How Johnson & Johnson evolved measurement models for higher impact

Aleksandar Petkovski, Senior Director – Marketing Science and Digital Analytics at Johnson & Johnson, speaks about the importance of capturing e-commerce in marketing mix modelling and building strong partnerships.

What role does marketing mix modelling play at Johnson & Johnson, and what kind of challenges does it help you address?

MMM is important to Johnson & Johnson. It’s a capability that we've driven and enhanced over the years. It has become much more sophisticated, which I didn't predict a few years ago. But the way we at Johnson & Johnson manage the marketing mix is very customized to our business. We run what we call a ‘Semi-Closed Loop Methodology’ model. In this model, we don't run MMM in isolation but go with multi-touch attribution in which different models talk to each other while remaining separate.

We have robust predictive analytics capabilities as a result of this methodology. It helps us run simulations and optimization at not just the brand but also at the portfolio level. When I think back to when we started it many years ago, it came down to developing the data infrastructure. It's not necessarily the most popular topic, it's not what people want to talk about. But the raw data and the way you cleanse it and bring it into the data lake is what makes everything work. We started with the data that we felt most confident about right at the beginning, and over time, continuity and partnering with our agency partners gave us more variables over the years. Today, we have a robust architecture of data that enables frequent updates of the marketing mix throughout the year.

We run hundreds of models a year with multiple frequencies across 17 brands. In terms of the future, we're delivering differently than before. We don’t have PowerPoint presentations because we couldn't survive off of delivering insights via a PowerPoint legacy deck with so many brands and frequencies. We have a more unique approach; in that, we plug in all our outputs including Marketing Mix and Multi Touch Attribution work into Domo. That’s our business intelligence platform that visualizes and reports the digital analytics and insights. Through that approach, we were not only able to create turnkey visualizations and harmonizations but also actually improve collaboration across teams leading to improved optimizations.

Even the deployment [of resources] has changed, from engaging with partners to constantly optimizing every month, quarter, and year. This gives us the necessary ROI improvements, and the effectiveness of our media campaigns becomes significant over time.

How do you account for e-commerce within this kind of model?

It's one of the most important enhancements we've done over the years. There are verticals like ‘beauty’ that have high indexing sales online. The measurement wouldn’t be that accurate without the eCommerce modeling. Tools are not perfect, but we’re talking about the marketing mix where we put a lot of effort into making sure that we have the exact numbers. As a result, we get high coverage with a greater degree of accuracy. In turn, the insights and predictions have more impact.

Which challenges did you face when bringing in the e-commerce component as part of your model and while assessing the accuracy of that data?

It’s much more work and data pipes that you work with upfront. There are nuances with every data set in e-commerce, especially with online sales. We don’t look for perfection, because then we'll never make any progress. We look at the information that online sales give us access to, and work that into our data infrastructure pipeline. The complication was that every single data set had a nuance that may make it hard to get it in a certain time frame or to align with your brick-and-mortar data. But we had to wrestle all that down before we felt good about modelling with online sales.

What advice would you give to marketers? What do you think is most important in terms of getting it right?

Not looking for perfection is probably the first one. Marketers tend to default to wanting something perfect, but the reality is nothing in this world is perfect. There are nuances and marketing science is a part of them. Ask yourselves questions like, ‘What are those areas where we would just have to take a leap of faith?’ That's the first thing to consider.

The second is knowing that nothing will work without investing in the actual capability, in talent. It’s a people business, and it is based on internal and external partnerships.

Finally, the third piece of advice is to have empirical evidence. You cannot just live and die by the marketing mix, you must also overlay other metrics or signals to help either produce more confidence in the actual results or go deeper into the meaning of the results.

Let’s talk a bit more about your collaboration with Amazon. How has your MMM strategy evolved since you started partnering with Amazon?

We've had a strong engagement with Amazon for several years. We’ve always shared information and data-run miles collaboratively. But the most significant step was to build a feedback loop. Essentially, we pass over our performance results and look over their feedback. This way, we know our categories best and they know their advertising solutions best bringing these pieces together. Once we get the feedback, we have a discussion over it and this is when optimization happens.

We bring the results of the feedback loop into the marketplace. Then we rerun the analysis when that campaign is done. It’s an iterative process of feedback that is key to driving value for both Amazon and Johnson & Johnson.

Do you feel that your Amazon retail data and your models are helping improve overall performance?

It’s been a gradual, significant improvement. Our organizations work in a highly collaborative manner. Bringing in e-commerce sales has been one of those necessary changes in our categories. The collaboration works in a domino effect: the higher the coverage, the more accurate the models. The more accurate the models, the better the recommendations, and the higher the impact. I can't stress enough how important it is to have online sales and e-commerce models as part of the marketing mix program.

Do you have any actionable recommendations to share about a successful partnership?

Sometimes, people even struggle with revealing information, let alone wanting to partner. It’s not easy to simply give and take information. It sounds very generic, but the key lies in being more turnkey, template-wise, to share information back and forth. You must also devote more time and resources to analyze and align the actual insights, rather than simply wrestle down the data.

Are there any improvements that you have specifically seen since initiating the partnership?

We see massive ROI growth when we have a close collaboration. It comes down to the basics, right from being able to share information back and forth to getting feedback from your partners. That way we are intelligent about our recommendations.