According to eMarketer, the expected global spend on social media advertising in 2015 is $23,680,000,000. That’s a big number, and the number for 2016 will be even bigger. What’s funny, though, is that the marketers spending all this money are constantly trying to spend less.

There’s a great quote from a turn-of-the-century department store owner named John Wanamaker, which goes:

“Half the money I spend on advertising is wasted; the trouble is I don’t know which half.”

That quote doesn’t hold much water when it comes to social media advertising. With the advent of Big Data and web analytics came the idea of 100% efficiency in advertising spend. 

Through click-tracking and cross-device technologies, digital marketers can now see the exact Return-on-Investment of every dollar they spend. Whether they use the acronym Cost-Per Click (CPC), Pay-Per-Sale (PPS), or Effective-Cost-Per-Action (eCPA), digital marketers are always concerned with how much they have spent relative to each individual “conversion,” which could be a sale, an email, a social media engagement, or a new user signup.

To bring that cost-per-whatever down, digital marketers mainly rely on a practice called “AB testing,” also known as “split testing.” The practice goes likes this: run two advertisements with one simple difference between them, such as the image or “call-to-action” used. Once the test has been run, you can evaluate how each version performed and decide which is more cost effective.

Setting up split tests, as well as researching new advertising platforms or tracking technologies, is all part of the scientific-ish skill called “Optimization.” It is essentially the practice of monitoring a company’s ongoing marketing efforts, running experiments and making changes in messaging or targeting that improve spending efficiency. This broad skill will only become more important to businesses as the global spend in social media advertising increases.

Optimization can only be used to augment and improve marketing efforts, though. You need images (aka “Art”) and writing (aka “Copy”) to AB test in the first place. Since testing paid social media advertising costs money, it’s important for digital marketers to develop an intuition for what works and to be able to create ads that perform moderately well in the first batch of testing.

Art and copy have been around a lot longer than social media, so there is already a shared intuition for what “looks good” and what “sounds good.” What hasn’t been around for a long time, though, and what is the third key ingredient for creating a cost effective social media ad, is targeting.

“Targeting” in social media advertising works differently depending on what platform you are advertising on, but it all boils down to choosing who you do (or don’t) want your ads to be served to. Since digital marketing is so focused on efficient spending, targeting the right audience is crucial. It costs money to serve ads, so serving ads to the wrong people is a direct waste of money.

Actually, let’s take a step back: Targeting is not as simple as “who your ads are being served to.” Targeting also concerns the how, when, and where of your ads; do you only serve this ad on iOS devices? Do you want this ad to run all day or only in the mornings between 6am and 9am? Do you want to reach people across the whole United States, or only in one zip code?

Related: How To Go Hyperlocal With Facebook Ads

There are an endless amount of variables involved in targeting a social media ad. On the Facebook Ads platform, for example, you can choose from hundreds of “Interests” that users might have. The word “might” is important there, as Facebook uses complex algorithms to assign targeting options to individual users, and as I found out by looking at my own Facebook Ads targeting options, they can be pretty inaccurate. That isn’t all Facebook’s fault, however, because a lot of the targeting options across all social media advertising platforms are determined by user input – such as gender, age, job title, education, etc.

The amount of different targeting options available and the varying levels of accuracy they each have dilutes the effectiveness of AB testing for better targeting. Also, since there are so many variables to test, such as the millions of keywords Twitter’s ad platform allows you to target by, it isn’t feasible to test all available targeting options for an ad.

That is where the need for creative intuition comes in: digital marketers have to be able to make good (or great) estimates for what targeting to use for each social media ad they start running. Otherwise, digital marketers would have to run large, expensive test campaigns to find combinations of targeting that perform well. Only large companies with large marketing budgets can afford to do that, so the digital marketers working for small companies with small marketing budgets have to develop that “sixth sense” for targeting. To help those digital marketers out,

Here are some tips on how digital marketers can approach social media targeting creatively:

1. Don’t think as much about who you’re targeting as what you want them to do.

As mentioned before, digital marketing campaigns are all evaluated by their “cost-per-whatever.” Let’s say that a real-estate company is trying to find prospects for filling office space. When running social media advertising, the company might find that using broad targeting for people aged 23-40 and interested in “Business” garner cheap impressions and cheap click-throughs, since the audience size is large and the ad platform has more users to serve ads to. 

But, how much of that audience is actually in the position to decide on renting new office space? Not very much. Targeting by job titles instead, such as “Human Resource Manager” and “Director of Operations,” would mean advertising to a smaller audience, which means more expensive impressions and click-throughs, but the cost-per-office-space-rented would probably be lower.

2. Remember that different networks have different audiences and different uses.

Continuing the real-estate example, let’s think about what the best network for that company to advertise on would be. Do all of your Facebook friends list their company and job title on their profile? Probably not. Do you think about work when scrolling through your Instagram feed? Probably not. The social network most associated with work and office life is LinkedIn, which is based on connecting professionals through where they’ve worked, what titles they’ve had, and what skills or experience they have. 

Related: 4 Ways To Approach New Social Media Networks For Business

Because the LinkedIn network is so focused on job titles and companies, their advertising platform has more accurate data on what jobs their users have. If the real-estate company from our example has space to rent that is well suited for medical offices, they can run ads on LinkedIn targeted at managers within medical companies.

3. Keep in mind that targeting is just as much about exclusion than inclusion.

Okay, so now the real estate company is targeting a very limited audience of managers within medical companies. There is still more they can do to shrink this audience size, though. They could exclude companies that are currently renting office space from them, and exclude companies that are too small for the office space they are trying to rent. Anything a digital marketer can do to exclude users from their advertising that they are certain will not “convert” will lead to a lower cost-per-whatever, which is why they should:

4. Make use of all the data you have, and do research on the data you don’t have.

Let’s say, for example, that real-estate company from before has a monthly newsletter for new listings on their website. The company can probably afford to exclude those people who have signed up for the newsletter from their social media advertising, since they are already getting the information about new listings in the monthly newsletter. That makes those emails a valuable piece of data, as again, anything digital marketers can do to shrink the audience size of their social media advertising and increase relevance will improve spending efficiency and save money.

But what about the data that companies don’t have? For example, the real-estate company might not know what job titles medical companies give to the people who are responsible for deciding what office space to rent, which is important information to know when targeting on an ad platform like LinkedIn’s. In that case, the digital marketer working for this company would have to do a bit of research to build up a list of job titles they want to target, which they could do through feedback surveys from existing clients, reading job descriptions for medical companies that are hiring, or just simply “Googling around.” In the end, it doesn’t really matter how they get the data, what matters is that they use it.

The world of digital marketing changes every day, with new advertising platforms cropping up everywhere with new technologies, new audiences, and new targeting options. If companies want their marketing spend to be as efficient as possible, they need digital marketers that are creative enough to keep up.