So many questions!
This is a tough one - you have a lot of data and it's all potentially meaningful.
When I've done stuff like this in the past, I've tended to keep it focused on ONE media event at a time. I'm going to assume these are all national campaigns, so there's no geo component to the analysis (that would be another way to slice the data). So your parameter notion is a good one: show a single media event and all the traffic associated with the same time period.
You might also think about a combined view of the media events: a single bar that varies in color depth or thickness depending on how many events are happening in a given time period. So for example, the June/July period would be the thickest bar, because there are 3 events going on.
Have you tried a dual-axis line and gantt chart? You've sort of done that manually with the colored bars but you could probably do it with the actual data.
Is there any way to attribute the reservation data to a media event? I'm guessing not, but sometimes there are promo codes or coupons associated with reservations, and you can use that to tease out directly-related data points. The issue, though, as you noted, is that there are so many overlapping media events.
I'd also be inclined to see if I could baseline the reservation data: is there ever a time when a media event is NOT happening? If you have nothing to compare a given number to, it can be very hard to tell if something is just the natural flow of activity (more rentals in summer than winter, for example) or if there's a delta between "normal" and "right now" that might be due to media activity.
This is ultimately the kind of thing you have to experiment with.
If I were you I'd probably create a dashboard with multiple vizzes to achieve this result. There is probably a way to represent the media launches on your lines using some calculated fields or the like, but if the client is mostly concerned with the impact of the media launches on rentals then I don't know that you need a time series chart to show that.
Here's a quick example that I believe shows relative success of various media events:
Great and thoughtful response!
The distribution for the media in this timeframe was all in the US, which was a good point.
I think the most significant aspect was to see how effective the media launches were. Given that a large part of the media coincided over a three month period (another good point) we can interpret that the media events were effective because the highest call volumes were seen in that timeframe. I asked the CEO of our company if we could possibly combine all the network tv launches together in one value and all of the cable tv values in another value to see if that would be easier to visualize and bring more meaning to the data. More on that later.
I did try a dual axis chart, but I was not sure how that would look with a Gantt chart. I agree the Gantt chart would be the best bet for visualizing the media events, but how would we do that with the time series data in the line chart?
Avis has chosen to launch their media when it has historically been the most effective; between March and September. The other challenge is that we only have January - August call volumes in this data set and we have to "project" the media data past that point.
I agree that this will require some experimenting. I would just really like to focus on an idea that might have a good outcome. I'll give the parameter theory a try. It might be more appealing to have an interactive experience; I am just concerned that they will want to see all of the media events together. Although, now that I think of it, if there is a greater concentration of media in a given timeframe, e.g. June - August, that might be a way to interactively show any spikes in call volumes.
This makes a lot of sense. Are you showing this example in a dashboard and lining up the media events with the call reservation data? To your point, you are only showing call reservation data with respect to the media events. That definitely gives me a different way of looking at this.
Yeah, I was only looking at media events in relation to each other. If you add the reservation dates where there are no media events it gets messy.
I've got basically zero experience with Gantt charts, so can't help there, but what you're asking seems like something Tableau can do. Or, it may require you to reshape your data in a way that supports the desired viz type. I'm sure someone on this forum can help.
You may want to look at average rentals for media events vs. non media events to get that comparison. Some sort of high level aggregation for your dashboard. And then you may want to experiment with excluding certain weekdays since it looks like weekends are always low - so that if one of your comparison groups contains more weekends it could bring down the average.
Keep in mind that I have no idea what I'm talking about. I was just hoping to help you think outside of the time series/gantt box, but that still may be your best bet!