The definitive guide to maximizing campaign responses
Automate Your Marketing
Most marketing campaigns today, even the ones targeted at existing customers, get only about a 1% response.
What is even worse is that some marketers treat this 1% response and 10% conversion from those responses as cause for celebration, ignoring the fact that their campaigns are rejected by 99.9% of their customers.
Most marketers find a 1% response rate on marketing campaigns targeted at prospects, acceptable. But that is because they know very little about prospects.
On the other hand, getting that kind of responses on marketing campaigns targeted at existing customers is really bad considering you know a lot about them.
Unfortunately, even if businesses better responses to marketing campaigns, they do not get any support from today’s marketing automation platforms.
Most platforms automatically deliver emails on the day and time that the customer is most likely to access emails.
Sadly, it wouldn’t suffice as a solution for growing concerns on maximizing responses on marketing campaigns. Because,
“Customers responding to your campaigns is a collective function of customer behaviour, demographics, communication preferences, incentive and creative”
In this article, we discuss a three step process to maximize responses to your campaigns.
- Capture key factors that influence campaign responses
- Build response maximization models
- Sustain the efficacy of response maximization models
To optimize campaign responses, you require significant amount of campaign data.
To generate the required data, you need to regularly run campaigns, heuristically.
Greater the number of campaigns you run, greater the possibility of generating the required amount of data for response maximization.
Alert: The first and foremost step in maximizing campaign responses is to ensure that your campaigns are highly relevant to your customers. To ensure relevance, plan campaigns based on your customer behavior and not business pressure while sharply defining segments for each campaign. Unless you ensure these, even if you religiously follow the recommendations in this article, you cannot maximize campaign responses beyond a point.
Step 1: Capture key factors that influence campaign responses
If you want to maximize customer responses to your campaigns, you need to maintain a contact and response history, in which the following information needs to be captured about each customer.
Segment Targeted:
This refers to the segment in which the customer falls for the specific campaign.
This is important because same customers can fall into multiple segments and multiple campaigns could be planned targeting the same segments.
Having this information will help determine which segment performs well for which campaign.
Marketing Action:
This refers to the campaign’s intended marketing action for the targeted segment.
For instance, cross-sell product A, increase usage of product B, offer next variant of product C and so on.
This information tells you the most effective marketing action for each customer segment.
Incentive / Offer:
This refers to the offer that you make the customer for taking you up on a recommendation.
For instance, 3X reward points on a new credit card sign-up, 25% off on the first transaction, etc.
This tells you the offer that incentivizes customers most for the targeted action.
Number of Exposures:
This refers to the first, second, third or nth exposure of the same campaign that is being sent to the customer.
This is important as we will know the number of times a customer sees a campaign before he / she usually responds to it.
Communication Channel:
This refers to the communication channel used to relay the specific exposure of the campaign.
Regardless of whether a customer responds to a campaign or not, it is important to log the channel used. .
This helps determine the channel preferred by each customer or the one that elicits the maximum responses.
Campaign relay week, day and time:
While many conventional email marketing / marketing automation platforms have this feature built in, they track week, day and time only for emails which isn’t sufficient.
You should capture these details for every campaign, regardless of the channel.
This would help determine when a customer is most likely to respond to a campaign.
Creative variant:
We are sure you would have heard about or even do AB Testing / multi-variate testing on your marketing portal.
You need to apply the same principles on your marketing campaigns where you need to test different variants of your creative such as different subject line, body, images, call-to-action etc. to see which one is effective.
You need to ensure you are using variants across all channels and not just email.
Even on channels like SMS, website and Voice, you could try completely different variants. Whereas in other channels such as your emails and push notifications, you can try different combinations of units.
Logging this information would help determine which variant or combination of creative variant can be used to maximize campaign responses.
Response Status:
This refers to whether the customer has responded to your campaign or not.
This is the most important factor to be captured as part of contact and response history as this helps in determining the efficacy of each of the above parameters.
Here’s an under-the-hood view of the customer and response history that is used by Acquigo to optimize customer responses to campaigns.
Step 2: Building response models using captured details
Once you collect the required data, the next step is to build statistical models that will find the best combination of each of the above parameters so that you can follow and implement those on your campaigns.
Remember, you may need to build different models for different parameters.
Before building the model, you need to ensure the data is of good quality by ensuring that your data doesn’t have missing values or outliers.
If present, those need to be addressed first by processing the contact and response history data, doing a missing value analysis, outlier treatment and balancing training data before training the models.
Once you process the data, the next step is to split the data into training data and testing data.
The next step is to build the actual response optimization model, where you run different algorithms on the training data.
After this, you need to test the accuracy of each of the algorithms by running the same algorithms on testing data.
The model that delivers the greatest accuracy should be used to find the correct value or range for each of the factors that influence campaign responses.
Pro-tip for the scaling up of model performances:
Over time, your models could become heavy and may start encountering performance issues.
To avoid this, we recommend using in-memory, distributed, fast, and scalable machine learning and predictive analytics modules to run response maximization models.
The machine learning workflow would require an automatic training and tuning of multiple models simultaneously.
To make this seamless, you need to create stacked ensemble models, using a process called stacking, that could find the optimal combination of a collection of prediction algorithms.
The performance of stacked ensemble models is generally far superior to individual models.
Step 3: Sustaining efficacy of response maximization models
The efficiency of machine learning models tends to deteriorate over time.
Hence you need to adopt a champion-challenger approach to modeling as follows.
- Once every quarter or once every 6 months, the whole model building workflow should be automatically refreshed with updated data.
- You need to pit the new best models (Challengers) against existing models (Champions) and compare the performance
- The better performing models should be deployed as the new Champion
- This sequence of activities should be done regularly.
Moreover, to ensure the response optimization models give importance to significant factors, you could leverage the power of Artificial Intelligence.
Towards this, you may use algorithms such as SHapley Additive exPlanation (SHAP) that could give you clarity on why your response maximization model draws its conclusion for a particular record and which variables are considered significant.
Here’s a visual representation of how the algorithm would provide an explanation for recommendations made by the response optimization model in the form of weightages assigned for each response influencing parameter.
Summing up
While maximizing responses to campaigns might look like its rocket science, the reality is, that it isn’t.
We have been doing this for over a decade for global brands as part of our consulting services, and we have embedded those learnings into Acquigo - so every aspect of response maximization is automated.
Even if you haven’t bothered maximizing your campaign responses until now and want to start fresh today, you could, if you had statistical modelling expertise.
To start with, you need to keep running campaigns and diligently collect contact and response data such that it is statistically enough to train and test models.
Once you build models and find the right set of algorithms to determine the best combination of response maximization parameters, you will start seeing your campaign responses go up.
And, when you keep enriching the contact and response data every quarter or bi-annually, you will see the effectiveness of models increase as evidenced by growing campaign responses – by even up to 40%.
Copyright © 2023. Powered by Xerago. All Rights Reserved.