Fixing brand marketing analytics

Things CMOs know: brand marketing is extremely valuable for growing an organization.

Things CMOs can struggle with: convincing stakeholders that it's worth investing in brand marketing.

The reason? It's tough to show the value of brand marketing through analytics, as you might not be able to clearly attribute brand-building activities to revenue growth, which makes it hard to convince people of its value as opposed to the clear ROI of performance marketing.

That's why Chris Kelly, founder and CEO of Upwave, joined our sister community (CMO Alliance) on their podcast, CMO Convo, to discuss how to make analytics work better for brand marketing, so you know how to grow the brand more effectively, and convince others that it's worth investing in.

You can check out the episode here, or scroll on for a write-up of what was discussed.

The value of brand marketing analytics for CMOs

Hi Chris. Welcome to CMO Convo. How are you doing today?

I'm doing well. Thanks for having me.

We’re delighted that you're here today because we’ll be talking about a topic that’s near and dear to the hearts of many CMOs – brand marketing analytics. But before we get into that, could you tell us about yourself and why you're speaking to us today?

Sure. I'm Chris Kelly, founder and CEO of Upwave. Upwave is an analytics company, and we run analytics for brand advertising. We sell a software platform to the world's biggest brands and their media agencies and media partners, and that platform helps them answer the age-old question of whether brand advertising works. We leverage machine learning to help understand that question that CMOs think about every day: does my massive brand budget pay off?

And if CMOs haven't got a brand budget, they could use analytics to pitch for more budget for that. They need to be able to show that brand marketing and brand activities actually work, and that can be very difficult in certain scenarios.

In startups, where it's all about demand gen and hitting growth figures, you might not have the time or the resources to devote to brand marketing because you can't show its effectiveness. Is that why we need a brand marketing analytics process – just to show that brand marketing works?

Yeah, it's to show whether brand marketing works, how it works, and why it works. All those questions need to be answered to justify the spend. CMOs often have to go to bat for their marketing budget, especially their brand budget, and they're usually pitching to a CFO whose job is not to understand exactly how marketing works – their job is to look at numbers and justify those numbers.

If you're a CMO trying to justify your top-of-funnel brand dollars, that can be hard because brand awareness doesn't pay off overnight. Brand is top of funnel. It generates demand but it doesn't capture it. We've had a lot of conversations with CMOs over the years where they will quietly complain – to put it nicely – about their CFO’s lack of understanding of how brand marketing pays off.

In the analytics community, we view our job as arming CMOs with the right answer. The job is not just to give them the answer they want to hear and tell them that brand marketing works, but to give them true data and analytics they can use in those CFO conversations.

We've coined a concept we call “the CMO’s dilemma”. To put it bluntly, that dilemma is to waste money or get fired. A lot of CMOs tell us that they know they have to invest in brand marketing to grow their brands long-term. They know that, and the research is clear on that; however, they can't wait years for their brand marketing to pay off because that's the path to getting fired.

You may know of some seminal research from two gentlemen in the UK, Binet and Fields. It measures the long-term impact of brand marketing, and it shows how you should balance this trade-off between short-term sales activation, i.e. performance marketing, and longer-term growth, i.e. brand marketing.

Their rough rule of thumb for all marketers is 60/40 – 60% brand, 40% performance. Your ratio may vary. It depends on what space you're in, what your product is, and what stage your company’s in, but that's the starting point they try to anchor people to.

We talk to a lot of CMOs who know that if they don’t do a 60/40 split, they’re going to wind up wasting money on performance campaigns that won’t generate new demand. Their attribution models claim that those dollars lead to sales, but they’re just advertising to people who are already going to buy, so it’s a waste of money.

So they have this dilemma: either do something they know is wrong or do what they know is right and risk getting fired – because CFOs will fire CMOs if they think they're spending money that doesn't have a clear payoff. As an analytics community, we have to solve that CMO’s dilemma and arm CMOs with the data to know today what the results of their brand building will be tomorrow.

The trouble with attribution

That disconnect between what CFOs want (or other key stakeholders – it's not all just the CFOs, to be clear) and what the CMOs know is right is why CMOs have the shortest tenure of any members of the C-suite. The best case we've seen is an average of three years, which is not enough time to see results from brand marketing activities and show the impact of those long-term growth opportunities.

Now, you mentioned that key word – attribution. In performance marketing, you can attribute ROI to your activities. You can see what your dollars are impacting and the return on investment; that's probably why CFOs like it. Is it possible to have that kind of model in place for brand marketing activities?

I agree about the short tenure, and I've heard even worse stats than that. Maybe it varies by country, industry, or stage of the company, but I’ve seen some data showing that the average CMO tenure is as short as 18 months, and it's shrinking. No matter what dataset you look at, it's getting worse.

I tend to nitpick over the semantics of attribution and impact. With attribution, you're tying ad spend to sales, but it only shows correlation, not cause. When you know someone saw the ad and then converted, the elementary thought is, “Cool! They saw the ad and converted, therefore that ad impacted the conversion.”

You may know the old parable about the pizza store coupons. It shows the silliness of blindly relying on attribution and not doing any kind of causal measurement or incrementality measurement, which you could argue are alternatives to just focusing on attribution.

The parable goes that there was a pizza store owner who wanted to drive more sales, so decided to give out coupons for five dollars off a whole pizza. The pizza store owner thought, “I'm going to have an attribution system. I'm going to hire a bunch of kids to give out coupons across town, and they’ll write their names on the back, so I know which coupon came from which kid and who is driving the most pizza sales.” Seems pretty straightforward, right?

They did this for a couple of weeks, and when they looked at the data, one kid was off-the-chart successful in driving pizza sales with the coupons. So they asked the kid, “How did you do this? How did you figure out how to drive the most pizza sales by giving out coupons?” He said, “It was easy. I stood at the door of the pizza shop, and I handed coupons to everyone who walked in.”

Of course, when you hear the punchline, you realize how silly it is. That kid obviously didn’t cause the sales; he just tagged the sales that were already happening.

It sounds asinine to argue that this happens at world-class brands and multi-billion-dollar companies, but it absolutely does – we've seen it. One example came from Uber. A few years ago, they were having problems with the performance marketing networks they were using to run ads.

Their attribution models said that these performance marketing partners were working, but then they saw some fraudulent inventory, so they turned off the performance marketing partners, and there was no dip in signups. The attribution model then started saying that all the signups that had previously been attributed to performance networks were just organic. No one was handing out coupons at the front door anymore, but lo and behold, they sold the same amount of pizzas.

It just shows that you have to be careful about assuming that, just because someone saw an ad and then took action, the ad caused the conversion. That’s why you need to use incrementality measurement through experimental design. You can do that through a variety of companies on a variety of platforms. It’s a smarter way to know how much pizza you would have sold (or how many rider signups you would have gotten) if you hadn't advertised.

Measuring the impact of brand marketing

When it comes to brand marketing, how do you measure its success? How do you show the impact of brand marketing on company growth?

We usually hear marketers split what they want to see into a couple of buckets and then maybe some sub buckets within those. In the first bucket, they have their delivery metrics, and in the second bucket, they have their outcome metrics.

The delivery metrics are about marketers wanting to understand whether they got ripped off. Did they get what they paid for? Were the ads shown to the right people in the right places? In other words, were they delivered properly?

Delivery metrics are important, and there are a lot of companies out there working on that. Some of them are public ad-measurement companies, who've done well by just focusing on the delivery metric piece. They'll check things like if your ad is viewable. If I show a digital ad below the fold and no one ever scrolls to it, that ad will never be viewed, so it’s a waste of money.

That viewability measurement is important to brand marketers, less so to performance marketers who are just tracking clicks and trying to see if the click-through rate went up. Brand marketers want to know if they got the exposure they paid for.

There are other types of delivery metrics too. Some companies focus on whether it was a human or a robot who saw the ad. That’s important, not just on digital, but also on CTV. There's a lot of CTV fraud out there, which your audience has probably read about.

It's also important that you reach the right person. I heard someone say recently that you have to make sure you're showing your mayonnaise ads to the mayonnaise eaters. When marketers think about delivering metrics, they also think about whether they’re showing ads to the right audience. All that stuff is on the delivery side.

And then they want to talk about outcome metrics, i.e. whether anything actually happened, which ties in more to your question. How do they know that their brand exposure is paying off? That brings us to a philosophical question that marketers need to align on – what do we mean by payoff? Not every brand campaign is designed to drive a sale right away or even at all.

People think the point of advertising is to drive sales, right? That's a simplistic mindset you'll certainly hear from CFOs. I don't mean to be mean to my CFO friends – listen, we love your work, and you're doing your job, but our job is to inform you of the nuances of these payoffs. Just because we agree that the point of advertising is to drive sales, doesn't mean the point of every ad impression or even every campaign is to drive a sale.

It's like saying the point of having an offense in American football is to score a touchdown, therefore, I'm gonna analyze my plays to see which one scored the most touchdowns and run that play the most. Lo and behold – the one-yard QB sneak scores the most touchdowns, so that’s the only play I’m going to focus on.

Well, that's silly. You only run a QB sneak once you get to the goal line. You can score from the one-yard line on the QB sneak but you can't run that play all the way down the field, right? You have to run other plays to get you down the field and to the one-yard line first.

That's how we think about brand and performance. A campaign can be successful without driving sales because it gets you down the field. That's brand marketing. Performance marketing gets you into the endzone. You have to use both in concert with each other, and you have to have both in your playbook.

In most categories, if I'm introducing a new brand that consumers have never heard of, it’s not feasible to believe they're going to see an ad, jump off the couch, run to the store, and buy my product. Think about the sales cycles of automotive and luxury brands. You're not going to see an ad and then walk the store and buy a Lexus or Rolex, right? Those brands invest years in building favorability, building slow consideration, and driving you down the field.

Even in certain CPG categories, marketers invest heavily just to keep you in the consideration set for a future purchase. Cold and flu products are a great example. If I don’t have a cold, it's unlikely I'm going to see an ad and then say, “That's a good product! I may need that one day,” and run out and buy it.

Cold and flu products are advertised year round so you keep them top of mind. That way, when I have my trigger event, which is when my kid gets sick, I walk to the store, glance at the shelf, and grab a bottle that looks familiar to me. In most cases, that’s what decision cycles look like in the cold and flu category.

Performance marketing isn't that helpful there. It's not like I saw an ad for 5% off the cold and flu product and then clicked to buy it. You had to convince me months ago of your product’s efficacy, so when I'm in a hurry because I have a sick toddler, I'm going to pick a product that I've already decided works.

It's silly, I think, to see the point of advertising as being purely to drive sales. Not every single impression needs to tie back to a sale. That’s just not the reality of how brand marketing works.

Interestingly, I think this misconception is more popular among tech people. I'll blame the Bay Area, where I live, for this mindset. If you talk to traditional brand marketers, what I'm saying is incredibly elementary, and it hasn't been new thinking for decades. This is just obviously how brand marketing works.

But a lot of people in the Bay Area see performance marketing as the be-all and end-all. They think that brand is fuzzy and brand marketing may be something that people do out in the Midwest, but not here in the Bay Area where all the smart people live. We have this irrational condescending view of brand marketing in the startup community. It's crazy because you're giving up so much growth potential if you're not properly utilizing brand marketing.

Driving direct and organic traffic with brand marketing

It's such a breath of fresh air when startups and companies take a brand-first approach. Gong is a great example of that, and the brand-first approach is really paying off for them – their growth’s been massive, and they've won a whole slew of awards.

They've also been quite lucky because their senior leadership is very marketing-focused. Their CEO’s a former marketer, and their CMO, Udi, is a very charismatic personality; you get very engaged with what he has to say about brand building, so he clearly has the motive and the ability to get people onside with his vision.

It’s gotta be difficult in companies that don't have that kind of marketing focus in the senior leadership. Often, CMOs are brought in at quite a late stage of a company's development, so there's already a close-knit community and trust between other members of senior leadership, trust that they might not have for the CMO.

In that kind of environment, the CMO needs to show results pretty quickly, and that's got to be a problem when it comes to brand marketing.

That’s a great point about the company’s DNA. Having a former marketer as a CMO matters, and I can see how that influences the organization's decisions.

The brand-first approach is harder to implement in B2B. My rants about people not fully appreciating brand marketing are mostly based on my consumer-facing experience. Most of the companies we have the privilege of working with are consumer-facing brands that can operate at a scale that you can’t operate at in a lot of B2Bs, especially startups.

I totally get that if you're a B2B startup with only six employees, one marketer, and a shoestring marketing budget, you should not try to copy Unilever, Gatorade, or Nike’s playbook. I can get how in the early days you think, “I only have six months of cash, and I’m in a six-person company. I have to be driving sales or we're going out of business. Everything has to be down-funnel.”

However, we've seen the light bulb go on at a lot of startups. As they move from an early-stage startup to a growth-stage startup they'll say, “Aha! Brand is the rising tide that lifts all boats. If I start investing in brand and people know my brand more, my click-through rates on my Google AdWords go up. My click rates on Facebook go up. Everything goes up when I invest in brand and increase my brand awareness.”

We’ve even seen this in full-on performance-marketing-focused organizations, like CPC arbitrage businesses. Lead gen sites that are about getting you to click on something on their site and arbitraging that are starting to think differently because they've realized, “Well, wait a second! I have to drive more traffic, and if I'm paying for traffic on Google, the arbitrage event eventually closes because someone else can always pay a penny more for the click.”

Google itself may even be the competition. Look at all the travel sites that have to compete with Google putting their own travel results ahead of even the page results in some cases. If you're just relying on Google traffic, you're in trouble in a lot of categories, therefore you have to drive more and more direct traffic to your website, or more organic traffic – people typing in your specific brand or site name.

The search bar is the ultimate test of unaided awareness. Driving an increase in direct or organic traffic is nothing but an exercise in increasing your brand awareness. That’s classic brand marketing, whether startups know it or not. When they're investing in ways to drive more direct traffic and drive organic traffic, they're actually investing in raising their brand awareness, and that's absolutely essential in some cases.

You can't have 99% of traffic be paid in certain categories, or your business model doesn't work. You have to invest in your brand in order to drive more organic traffic, more word of mouth, and more direct navigation traffic.

I've seen a sea change in that over the last few years. Where brand was once a dirty word to many of those businesses I’m describing, they’re now realizing, “Oh, wait a second! I do need high brand awareness. That's how I get people to come directly to my website without my having to pay $5 Google tax for every visit.”

Brand marketers in denial

Thinking about brand marketing and brand awareness at an early stage must make it a lot easier when it becomes essential to have a strong brand. If you've been fully focused on performance marketing since the outset, it's gonna be a lot harder to pivot to building brand awareness.

Yeah, and company DNA matters in that respect.

It’s funny – I've seen a couple of examples of companies being fully on team performance and anti-brand. They talk about that as their marketing strategy, but then you look closely and you realize they’re not even living that 100% performance marketing ethos; they’re benefiting from the wonderful brand tactics they’re doing.

I won't name names, but there was one particular startup – a D2C poster child that had a big exit and got a lot of attention – and I met some of their investors, who talked about every dollar of marketing tied to a sale. That was like a religion within the company. Every single penny of Facebook ads had to tie to a sale.

But when you dig in and ask them more about their growth, it's like, wait a second – that's not how you're growing! You guys put out beautiful branded content that spreads organically online. You’ve made viral YouTube videos. You're creating immersive stories with a specific voice and tone that is consistent across the videos. That's world-class brand marketing you're doing without even realizing it.

I even talked to massive public companies where the marketing team would say, “We don't make brand investments; everything we do is performance. It all has to tie to an immediate sale or signup.” And I've asked them, “You know, you guys sponsor sports stadiums. What are you talking about?” Nobody’s going to walk by the sports stadium, whip out their phone, and sign up for your products. You’re investing in the top of your funnel when you invest in a stadium.

So there's sometimes a disconnect. Maybe brand sounds squishy so you don't even talk about it. You just put it in another budget and call it sponsorship, but it's the same however you label it – top-of-funnel investments that raise your awareness and favorability.

How brand marketing analytics processes work

Can you give us a brief overview of how an ideal brand marketing analytics process works?

Brand marketing analytics is all about measuring the outcomes we were talking about earlier. You want to make sure you’re driving some top-of-funnel lift – incremental lift in a brand KPI. For example, if a brand KPI is favorability, you want to show that people who saw the ads were more favorable to your brand than people who didn't, thanks to that exposure.

A lot of companies use an experimental design, meaning they’re measuring people who are exposed to the ad (hopefully in a privacy-friendly, consumer-friendly way), measuring people who are not exposed to the ad, and then gathering sentiment data. You're looking for a change in sentiment – in this case, favorability.

Let’s say a brand already has good awareness and they want to spend a million dollars next quarter to increase their favorability. They’ll run ads and use a partner to track who was and wasn’t exposed to those ads across channels.

Next, they gather favorability data from both sets of people, through consumer interviews, for example. If they’re doing it right, they’ll see an incremental lift in favorability in the exposed group versus the control group.

That's the incrementality model I talked about earlier as a better alternative to attribution. You're not just showing that this person saw the ad and liked us – you're using a control group to gauge the effectiveness of the ad. It’s analogous to how drug studies are done. A lot of brand marketers these days are using decades-old experimental design techniques from drug studies.

Now the question is, how do you know that an increase in favorability will turn into a future sale? There are a few ways to Check. First of all, you have to think about where you get your sales data. If you’re in a B2B SaaS company, that's easy – the sales happen on your website, so you have first-party conversion data.

Not all brands are so lucky. If you sell through retailers, you don't have that first-party conversion data – the store does. Sometimes you can get that data from the retail partners themselves, but usually not. Usually, you’ll need to get that consumer data from a third party. Some companies pay consumers to upload receipts from grocery stores, so they have a panel of purchase data.

It's different in automotive, and it's different in pharma. Some companies focus just on gathering HIPAA-compliant anonymous prescription data. My point is that whatever your category or business model, you have to find a partner who can get you your conversion data.

Marketers out there might be thinking, “Okay, got it. I can see that favorability lift we just talked about; I can see conversions that happened a few months later – I can tie those together.” And sometimes that’s true. Sometimes some identifier persists (of course, cookies are dying, but that’s a whole other conversation) so you can tie that lift directly to your conversion.

However, the odds are slim that the person you have lift data for makes the sale in time for you to connect that conversion. You might also have a sample size problem, so often you're modeling the funnel, then back-testing it with purchase data at the population level or the market share level.

There’s another approach you can take if you can’t tie an individual household with sentiment lift to an eventual sale. You can model the funnel to better understand the length of the purchase cycle and the lift in different parts of the funnel. Then you can see if there is, at a bare minimum, a correlation between a change in the top of the funnel and a future change in market share.

We've also seen marketers graph their brand’s change in favorability against competitors’, then use their purchase data to look for changes in market share. Sometimes there's a lag. In certain categories, it’s not uncommon to see a five-month lag between a change in top of funnel and a change in market share, but if you have enough clean data, it can be very predictive.

Going from a lift in favorability to conversion doesn't happen overnight, which makes sense, right? You're not going to show someone an ad, have them like your brand, then run to the store and buy your shampoo overnight. There's a purchase cycle in these different categories, and it may take time to convince me to buy a different shampoo or a new brand of whiskey. It’s going to take more than just one exposure.

Those are some wonky and detailed approaches, but it's interesting to see what's possible these days and to hear what marketers are doing. This is the data they're bringing to a CFO, so they can say, “Look, I increased favorability, and I know that sounds fuzzy to you, but I have analytics to prove that we can predict a 2% lift in market share in the next five months.”

We can put a dollar value on that, and then we can put an ROI on the brand dollars that we previously were not able to put an ROI on.

Ideal timescales and sample sizes for your analytics

You mentioned some timescales and sample sizes there. Say a CMO wanted to experiment with this process for the first time – what kind of timescales and sample sizes should they be looking at? Is it very much dependent on the type of company and the vertical, or is there a rough ballpark that they should aim for?

Timescale is fully category-dependent, but longer is always better. There are, at the high end, some luxury categories where it could take literally years for the full effect of advertising to pay off. Those gentlemen I cited earlier, Binet and Fields, have done some research on this too, and I think their rule of thumb was something like two years. You should be ready to think in terms of years is the short answer.

In categories like luxury and automotive, it takes longer for your brand marketing to pay off than in categories like potato chips and soda. But even for weekly or monthly purchases, your decision cycle doesn’t necessarily map directly to that purchase cycle.

I may be a Coke buyer, but just because I buy Coke at the store every week doesn't mean that if I like a Pepsi ad this week, I'm going to buy Pepsi when I go grocery shopping next week. It might take months for a new Pepsi campaign to turn into a purchase.

So CMOs should be running their brand marketing analytics over at least months, ideally years. When you're talking about years, that’s where individual identifiers start to break down – it’s hard to track the same person for two years in a privacy-friendly way, so you have to use some of those other techniques I talked about instead of tracking.

In terms of sample sizes, the lazy answer is the bigger the better. The exact answer depends on the types of data you’re gathering. You could be tracking ad exposures on any number of platforms, you could be collecting sentiment data through online consumer interviews, or you might be looking at sales.

For all those types of data, more is better. The question is, what's realistic? Digital exposure’s the easiest – you can basically track everything and get tons of data. On linear TV, you're usually working with data partners that have television panels, and those panels have millions of households, so your samples will be pretty large.

For sentiment data, you’ll usually have smaller samples. You might have tens of thousands at the high end and high hundreds at the low end. We see marketers interview as many as 10,000 people to get data to use for their brand lift measurement from a big campaign. For a smaller campaign, you can learn a lot even just by interviewing 800 people who were exposed to your ad and 800 people in your control group.

Purchase datasets tend to be larger. If you're buying from a company that has grocery store data, they might have a panel of two or three million people, so maybe they'll have thousands or even tens of thousands of people who bought your product last week. That's the range of samples you'd expect if you’re buying third-party purchase data. If you're lucky enough to have first-party conversions, presumably you have every single conversion tracked.

Resources to build a case for brand marketing with your CFO

Chris, this has been eye-opening. One last thing: say our audience wants to investigate this further, maybe find some resources they can take to their CFOs to build a case for brand marketing in their company. I'm assuming Upwave’s got plenty of content, but are there any other resources you'd recommend?

Sure. Since you invited the plug, we have some resources that are freely available at upwave.com/whitepapers.

For other organizations, check out the Advertising Research Foundation. They're great thought leaders in measuring the effectiveness of ads, so that's a great place to start. They’ve also acquired another group called CIMM, which does a lot of white papers around media measurements.

You should also check out those gentlemen I mentioned a couple of times, Binet and Fields. They’ve done some phenomenal work on how to think about short-term sales activation versus long-term brand building, and they’ve looked across tons of categories over many years, so it's really interesting data.

There’s also the VAB, which focuses on video advertising. They’ve put out some really interesting content. Oh, and I'll close with the Association of National Advertisers. They put out some good thought leadership for advertisers by advertisers about what's working.

So yeah, check out all those resources, or feel free to contact me. I'm happy to point your audience in the direction of any third-party research I mentioned today and evangelize the great work that researchers out there have done.

Awesome. Thank you very much, Chris. This has been great. Thank you very much to our audience as well. We'll be back again soon.