[manigandham]

> So to answer your question: major brands have billions for branding and it's a bunch of people's jobs to spend that money and convince the people they work for that it's money well spent. And if it's not money well spent, they'll find someone who will tell them it is.

I'll be the first to say there's a lot of mismanagement, incompetence, politics, etc that leads to this but it's also one of the most data driven industries around and there's a lot of proof behind the results. It's not all just random guessing.

[scottmmjackson]

I'll go even further: it's a big industry. Companies of all size and sophistication buy advertising in huge amounts.

You have some players who have honed their advertising strategy for years to try to get the biggest bang for their buck- and they're willing to pay to have professionals work out how to hone it further.

You also have unheard-of locals who have pinstriped sales reps telling them about how much money they're going to make if they buy this billboard or phone book page, and agencies paying big bucks to get the pinstripiest, salesiest representatives to the right suckers as efficiently as possible.

Then there's everything in-between for all the people who are in way over their heads but want to at least imitate the ones who are doing it right.

[shostack]

Despite all that data though, there's still very little in the way of a clear approach to figuring out cross-channel attribution, valuing view-throughs etc.

[manigandham]

This is a case of it being simple but not easy.

The technical strategies are pretty straightforward but it's all the business policies, silo'ed data, bad integrations/tech, privacy issues/constraints, and (the worst of all) politics and outdated thinking, that cause these issues.

Attribution isn't that hard, it's basic analytics and statistical analysis - but half the agencies don't have any understanding of math or tech and just use last click wins with some unreliable vendor and probably poor implementation which ultimately hurts everyone.

[shostack]

As someone who has invested countless hours reviewing attribution reports and has seen how it is handled by companies of all sizes (including up to Fortune 50 brands), I respectfully disagree with your statement that "attribution isn't that hard."

I have the fortune to also work with an incredibly bright Data Science team (several of whom have phenomenal stats backgrounds), and they all agree with me.

Many companies and agencies know last click has very real limitations. Likewise, for anyone that has started to go down the rabbit hole, you quickly find all of the other static models have similar limitations. Dynamic/data-driven attribution at the user path level is the way forward, and Adobe's econometric attribution modeling tools are the closest I've seen to getting it right. But even that has limitations (cost being just one of them). The free reports in GA and AdWords are a great start, but likewise have their own issues.

There are a LOT of variables in terms of sample sizes, data accuracy, inability to effectively isolate an experiment group due to other marketing efforts, etc. that all throw other major wrenches into this.

All of that said, I'd genuinely love to hear your solution for how to definitively solve attribution from an analytics and statistical analysis perspective. As much as I disagree with your statement, I realize I don't have all the answers, and if you have them, I (and many others) want to hear them.

Personally, I think this is the biggest challenge the industry faces right now. My gut says display and video CPMs are overvalued, but better analytics and better data are needed to really help advertisers answer the questions of things like "what is a view through worth?" or "how much revenue should I attribute to this display/video campaign?"