gathers information at various data points (bid vs cost, bid vs CPC, bid vs. clicks, bid vs impressions, bid vs. position etc.) and predictive modelling technique is used to analyze and generate data.

There are two types of models a) Cost Models b) Revenue Models. These reports and models can be seen on dashboard. Let’s discuss the role of cost model and its contribution below.

Cost Model Accuracy guides how the bid unit is performing as per its predicted performance.

The model can be accessed through ‘Portfolio cards’ tab and then selecting ‘Model accuracy’ option. It can be studied at following level:

a) Click volume level

b) Bid unit level

c) Device level

d) Mobile by ad group level

Close to 100% cost accuracy is the best figure to have in any account. If it is higher than +10 to +20% depending upon the use case basis, cost model should be studied in detail at all levels to understand from where the major cost is coming from. ‘Bid unit level’ is an important section to focus on in depth. It actually provides a detail on list of bid units where how much cost and click is incurred in actual performance than predicted value. This performance data then can further be analyzed where cost and clicks are higher than the predicted values. Once the list of bid units is shortlisted which are spending more than required then the next step is to study those bid unit’s search terms in detail. To do so, search terms at search engine or AMO level can be evaluated to ensure keywords are matching to right terms particularly of those keywords where our models are showing discrepancies. By setting those terms as ‘exact negatives’ and ‘cross positives exact’ at right ad group level help control the following metrics:

• Inaccurate impressions
• Irrelevant clicks
• Exact match types ratio
• CPC’s

The above action helps in immediate control of spend on irrelevant search terms. This helps not only in controlling the day to day spend but also further metrics such as CTR as it controls the unwanted impressions. CPC’s also get improved as exact match type increases and also the appearance of ads on relevant keywords only. As relevancy increases, it helps in improving the quality score further which in turn improves the CPC’s and avg. position. As a result of these basic metrics improvement, CR also gets better as the cost and the overall relevancy factor are getting controlled.

Therefore, the cost model accuracy information plays an important role in understanding the spend level path and helps taking the optimization action in right direction.

¹Definition taken from AMO training material from Gauri Bhat.