Welcome!
Welcome to the Marin Software Support Center. We're glad you're here. Here's what you can look forward to:
  • Real-time search so you can find what you need faster than ever.
  • Easy-to-follow video guides for our most popular articles.
  • Interactive simulations and Live Screenshots to make learning easy.
  • Regular content updates to ensure every word you read is accurate and up-to-date.

Bidding and Optimization

Home > English > Bidding and Optimization > Bidding Basics > Bid Strategy Forecasting

Bid Strategy Forecasting

Introduction

Clients using our bidding algorithm can also use our forecasting model to predict future trends in clicks, conversion, revenue, cost, and profit based on historical data. These predictions can be used as guidelines to adjust bidding goals based on current business needs. The trends are calculated using Marin's predictive keyword-level auction and volume models, which are then aggregated to present a Bid Strategy level graph.

Please note that the organizational structure used for bidding is referred to as Bid Strategies in MarinOne and Folders in Marin Search. In this article, we will be referring to them as Bid Strategies except where a user interface element may refer to them as Folders, in which case we will align the name with what's visible. 

 

bsf3.jpg

How To


In MarinOne  

  1. The first step, as always, is to log into MarinOne. 
     
  2. Then, click on the Optimization button ('light bulb' icon) in the bottom left-hand corner. 
     

    fc2.jpg
     

  3. Next, make certain the Bid Strategies sub-tab is selected in the top navigation. This should be the tab you land on by default. 
     

    fc3.jpg
     

  4. From here, click on the Forecasting toggle in the top-right of your grid. Click once to slide the toggle over to the right (it will turn green when enabled). 
     

    fc4.jpg


    Turning the toggle on will, by default, enable account-level forecasting, including Budget Pacing and Budget What-If features. To learn more about these tools, check out our dedicated article
     
  5. Before you review your forecasting, double-check your date range by clicking on the Calendar icon above the grid. You can choose between This Week, This Month, This Quarter, Next Month, and Next Quarter. 
     

    fc5.jpg
     

  6. To see forecasting for a single Bid Strategy, simply click the check-box next to the relevant Bid Strategy.
     

    fc6.jpg
     

  7. You can hover over the chart for more analysis or use the Target slider to the left of the chart to experiment with different outcomes. 
     

    bsf1.jpg


    The graph shows predicted changes in different key performance indicators based on changes to the Bid Strategy target. Note the numbers predicted by Forecasting are estimated average daily values for different metrics.
     
  8. To see the predictions, click on the graph to select a new bidding target (the target is on the x-axis of the graph). You can view trends in two metrics on the line chart and all the predictive metrics and their respective changes are also shown on the bar charts as you select a new target.

    The most recent four weeks of historical data is used to build forecasting models which predict the metrics shown. The historical data is time-weighted to treat recent data as more important.

     
  9. You can export the forecast chart to a dashboard if you wish, simply by saving it as a Saved View, then adding it to a dashboard. For more information about dashboards, check out our dedicated article
     
  10. If you see an outcome you like, simply click Apply Recommendations to target your new spend goal. Marin Bidding will adjust your Bid Strategy targets accordingly in order to reach your goal. 

In Marin Search  

When logged into Marin Search, click on the Bidding tab, click into a specific Folder, then click on the Settings tab. From here, scroll down to the forecasting chart. You will see a clickable image directing you to MarinOne -- click on this to launch forecasting in MarinOne. You can then follow the instructions outlined in the MarinOne section earlier in this article. 

Note: You may need to re-login once you launch MarinOne. 
 

fc1.jpg

Auction and Volume Models


Predictive Auction and Volume Models are built for both head and tail terms in each Bid Strategy.

Keyword-level models are built for all head terms with more than 28 clicks in the past 28 days, while cluster-based models are built for similar tail terms in a Bid Strategy with less than 28 clicks in the past 28 days. For both head and tail keywords, the Auction Model predicts the relationship between Bid and Cost per Click (CPC) based on historical Average Bid and Average CPC. It uses a half-logistic regression to model this relationship. The Volume Model predicts changes to expected daily clicks based on Cost per Click using a full-logistic regression. Cross-Keyword information is leveraged when keywords don't have enough movement to guide the parameter bounds. Using the keyword - and cluster - level Auction and Volume models, we can predict the change in expected clicks based on Bid values and expected cost.

Predicting Conversions and Revenue


The expected clicks are predicted for a given bid by the Auction and Volume models as described above. Using the conversion rate predicted for each keyword by our Bayesian based algorithm multiplied by the expected clicks we predict the conversions for each keyword in a Bid Strategy. Similarly the revenue for each keyword can be predicted by multiplying the revenue per conversion by the expected conversion.

Confidence Intervals


Confidence intervals can be displayed for all metrics on the forecasting chart. The interval shown is the estimated range where forecasting is 90% confident with the predicted performance metric.

Note: The Confidence intervals are only displayed if you are using the algorithmic bidding model. 

Last modified

Tags

Classifications

This page has no classifications.

 

wiki.page("Internal/Mindtouch_Launch_Sandbox/js.cookie.js")