Hi all,

I am currently working on a project where I try to investigate competition in online traffic / online advertisement.
I have a panel of daily data on online traffic T = 365 from N=14 domains/companies in Stata.
My data contains information on traffic in the various marketing channels these companies serve (banners, pop up, email ads etc.).

I can see in the data that one of these companies has quit serving all online ads altogether from one day to the next. In fact, this company was by far leading in all advertisement efforts, prior to them dropping out. I know why and the reason can be considered exogenous.
All companies serve the same market and offer the same highly substitutable good.

After some time series Econometrics and descriptive analyses, I can confirm that none of the other firms responded to this drop in advertisement.
As I thus made sure that there are no dynamics, I now want to apply a DiD Analysis to establish the causal effect of a stop in advertising on traffic.

Preliminary, along the lines of:
reg Traffic i.Treat_Domain ## i.Treat_Time, robust
where i.Treat_Domain = 1 is the Domain that stopped advertising.

I am now worried about the common trend assumption that needs to hold. I tried using any of the firms in my sample as control, but it does not seem to hold by mere eye-balling. Find attached a graph using one of the other firms as non-treated firm.
I am a bit helpless as I have not found any comparable setting in the literature I consulted. What is the standard approach here?

As this is small N, large T matching seems not to be possible.
I read about synthetic controlling on observable company characteristics, could that work out?; e.g using --ssc install synth-- in Stata.
Array

I am working with confidential data, so I can't upload a data snippet. I hope my description was clear enough nonetheless.

Thanks,
Markus