Google automatically refunds advertisers for fraudulent clicks each month. It's not a lot (maybe $1,000 refund on an account that spends $600k to 800k/month), but it's a line item on the invoice for sure. You have to trust them for what is a 'fraudulent' click though.
This is spot on! When I first transitioned those campaigns from eCPC to tCPA there was a big learning period. As the campaigns scaled and collected more data the actual CPA started aligning closely with the target CPA, but it took some time. The more conversions you do the faster that alignment will happen. After 2+ years on tCPA (and many behind the scenes improvements to the Algo by Google), I could get that delta to match within 3-4 days of making a change to the target CPA. That was with thousands of conversions a day though, so scale absolutely helps.
If you want to learn Google Ads I'd recommend getting certified for Search, Display, and Shopping. If you want to learn Facebook and Instagram ads then that's a whole different animal.
You're right to be skeptical! There's no mystery behind this curtain, just a guy who spent a bunch of money on advertising and wanted to write about it. My unlimited budget was strictly tied to profitability. Being profitable was the only way to scale because I had to maintain certain margins. If I hit the right margin, I could spend as much money as I wanted because the supply was there. Thanks for your comment!
Unless the armies of bots out there started spending real cash for products after those ad clicks, I'm going to stick with "a lot of of people click on ads".
It's the top result for pretty much any search. And it often is relevant. Why go down to the actual results if you can just click the first link and get what you want? And most users don't use ad blockers.
Author here and I think you're spot on! We had a tight focus on ROAS and overall profitability for those campaigns, and optimized based on that. Sure, there are companies that can waste millions of dollars on ad spend, but the company I worked with was heavily invested in making paid ads work and had the right feedback loops in place to ensure we were making the right decisions.
How did you approach incrementality? I have some experience in this area, and get frustrated when marketers talk in terms of ROAS/ROI, without discussing the return compared to other options.
The ultimate goal of the marketer is to use the most effective, scalable channels. The optimization isn't against a fixed goal (CPA), but against the entire universe of ways to spend each marketing dollar.
Hey there, thanks for your comment! You're totally right to be skeptical. I didn't include any numbers on ROI for 2 reasons:
- This was my previous job, and I don't have the data anymore :)
- I don't want to reveal numbers that could harm that company or inform their competitors.
I will say I was 100% focused on profitability. I'd be a pretty shitty marketer if I doubled spend without focusing on profit.
If you were focused on profitability, you wouldn't run Target CPA. Your marginal cost per action is going to be much higher than the target CPA, making your marginal acquisitions unprofitable, lowering your total profit. That's in the best case, if you have unlimited budget. In the unlimited budget case, Google is at least incentivized to bid on the lowest CPA users first. In the constrained budget setting, they can give you absolute garbage, and it's no skin off their backs. It's even worse if not all of the actions are incremental, which for an established company with unlimited budget is guaranteed.
The past few years, Google has been adding features that make absolutely no economic sense (assuming rational firms), but assuming that most of their clients are unsophisticated and then taking advantage of them has no doubt earned some Google Ads PMs huge bonuses.
Thanks for sharing your thoughts! That hasn't been my experience, target CPA was absolutely the driver of both volume and profitability for that business. You have to know how to work the algorithm and feed it the correct data. I was doing thousands of conversions a day so we had scale in our favor. If you're not careful tCPA can spend a lot of money for sure. YMMV though.
This is a basic application of the mean value theorem. Moreover, assuming 100% incrementality is what leads marketers to do silly throw money away things like Mailgun bidding on "mailgun" even though all the links on the first page of results are already to pages Mailgun controls.
Actually, that's not always that silly - if you have a structured campaign that uses ads to send people searching for specific things to the parts of your funnel you have optimized for them you can increase conversions and have a good ROAS even over the same keyword performance organically. It can totally be worth it to do.
Hi everyone! Author here. I'm humbled that someone felt this was HackerNews worthy, really unexpected! I'll do my best to respond to comments, but if I don't get to yours feel free to drop me an email, link is on the website.
Thanks, Nick! What are some of the most difficult actions to programmatically "prove"? Which actions seem like a good idea, but perform less profitably vs. other actions?
