Jonathan Manalo

Data Engineer. Data Scientist. Data Analyst.

An example of my work with media mix modeling. The robust methodology behind this approach enabled us to achieve great success for multiple clients.

Background

Our client was struggling to understand the public’s perception of their brand, especially in regards to the impact of their advertising and marketing campaigns. We proposed a solution involving social media monitoring, media mix models, and analysis.

Questions

For the analysis, we aimed to specifically answer the following questions:

  1. What are the main factors that increase the negative conversation?
  2. What are the communications tactics most effective at driving down the negative conversation?
  3. How do the paid and owned efforts interact to drive down the negative conversation?
  4. What is the return on spend for driving down the negative conversation?
  5. How are our tactics performing at driving down the negative conversation YOY?

Methodology

To answer these questions, I performed each of the following tasks unless otherwise noted.

  • Guided the team in building a customized negative sentiment model using Natural Language Processing (NLP) techniques.
  • Built a data warehouse that captured client social media.
  • Used SQL and python to manipulate and transform the data.
  • Tagged the social media content using our negative sentiment model.
  • Leveraged EViews to create a media mix model using SARIMA methods.
  • Analyzed the data and modeling results.
  • Crafted the story.
  • Created the deck (select slides shown below) with conclusions and recommendations.

Results

When presenting, I like to use storytelling. I begin by comparing past and current strategies. image-center

I then summarize the changes in their KPI of interest, in this case, negative conversation. (Note, we want to suppress negative conversation, so lower is better.) image-center

Next, I dive into the specifics of the negative conversation trends, identifying key events that triggered significant fluctuations. image-center

Lastly, I provide how each media channel contributed to this overall impact. image-center

Analysis

For the results, I focus on significant discoveries, including those that were anticipated and any unexpected surprises.

As before, I summarize and compare the past to the present. image-center

Next, I provide how each media channel performed in supressing the negative conversation. image-center

Afterwards, I share the efficiency of media channel in terms of spend… image-center

…and per impression. image-center

To provide a clearer picture, I dive into specific trends and assess how each channel is performing in real time. image-center

For those interested in statistical details, I include factors, their impact, and confidence levels. However, the information remains accessible to a non-technical audience. image-center

Conclusions

Due to confidentiality, conclusions and recommendations are omitted. However, for all our clients, I have created slides to answer the main questions and provided detailed next steps.

As with all my presentations, I encourage questions throughout and save time for discussion at the end. I believe this fosters excellent collaboration and generates invaluable insights.

Previous