Using Data Analytics to Optimize Marketing Spending


The Solving for B logo
Solving for B°
Using Data Analytics to Optimize Marketing Spending

Aligning your marketing goals with wider business objectives can be a major challenge, especially when it comes to budget allocation. How do you make informed decisions when determining marketing spend?

In this episode of Solving for B°, Dr. Hari Sridhar, Associate Professor of Marketing at Texas A&M University, and Dr. Vikas Mittal, Professor of Marketing at Rice University, explain how brands can harness data analytics to create effective marketing spend strategies.

Learn how to allocate marketing budgets and how to avoid common pitfalls that misalign the marketing plan and the business strategy.

Read the Transcript

This transcript has been edited and formatted for readability.

The Need for a Systematic Approach to Determining Marketing Spend

Chris: Thanks for joining us today, guys. To begin, can you talk a little about the need for a systematic approach to determining marketing spend?

Hari: Sure. One of the big criticisms that marketing folks get is that they make up a budget. They go to the CEO and say, “Can I have $3 million to do this?” And that's fine, except the CEO is likely to come back and ask, “How do I know this is going to pay off?

One of the problems with having an ad hoc approach is that nobody really knows the right amount to spend and whether it's paying off. The whole idea of trying to make it more systematic is to build credibility to the function and create accountability in the marketing function. Right?

Vikas: I think it relates to the old saying from John Wanamaker, “I know half of my marketing dollars go to waste, I just want to know which half.” One of the main reasons why people think like this is that in a lot of companies, marketing is seen as a communication function.

Additionally, marketing CMOs typically go to the CEO with an ask. But a CFO always goes to the CEO with an investment budget. “If we do this, this is how much more we expect to make,” or, “If I cut these jobs, if I do the streamlining, this is how much we expect to save.

To Hari's point, what we are advocating is that Chief Marketing Officers should be able to relate their marketing spend, and different components of their marketing spend, to tangible outcomes. Whether its sales, margin or EBITDA, so they can position the spend not as an expenditure but as an investment.

Chris: Right. Whenever a Chief Technology Officer, for example, says, “I need this budget because I'm going to upgrade system,” you're going to see some tangible evidence. Correct me if I'm wrong, but traditionally the CMO will have little to no data for support. But what we're saying is it's important to add some sort of expectation of results.

How Do People Determine Marketing Budgets Currently?

Chris: In lieu of having analytics to help determine the budget, how do people determine what budget to ask for? 

Hari: Well, you'd be surprised that more than 50% of companies still don't have any real systematic method to do it. They have methods that are systematic from a procedural standpoint but they aren’t necessarily – tying it back to Vikas' point –  creating any accountability in budgeting.

There are three methods for determining budgets that I can think of which I call “Heuristic Based Methods.” They seem systematic, but they're very dangerous when you think about what value they are creating.

Person Data Sales Method

Hari: One of them is what you call “The Person Data Sales Method,” which is to say, “If I'm a $10 billion company, a certain percent of my annual sales should be devoted to my marketing function.

Now, this is systematic because a certain percent of your sales goes to marketing every year. But nothing in the heuristic points to the value that is being created by marketing. It doesn't allow to measure if marketing is actually creating incremental sales nor determine the activities that should be tied to marketing.

All You Can Afford Method

Hari: Another method is what you call “All You Can Afford,” which is basically saying, “I don't have any money left, so you're not going to get any.

Squeaky Wheel Method

Hari: The third method is what I call “The Squeaky Wheel Method,” which is to say, “I'm going to go up there with the CFO and the CIO and we're all going to fight for who's going to get most of the budget.

So all of these sound like they're systematic because you go through the same procedure every year, but they don't really create any value.

Determining the Effectiveness of Your Marketing Spend

Chris: And they lack data behind them, correct? That's the key component to this. The justification for why a company needs to spend that amount of money.

Vikas: Right. And linking it to the sales and margin, or EBITDA, to show the amount of money spent on advertising, the money put into trade shows or the money put into print ads, and the lift that the company got in sales from those efforts. So there's tangible evidence for the CEO to notice that this spend was not an expenditure but an investment.

Chris: Many of these models, or this analysis, can really tell companies what particular aspect of marketing works, and in what capacity. It's not just “Spend X amount of dollars, and you'll get X results.” It's “Spend this amount here, and this amount here, and this is what you can expect.

Hari: Right. To give you an example, in one of the companies we talked to recently, we said, “How are you spending your marketing dollars?”

They said, “Well, we allocate most of it to trade shows or to a trade press magazine. We always spend 80% of the budget on trade shows and 20% on the trade publications."

We responded with, “Okay, why do you do that?

Well, the trade show is a ‘got to be there’ event. It's a yearly event and we get a lot of our leads from there.

Okay, what have you done over the last 10 years to see if that's the right strategy?

Oh, we know it's the right strategy.

Something in there tells me that we need to examine the strategy by looking at what each of these two mechanisms generates. If they didn't go to the trade show next year, what would happen? Those are the kind of questions you have to think about, as a thought experiment, to figure out whether marketing is paying off.

How Companies Should Determine Their Marketing Budget

Chris: So how should companies allocate their marketing spend budgets?

Hari: There are many analytical ways to do it or at least one analytical way that has many terminologies associated with it. Some people call it “The Attribution Approach.” Others call it “The Statistical Approach” and there are people who call it “The Response Model Approach.

Regardless of the name, these strategies are looking to answer the question: “What is the incremental value of every dollar you put in the market function to your sales or your EBITDA dollar profit?”

If you spent $100 last year, and this year you spent $101, what does it really mean in terms of your sales? That's what this model speaks to.

Vikas: Think of it as a see-saw. If you press one lever – which might be advertising – what is the independent lift that you get in sales?

So the issue is that if you just keep spending more on sales, trade shows and print advertising, you will obviously get some benefit. But Hari is saying that by quantifying the relative benefit, it’s easier to make decisions and adjust strategies.

For example, knowing that a one percent increase in advertising gives you a 10% increase in sales and a one percent increase in trade show gives you a 12% increase in sales, allows you to make wiser decisions.

With this information in hand, you know that you should be putting a little bit more on trade shows. That quantification becomes critical for the CEO.

Now, if you go to the CEO and say, “Look, we want to get a 22% lift in sales and this is the budget allocation that will get you the 22%.” Even if you don't get exactly 22% – let's say you get 20% or 24% – the statistical model can help you identify why you didn't get the sales you expected. So you get smarter for next time.

That's the crux of this approach. If you fill your gas tank three-fourths of the way, and you were expecting to go 200 miles, and you only went 190 miles, you now know how to get that done.

Chris: It sounds like this is a process that should be iterative and should be monitored. What do you think is a good timeframe for monitoring?

Hari: Well, to determine if you have executed a good marketing strategy you must know the size of the budget. And you should be able to say with precision how you want to allocate this size of the pie across different vehicles.

So you want to be confident that if you spent a million dollars in three different vehicles, you can repeat this process every time. That's the first big thing you should be trying to do every year.

The typical way you do this is going back to see how you've done it in the past. Usually, you collect historical data on some outcome of interest. Say sales, leads or profits.

And using statistical approaches, try to track whether increases or decreases in any one of these vehicles have actually cultivated sales. In other words, the best data you have is all the marketing you've done in the past.

Once you've got this historical data and you run it through the statistical processes, you should be able to determine whether a marketing vehicle works. Once you know this, you can decide the size of the buy you want to have and the slices of the pies.

Interpreting Marketing Data

Chris: Are these models sophisticated or detailed enough to show you the point of diminishing returns in certain activities?

Hari:  Yes, in fact, the first fundamental principle behind these models is that, in general, your marketing investments have diminishing returns to scale.

That is, over time, they'll begin to become incrementally less effective. The trick in these models is to determine when each of these vehicles is reaching that point.

That way, when you have a vehicle that’s already a spent force, you can start allocating money towards the other vehicles. That's the intuition behind how these models work.

Chris: So, it sounds like it's pretty crucial to not extrapolate too far ahead. Even though the data tells you something, that doesn't mean that you just multiply everything by two and it's the exact same. You have to be precise.

Vikas: Hari just did this for a large company – a $10 billion to $12-billion company that makes adhesives. They faced similar issues where they had been doing all this spending, but they were not sure if they were getting anything out of it. So, Hari, maybe you can talk a little bit about it.

Hari: Sure. This started when they began to feel like their budgeting process wasn't right. When we started talking, we realized they had no real systematic process to do budgeting.

So every year they'd face the same alarming question, “How much should we really go and ask the CEO for?

We decided to try to look back at five or six years of decisions they had made. “Let's look at every month of spending on trade shows, magazines and sales force, and let's try to see if you have any valuable data on your sales."

Like I said a few minutes ago, we asked them which vehicle they thought was most effective, and they said, “It's got to be trade shows. It's this huge event every year, all our suppliers and competitors are there. That's where we put most of our money.

Once we ran through the models, we found that the incremental effects of the trade shows were basically close to nothing. So we said, "Well, this is a problem because you believe this vehicle is very important, but in fact, the data says it doesn't really generate any incremental lift.

So we walked through the process with them because this is as analytical as a cultural shift in companies. And we were able to explain that if they pulled back on their trade shows by about 20% and allocated more money to trade press magazines and sales force, they'd likely get an 18-20% bump in sales.

This was a little bit of a shock to them; they didn't really believe the numbers. But the whole point, like Vikas said, is not really the specific numbers, in some cases, as much as the directional guidance. The big insight here was, maybe trade shows are not the main vehicle.

When they shifted it around, they said, “Okay. We're beginning to see more bang for our buck,” because they didn't really have to increase their budget; they just had to redirect it.

And research shows that, in fact, redirecting your money in the right way can actually yield more returns than going and asking for a bigger budget. That's what these models help you do: work with the same budget, but being smarter about how you allocate the money.

Vikas: Sometimes these models are extremely useful. A company probably has already committed to an investment, but they don't have a good way to figure out what the payoff from that investment is.

For example, a company had developed a free app for their clients. This is a B2B company, and the app basically helps the client figure out what specific tools they should buy, based on the client's needs.

This app was given out for free, and it was a million-dollar investment. But the company had no clue if they were getting any benefit from the app. Except for the CMO running around saying, “The app is beneficial because it's helping in relationship marketing and customer engagement.” But nobody knew what the benefit of all of this was.

The ability to quantify the benefit, and how you can talk about it, really helped the CMO get a huge budget bump from the CEO.

Hari: I think the key intuition, in that case, was to see what does this free app does to the bottom line. We were able to look at the buyers who actually downloaded and used the app, and the buyers who did not use the app.

Now, of course, some of the buyers who downloaded the app already liked the company, and there are ways to correct for things like that in our models. But the crucial thing here was, simple usage, repeated usage and engagement of the app were leading buyers to keep this company at the top of mind, which eventually led them to buy more.

After 18 months of tracking, we were able to show that in fact, 16-18% increases in sales emanated from the usage of the app. Once we translated that to sales dollars, it amounted to about $6 million.

The CMO went up to the CEO and said, “Well, the incremental benefit of this free app was $6 million, and I'm just going to ask you for half of it to make the next version of the app.” And the money was sanctioned.

Ultimately, when a CMO has a budget to make, with the confidence that each of the activities is creating an incremental lift in sales, he or she can write a budget a lot more confidently and demand more money.

Not only with confidence that it works but also with some projections in terms of what it'll do. This is a good tracking mechanism for a company to have.

Common Mistakes of Determining Marketing Spend

Chris: This reminds me of our conversation we had in preparation for the episode, when Vikas mentioned that there are three questions that should be answered by these models:

  1. What should the total spend be?
  2. How should the total spend be allocated?
  3. What are these dollars going to get me?  (Or how is this going to affect my bottom line?)

This is a really good example of a way in which these models can be used. Can you guys talk a little bit about any of the pitfalls or the mistakes that are most common when determining marketing spend?

Hari: Sure. One of the things I want to point out – and maybe Vikas can talk about it – is a phenomenon in which companies have these initiatives that they believe are actually paying off, but there's a huge fight among executives within the company about whether they pay off or not.

I think that solely relying on analytics to determine marketing spend effectiveness is a problem. In my mind, it should be a way to inform the CMO about how to go about decision making. But it shouldn't become the basis for an argument like, “Oh, he didn't hit the exact number.

Vikas: Just like anything else in life, these models, despite the fact that they can be very precise and provide a lot of information, should not substitute for good decision making.

What we've seen a couple of times is the CMO – or anybody else – will take the result from a model and will just sit there and bang on a table saying, “Look, you got to give me this kind of money.

Even if the model gives you an answer, this answer is more directional in giving you a general idea of what that answer really is.

A second thing you've got to recognize – and we've faced this quite often – many times, even if the correct answer is staring people in the face, they're unable, or unwilling, to do the right thing. And it's simply because of culture or that they’re used to a way of having things done.

You've got to be aware of all these things. For the first couple of iterations of doing something analytically, don't expect people to believe you.

And don't expect people to do the right thing just because you've given them the right answer. You just have to be mindful of the process of getting buy-in and getting people to understand and believe in this.

And the third thing that I would say is, for these types of approaches to work, the CMO and the marketing department have to believe in it.

This is a legacy issue. Frequently, many marketing departments that tend to be calmer or more of the creative type, they may not have the quantitative training necessary, so they are skeptical.

If you think about finance and operations, and all of those departments, half the reason it works for them is that they believe in what they're saying. So I think the people who are behind these models must believe in them.

Hari: Another thing that I'll point to is that many times, these analytic approaches rest on the quality of the data that companies have.

In many cases, you have companies that work with six or seven years of data and all they have is yearly data, in which case, you have seven data points on which you're deciding the future of the company. You probably don't want to do any analytics if you don't have any reasonable data.

There's really a whole process that you want to put into place, whether you have the right kind of data, and the kind of claims you can make with the data you have.

If you just do analytics because it's the in thing today, to the point where you're really building more imprecise processes than the ones you already had in place, actually destroys the mission of marketing spend.

My two cents on this issue would be that there's a whole slew of already well-established processes that go into determining whether the quality of data you have is good enough for you to be going through an analytics exercise in determining your marketing spend.

And if you find that you don't have the data needed, you can put into place a process to collect that data. But you definitely don't want to do a half-baked analytic exercise and put the whole marketing department's budget at stake.

Collecting Data

Chris: It seems like there are a few necessary elements for an undertaking like this: buy-in and a significant amount of reliable data…

Is there anything else that is absolutely necessary to put together the data for marketing spend?

Vikas: This is sort of like medicine. Reading a couple of articles on WebMD doesn't make you an expert. This really is quite complex statistics. So it's not a simple average on a spreadsheet where everybody can run around being an expert.

The one thing, if you decide to do this, is to get the right type of data. But the right type of data doesn't mean more data. It means that an expert in this type of modeling should look at the data and find the right analytical method.

And, with the answer they come up with, they should be able to tell you a level of confidence. So the upshot in all of this is, “Kids don't try this at home.” You really need somebody who knows about this to guide you in this sort of exercise.

Chris: Great. Well, guys, I think that about covers the topic for today. I really appreciate the time. Thanks for joining us, and we'll let you get back to your day.