Hacker News Reader

>What matters is incremental ROAS - not the conversions following exposure, but rather the conversions that would not have happened but for the exposure. [...] there's no untouched part of the market that's never seen a Coke ad against which you can do an A/B test.

Sophisticated advertisers like Coca-Cola are aware of the concept about incremental ROAS. They can't do the exact A/B test scenario of isolated consumers you're talking about but decades ago, they did do A/B tests in different tv markets where one city had more ads than another and the city with more ads had higher sales. (What the industry jargon calls "lift" from advertising exposure.)

So even brands that are already very well-known by most of the public still constantly do A/B tests to measure incremental conversions in all media including digital, magazines, sports sponsorships, etc. Back in the late 1990s, many advertisers noticed that running banner ads on Yahoo didn't work which contributed to their stock price crashing. Recently, a lot of advertisers (e.g. Proctor & Gamble) quit spending ad dollars on 2nd and 3rd-tier ad exchanges because their A/B measurements showed they were a waste of money. (The "1st-tier" ad exchange examples would be Google & Facebook.)

> Recently, a lot of advertisers (e.g. Proctor & Gamble) quit spending ad dollars on 2nd and 3rd-tier ad exchanges because their A/B measurements showed they were a waste of money.

Can you share a link? I work in digital marketing and would be super interested in more info on this.

Whenever I buy woodstain I buy ronseals quick drying woodstain because it does exactly what it says on the tin. This is because I watched tv in the 90s as a teenager and it was ingrained.

The benefits of brainwashing can pay dividend over a lifetime.

Your point actually proves that, advertising gets the customer through the door, but, your product needs to be actually decent to keep the customer.You saw the advertisement, you bought the product and it worked as described, and continued to work in that way in the future.

The product is successful not because of advertising, because it is actually good. I find it weird that, most people spend huge on advertising, but sometimes neglect the product.

A lot of products are good though...So your point isn't very convincing that products aren't successful because of advertising given the zero sum game of hooking a customer for life that OP describes.

I have no idea. It’s rare I stain wood, I have never compared any other stain, I don’t know if it’s any good or not, or indeed the price. They may charge twice as much as bobs woodstain but it’s not worth me thinking about.

What you're describing is pretty typical though. You buy a product either because of advertising, because it "looked" like a good choice in the store for whatever reason, it was recommended on Wirecutter, the Home Depot worker or a contractor pointed you towards it, etc. And it seems to do the job and, for the amount you spend on that item annually, it's not really worth doing your own comparison which you probably aren't in a position to do scientifically anyway.

It also doesn’t mean that the advertising is money wasted. Coke spends that money to remain the go to brand, and surely values being the cultural default very highly. Saying “Coke would still sell without ads” really misses the point for why Coke advertises.

It reminds of car ads. Apparently (correct me if I’m wrong), but OEM ads aren’t about converting new customers, but they’re about trying to convert recent buyers into lifetime buyers. It’s to build in the consumer the attitude of “we’re a Ford household”, not to convince a Chevy driver to buy Ford for the first time.

I don't disagree with you. My point wasn't that Coke-style brand advertising was worthless, just that it's extremely difficult to measure without much longer experiments - specifically because it's been so successful.

I remember an experiment about coparing cola and psi, they found advertisement of cola was so successful that people saw a cola bottle will let them feel it more delicious. I assume they want to keep this advantage in internet era

Wouldn’t it be better just to have it as product placement rather than advertising? Warren Buffet makes it look great, 5 cans a day. With his longevity and investments he’s great for them.

Product placement is a subset of advertising, not a different thing.

The automobile analogy brings up another benefit of advertising. Items that are sold as socioeconomic status proxies don’t work unless everybody knows the brand.

It’s not just that Volvo wants me to think I drive a safe wagon that is a sensible choice for middle-class, educated people who watch public television.

I want my neighbours to think these things about me even though they drive Ford and GM. That advertising assures me I’ll get both a car and a cachet.

If Volvo had a way of only selling to people who want Volvo cars for the utility, but few others would have heard of the brand, it would have less value.

You make a good point. But there is one more reason for advertising than just conversions. If Coca Cola were to stop advertising, it would free up substantial ad space allowing for their competition to advertise cheap. These large brands don't just advertise for branding/conversions. They plan their spend budget so as to actually influence the bid rate. This makes it harder for newer brands to compete and get any viable ad exposure at affordable prices.

TY. Incrementality is king. Measuring that as a marketer and advertiser is getting more challenging for a variety of reasons. The future, primarily for large brands, but increasingly accessible for smaller and smaller companies is in statistical analysis with properly controlled experiments.

It is the only path I've seen that can account for all the noise in the measurement ecosystem. And even then it is far from easy to do "right."

> The future ... is in statistical analysis with properly controlled experiments.

The future? No, statistical analysis was always at the core of advertising, from the moment statistics was invented.

It's only recently in the last few years that people forgot this, and only because techies and programmers without statistics knowledge successfully "disrupted" the advertising industry.

But internet advertising sucks balls compared to classical advertising forms, especially from the point of view of ROI to the client.

So we will eventually "undisrupt" this mess and make internet advertising more like the classical forms.

t. working in ad tech for the last 17 years.

Fair point about statistical analysis. My comment was more that the tools and data to provide that visibility are starting to trickle down to smaller companies, sort of like how GA did it for web analytics. For better or worse.

Disagree on traditional vs digital though. It is totally circumstantial which performs better.

Incremental ROAS is the right metric in theory, but very hard (or impossible) to measure in practice, in particular for the small businesses with smaller ad traffic. Hence the industry falls back on measurable proxies such as ROAS.

I assume there must be some effect in getting new drinkers.

I’ve tried lots of colas. I’m never buying a Pepsi if there is Coke. I will always buy soft drinks in this order: Cherry Coke, Cherry Dr Pepper, Dr Pepper, Coke, maybe I’ll just have water. At this point no amount of advertising will change my mind.

The advertising doesn't simply need to affect your personal preferences to affect your buying habits. If you were buying a few different packs of soda for a group of people (e.g. for a party), how much cheaper would the case of Pepsi need to be than a case of Coke for you to buy it instead? Could that be affected by advertising that made you feel like, regardless of your personal affinity for Coke, Pepsi is the acceptable default choice?

Imagine further that this isn't a product category that you feel quite so strongly about (most people don't have a 4-deep ordered list of brand preferences for common product categories). If you were making a quick choice between Sprite and 7-up (again, for a party), are you sure that advertising couldn't possibly influence which of the brands is most accessible to your brain?

This is a common anti-pattern when smart people start thinking about (brand) advertising.

“I have such strongly held opinions about my favorite soda! No amount of advertising will change my mind. I am immune to advertising!”

And then they go to the grocery store, and remember their girlfriend told them to pick up some fabric softener, which they know nothing about. So they look on the shelf and they think “Ok, Downy, I’ve heard of that one. It’s probably fine.” And away they go.

And the next time they need fabric softener, they reach for it again, because it worked fine last time, and what’s the point in spending any more time thinking about fabric softener?

This is what brand awareness advertising is meant for. It’s meant to change your weakly held preferences, not your strongly held ones.

This is precisely my point about getting new buyers but couldn’t think of the terms. There must be some clear up shot to spending so much on advertising.

That’s why these companies invest so much in advertising. Market share is gained or lost in an instant.

If Coke switched off their advertising they would have gained this market share in recent years?

https://www.statista.com/statistics/225464/market-share-of-l...

Is there a control to prove it was advertising? They bought a lot of brands and there isn’t evidence that online marketing was the reason.

The point you're talking about ROAS/incremental ROAS is quite moot actually. Every online ad tracks the user from the moment they click the ad to whatever events they make (e.g. add to cart, purchase, etc). So the measured ROAS is exactly for the users who'd come via the ad and not any others.

I don't want to be offending but this is honestly a very basic point, of course one would only measure the Return On Ad Spend from the user's acquired by said ad and not any others... Advertisers are familiar with all such basic statistics and Google/FB ads give you very easy tools to track any person who clicks on an ad throughout their entire journey.

Edit: I was mistaken as described in the comments below

You are not understanding the point. You cannot actually measure incremental ROAS with basic statistics and easy Google/FB tools, because you would need to know how much they would have spent WITHOUT being exposed to the ads.

You can run lift tests on Facebook which measure exactly that.

You can't measure it across multiple advertising platforms very effectively though.

I don't know what lift tests are. It should be easy enough to devise a tool for measuring incremental ROAS under the following assumptions: a) all purchases happen online; b) I have a complete history of ad impressions and ad clicks for each user who made a purchase

As you said yourself, b) breaks down for multiple platforms. It also breaks down if users have disabled tracking.

> It also breaks down if users have disabled tracking.

Almost all users who disable tracking also block ads, so you are still correctly measuring the effect of your ad spend.

Lift tests are these things: https://www.facebook.com/business/m/one-sheeters/conversion-...

They are essentially what one needs to measure incremental ROAS of one's advertising.

Now, there are obvious caveats (online, tracking etc) but they're much much better than what's available for other platforms.

Well, if we are talking about a recently formed company who (1) doesn't do any offline advertisment and (2) only does fully tracked online ads, you can calculate ROAS fully right? (The only exception would be word of mouth sales)

It's a big exception; otherwise you are just assuming that the only way people can get to your product is through ads. But this is equivalent to assuming "incremental ROAS = ROAS". So you propose to solve the problem by ignoring it. ;-)

PS: Of course that works if the difference is indeed small enough in your case so that it CAN be ignored.

Yes, obviously ROAS is only computed over the people who saw the ad. That's not in question. The point is that knowing how many people saw the ad and then bought the product doesn't actually tell you how effective the ad was at improving your business.

Imagine a product, let's call it Oxygen, that every single person buys $10 worth of every month. It has 100% market penetration.

One day, Oxygen Corp decides to take out a $1M ad buy. They reach 1 million people, all of whom then go on to buy $10 of Oxygen, as measured by cross-site conversion tracking. $1M ad buy to move $10M of product - that's a whopping 10.0 ROAS. Must be the most effective ad campaign ever run, right?

I see what you mean and stand corrected; I wasn't assuming the case where the product sells without advertising at all.

From a gut feeling I can say that the customer came from some result of previous advertising (as we're talking about products like sodas not life-essentials like oxygen), but I guess there's no way to know since those previous ads being tv/print ads were not tracked.

If we suddenly stopped all other forms of advertising and only used online ads, in 10-20 years each customer can be tracked exactly to what ad created the first impression about the brand and thus be more accurate (still excepting marketing like word of mouth though)

You can run that same study with a control group, and sophisticated advertisers do. Show some of the users ads for your product, and other users ads for something unrelated, and compare their purchases of your product.