On-Line Campaigns Management

On-Line Campaigns Management

Until now, there has been no reliable way to predict if a campaign is going to produce a positive ROI without a few days of activity. This inability to target the most successive campaigns prior to a time span of a few days leads to an un-optimal spending of resources that could be placed into the more profitable campaigns.

With AlgoTrace’s automated prediction solution you will be able to decide in which campaign to invest your efforts and resources, to achieve the best ROI. Our solution will predict the success or failure of each campaign prior to launching. This new ability will give you the opportunity to both maximize profits and avoid costly mistakes.

Ability to know each campaign success rate before it launch
On demand analysis of all current campaigns
A list of all relevant success Influencers
A What-If wizard for simulation
A dashboard for relevant data inquiries


You will get a pre-made data set that is based on AlgoTrace knowledge gained through conducting multiple successful campaign management models. This will allow you to jump start your own project and the start to finish time will be reduced to only a few days.


Present Situation
Every time a company launches a digital campaign, the numbers detailing the performance of the campaign start to roll in. There are dozens of KPIs and other metrics that are important for understanding the performances of each campaign but only a few of those metrics can really help in predicting the campaign’s future success.
A campaign’s success depends upon predicting and detecting which campaigns are likely to be profitable, based on a specific metric for success set by the operator. For example, the campaign’s operator could consider a campaign that has a profit of 30% and a reach out of at least 5,000 clicks to be a success.

The AlgoTrace Solution – Over 90% prediction accuracy
We consider ROI prediction as a binary classification task.  By using AlgoTrace’s automatic prediction engine, the campaign operator can get a prediction of the likelihood of success for each campaign before it is actually started.
We use historical data as an input for predicting the likelihood of meeting a positive ROI as a metric of success. The output is the answer to the question, ‘will the campaign be profitable or not?’.  We outline below the process for predicting the success of each individual campaign.

3 Steps to Predict Campaign Success in 3 steps:

1.Gather data on your past campaigns
2.Automatically create a “best-of-breed predictive model”
3.Use the model in production environment

STEP 1: Gather Information

AlgoTrace’s implementation team will send you a specific set of features that are directly related to campaign success prediction.  You can use the preliminary features set or you can use any other features that you consider more suitable. The objective is to feed the relevant data into the software which will then use that data to predict campaign’s success. In order to conduct a prediction project, you would need either a few months data or at least 200 previous campaigns.  You then need a specific column – the  “Target Variable” column which would indicate if a campaign was a success or not (“1″ – success,”0” – failure).

Relevant features example:

1. Campaign Type
2. Banner Size
3. Is Interstitial
4. Device Type

The data that contains the selected features and the target variable should be sliced into two files. The first file will contain 80% of the rows and is the “training file”. On this “training file” the algorithm engine will learn the hidden relationships between the success state of the campaign and the rest of the campaign’s parameters in order to predict campaign success for new campaigns. The second file contains the remaining 20% of rows. This will be the “test file” and the results of the prediction will be tested on it.

STEP 2: Make a Prediction

Now it’s the fun part:  AlgoTrace’s software automatically performs all the necessary steps to get the best prediction results. The prediction engine transforms and evaluates hundreds of options for each model until it reaches a final list of models and presents them to the user. Each model comes with all relevant accuracy and stability metrics.
All that is left to do for the user is to pick the model that interests him the most and prepare it for use on fresh data. Fresh data can be manually uploaded to the software or it can be used within a production environment.

STEP 3: Moving to Production

There are three main options for using the software:   The first option is to use it as a screening tool that presents the user with the best solution for event classification. The user can then draw managerial and technical insights out of the model’s explanatory information.
The second option is to use the model in on-demand mode. In this mode, the model is saved and once in a specified period of time ( a day, a week, a month., etc.) the user feeds the model with new raw data. The model produces a new prediction for each ready to be launched campaign.
The third option is to use the desired model in a production environment. There is a formula for specific models that can be copied and pasted into an SQL or other database related engine. In this case, the model can run at any time interval and use whatever specific parameters the user wishes. For example, the model will run once a week for a specific country and once every two weeks for another country.


Media buyers and Ads analysts can use AlgoTrace’s prediction solution to predict the success potential of each campaign they intend to launch.
Based on historical data, the prediction engine can be used for campaigns portfolio prediction or to get marketing insights based on different campaign attributes. The implementation time of predictive model can be dramatically reduced by using pre-set features that are provided by AlgoTrace’s implementation team.

You can have an end-to-end prediction solution for predicting campaign success in a matter of days.
You can reach us at info@algotrace.com