You’re beaming, proud, and ready to rake in a massive amount of leads. Why? Because—you’ve got two brilliantly designed sets of ad creative and you’re ready to set ‘em loose to the hungry, scrolling consumer masses.
How do you know if your campaign will be a boom or bust? Can you even test such a thing in an easy and straightforward way?
Time for some split testing….
For those who haven’t yet implemented split testing to increase conversions, an explanation is in order. Simply put, split testing (also known as A/B testing) allows you to test different advertising strategies on commonly divided audiences to see what works and what doesn’t.
Want to see which bidding option, creative, or ad placements perform best? Split testing is the answer.
The Facebook split testing API does several great things:
First, let’s start with a simple example. Let’s go back to those two stellar ad creatives. At this point, of course, you don’t know which one’s going to perform better. The first thing you should do is set up your two ad sets, with each one of your creatives in each ad set (in other words, one ad in each set for a total of two ads). Keep the copy the same for each ad.
For the purposes of this example, then, our plan is straightforward:
To run the split test, you’ll need to set it up in Facebook. (If you’re a Marin customer, contact your account manager for help with this.) The test can be 50/50 or 33/33/33, etc., depending on the testing variables, but note that 50/50 is the most commonly used model. So, if an audience has 10 million people, the ad sets will have 5 million people in each audience.
From here, we select the image as the variable to test. Our main KPI is conversions (downloads), and we’re allocating $5,000 per ad set. As we mentioned above, our audience is 5 million per ad set. We’ll run the campaign for two weeks to ensure we have a broad reach, high budget, and long duration.
Since we want to see positive results before we extend our campaign to other markets, we’ll start only in the UK first.
Ready, set, test, measure.
When scoping your own split test, make sure that the test will have value for you, and that you’ll see clear results that you can use to refine and improve your campaigns.
The first questions you should answer include:
Back to that riddle—is it a boom or a bust? To determine which test worked best, choose the variable that has the highest efficiency level based on your objective.
In our example, our objective is conversions and the main KPI is downloads. So, we can consider the ad set that has the lower CPA as the best performing.
And there you have it. Easy, right?
Best Practices and Recommendations for Maximum Success
If you’d like to dig deeper (and we recommend that you do), here are a few best practices.
Define an acceptable confidence level
Before you create a test, determine an acceptable confidence level. Test with larger reach, longer schedules, or higher budgets.
Choose one variable to test
This allows you to define the exact difference in ad creative that drove better performance.
Define main KPIs before the test
This will allow you to determine the best performing variable.
Ensure both test sizes are comparable
When testing for volume metrics such as number of conversions, scale to ensure both test sizes are comparable.
Start testing on one specific market or campaign
This will allow you to monitor and analyze the test results more efficiently, which will in turn allow you to draw better conclusions. If you find this useful, you can conduct further tests on different variables and expand to other markets.
Test based on one large audience
The audience should be big enough to be split and to allow you to gain sufficient insights.
Allocate the same budget to the test groups
If you’re running your splits at the campaign level, make sure both campaigns have the same lifetime budget. If you’re testing on the ad set level, both ad sets should have the same lifetime budget.
No changes to the test groups
Any changes could compromise the split testing and prevent you from seeing clear results.
This is the final post in a series on transparency. In today’s article, we look at data transparency, why it’s important, and the elements of a transparent data model.
We’ve reached the last article in our series on transparency in programmatic display advertising. For those of you just tuning in, let’s quickly recap:
To wrap things up, this article looks at the importance of having access to all of your data—not just the numbers a publisher wants you to see. Let freedom—and transparency—ring.
When it comes to data visibility needs and desires, advertisers and agencies run the gamut—some are happy to pay on a CPC basis without knowing what CPM the vendor paid. Others are okay tracking to just a few KPIs. Still others want it all. What’s the best way to go?
The key to transparency is precision—whether you’re running a direct response or a brand campaign, having the right data measures the impact of either type of campaign, and allows you access to what you need to measure ROI.
So what do advertisers most often expect from their ad tech vendors and publishers? Our take is that list is long, but we’ll focus on the essential data you need to transform “unaware” to “X-ray vision.”
To gain greater intelligence about the effectiveness of your display campaigns and to better understand the customer journey and attribution, at minimum, you need access to:
This level of transparency allows you to understand true ROI. It also lets you keep an eye on how your bids are managed, unlike black box vendors who don’t reveal a history of bid calculations or otherwise provide any idea of what—and how—you’re actually doing. With the black box model, it’s much harder/impossible for you to look into the data and understand what’s driving conversions or an algorithm. Conversely, being able to do this provides you with a high level of visibility into the effectiveness of your campaign.
Despite industry calls for greater data transparency on many levels, mum’s still the word from most vendors and publishers. Some players have their reasons—the main one being to maintain a competitive edge. Understandable. However, as more and more brands pound at the locked doors of black box vendors, it’ll become less to those vendors’ advantage to keep the secret sauce secret.
If a vendor is transparent about not being transparent, is that “transparent”? We say “no.” As vendors realize this, perhaps there’ll be some level of agreement that to survive in an increasingly tug-of-war environment, neutrality and openness are key.
That is, companies must heed the call to “show me the data.”
In sum—the best way you can improve campaign performance is to know exactly how bidding, reporting, and attribution are all working independently and in concert. Once you have this level of data and insight, you’ll know exactly how your budgets are working for you (or not), and you can take active steps to enhance effectiveness. At that point, you’ll have a crystal-clear view of your marketing efforts—and your ROI—to give you the confidence, control, and independence you need to execute measurable marketing initiatives.
Holiday shopping’s in full swing. If you’re running retargeting campaigns, make sure they’re as prepared for the season as you are. Online sales are forecast to increase between seven and 10 percent over last year to as much as $117 billion.
We made your list, so check it twice, and take these steps to boost campaign performance during the holiday season.
You’re likely going to see a boost in site traffic (especially if you sell anything that can be given as a gift), which means you’ll see a boost in impressions served and in advertising funds spent. Make sure your campaigns have a proper budget set to guarantee you have enough ad money available for the day, so that you don’t miss out on these potential new customers.
We recommend a 25 to 50% budget increase for the holidays, but you know your site traffic best. Whatever percentage of traffic increase you’re expecting, boost your budget about that same percentage.
Almost all advertisers will increase their spend for the holidays, so you’re going to have serious competition.
With so many advertisers fighting for ad space, it’s not uncommon to see your CPM costs rise during this time of year. To prepare for this surge, make sure you increase your CPM bids across your campaigns. Bidding higher will make your campaigns more competitive and will give you a better chance of serving more ads by winning more impressions. We suggest increasing your CPM bid by 50-100% of the current average CPM cost for the campaign.
Holiday-themed advertising only gets people’s attention during one time of the year, and you should join the conversation your customers are having. Using ads that mention specific events like Black Friday, Cyber Monday, or any of the major holidays can grab a visitor’s attention.
Send a happy holiday message, mention that there are only X number of shopping days left, and give them a reason to click your ads. Use the holidays as a chance to create urgency and you could see a boost in clicks and conversions.
We hope these suggestions are helpful and lead to a profitable holiday season for you and your business. As always, please feel free to contact us with any questions or comments.
From our team to yours—happy holidays!
We’re headed into another peak retail season, which runs from now through Christmas Day. Considering not-so-recent trends in Shopping and mobile, many marketers are hedging their bets on this being the biggest online retail season yet. Preparation is key, and understanding what went well and what didn’t last year—and when it did and didn’t—will help guide decision-making in the coming weeks.
When the volume is so high, each day could make or break the quarter. Here are three things you should be doing on a day-to-day basis to increase the likelihood of favorable outcomes.
Your buyers should have a list of products they expect will be major sellers this season. These could be products where inventory is so deep no one can compete, or buyers purchased at a bulk rate and can offer the best pricing.
Work with your buyers to understand what these products are, and optimize them on a per-item basis with SKU-level product groups in a High Priority campaign. Monitor these daily and keep an eye on inventory—when they start to sell out, pull back so that you don’t end up aggressively pushing nearly sold-out products.
In addition to the proactive management of products you’re bullish about, the high volume is going to yield insights of its own. Monitor your broader product groups—defined by Category, Brand, Custom Labels, etc.—for segmentation opportunities.
You’ll start the season with a single bid for a Brand product group. But, as volume dictates, some products or sets of products within the group will warrant segmenting and assigning a new bid based on how they’ve performed to date. This is a crucial step to optimizing and hitting performance goals on an ongoing basis.
As you structured your campaigns, you established the levers and switches you’re going to use to effectively manage your product mix and hit performance goals.
The most important pieces of all this will be to understand how you want to bid these levers and how to stay on top of everything. Be considerate of sales, key dates, top products, and inventory / stock levels. A combination of proactive strategies (e.g., Brand X is 20% off next week) and reactive strategies (e.g., Brand Y is selling amazingly well over the past week) will be necessary to generate the best results.
Be aggressive where you expect the best returns and don’t hesitate to pull back on things that aren’t producing. Good luck!
In PPC, there are two main approaches when it comes to bidding workflow—manual and automated. Over the years, there’s been debate among search marketers on the pros and cons of each approach. Search marketers have differing opinions on which yields the best outcomes.
One of the main arguments in favor of manual bidding focuses on the control that it affords the search marketer, in contrast to the hands-off nature of automated bidding inherent with publisher bidding—like AdWords “Smart” bidding and most (but not all) 3rd party proprietary bidding algorithms.
In nearly all automated bidding approaches, the search marketer sets a goal and the bidding algorithm reviews historical performance, and then calculates a bid with limited transparency from start to finish.
The apprehension some search marketers feel towards automated bidding derives from the opaque nature inherent in most approaches. This fear is realized when a campaign is underperforming, and the search marketer becomes at a loss for what’s amiss, or how to improve it.
Putting that fear aside, let’s reflect on the many benefits of automated bidding, which is the reason for its proliferation.
Here are just a few.
Automated bid management is a huge time saver. Think about it—how long would it take you to manually change a million keyword bids? How confident would you be that each bid is optimized to maximize your return?
If you’re being honest with yourself, the answers to those questions should naturally steer you towards automation as the optimal solution. Automation augments the search marketer by executing repetitive tasks, serving as an ‘enabler’ for the search marketer to focus on growth opportunities or account strategy while keeping tabs on daily performance.
Automated bid management platforms produce accurate bids through regression modeling that looks backwards to predict future outcomes. With millions of dollars at stake, these algorithms are typically built with risk aversion at their core to produce low error rates. By their very nature, they make changes at scale that’s quite literally impossible for any individual, or even team, to compete with.
The reality is, sophisticated marketers with material budget use an algorithm to bid on their media today. If you aren’t, you’re putting yourself at a disadvantage.
Automated bid management platforms allow advertisers to define the goals and milestones for the algorithm to work towards. The marketer remains the operator and the brains of the operation, with the bidding algorithm working as his proxy.
Learning from massive datasets to create better future outcomes is at the heart of bidding algorithms. Today, this type of mathematical analysis is popularly called “machine learning” and “artificial intelligence.” Most ad tech companies have years of experience with these techniques, but largely fly under the radar in popular press, with newfangled applications like self-driving cars getting the headline coverage.
So, how do you get the best of both worlds? Simple—employ automated bidding with full transparency. That’s not an oxymoron. That’s a real thing offered by a few leading independent marketing partners (not to toot our own horn, but Marin Software is one such example).
Fully transparent bidding solutions (i.e., the bidding system shows you the step-by-step logic of the bidding algorithm) allow users to see all the details behind their bid calculations for each keyword. This includes the bidding model(s) employed, the details of the dataset used, performance bumpers activated, and any other pertinent details behind the decision-making. If automated bidding is fully transparent, many of the arguments opposed to automated bidding lose their heft.
Information Available in a “Fully Transparent” Bidding Solution
The level of information available for each keyword in a “fully transparent” bidding solution varies. That said, at Marin Software, we show the logic of our algorithms “line by line,” which allows users to see a full breakdown of bidding decisions, including:
Contrast this to the information displayed in a “black box” bidding solution:
Fully transparent bidding solutions allow PPC managers to review the logic used to reach a bidding conclusion. In addition, the search manager has the option to overlay bidding rules to ensure the algorithm behavior is consistent with their risk tolerance and strategy to hit certain goals and milestones.
The best fully transparent bidding solutions also allow you to preview bidding calculations before they’re pushed to publishers, and manually override bids on specific keywords if needed. This gives PPC managers the full control of manual bidding with all the time saving, efficiency, and data processing power of automated algorithms.
If automated bidding isn’t currently part of your strategy, we hope this post helps break down the nuances of different approaches. Although it also explains the pros and cons, it advances the argument that if you aren’t using a transparent bidding algorithm in today’s environment, you’re hamstringing yourself, because it’s near-certain that your competitors are employing an automated method of bidding to try and out-compete you. If you’d like to learn more about Marin Software’s approach to bidding, click here.
Shopping doesn’t end after the holidays—according to a National Retail Federation survey, 65% of shoppers plan to keep shopping after Christmas. Use this time of year to convert them to loyal customers with new demand generation and cross-sell opportunities.
Don’t let your holiday campaigns go to waste—keep aiming for more purchases. Post-holidays is a great time to re-engage to drive demand.
For more tips on winning the holiday shopping game, download our Social Advertiser’s Holiday Guide.
You’ve got your product feed set and sending to Merchant Center. You’ve created campaigns mapped from the attributes that make the most sense for your style of management. And, you’re pretty familiar with the set attributes you can use to define product groups:
|Item ID||Brand||Category||Product Type|
|Condition||Channel||Channel Exclusivity||Custom Labels|
What about those custom labels, though?
Google allows for up to five custom labels per feed. Here are some effective and creative ways you can use them.
One clever way to use a custom label is to group SKUs together under a Parent ID. Think of this as a “Parent” and “Child” relationship where we group the Child SKUs under a Parent label.
For example, suppose you offer a coffee mug in different colors. Each variety has its own unique Item ID. You could assign the same “Parent” identification number to each of the variations, and then set a product group to define the set of the products—as opposed to having to map these individually by Item ID.
Custom_Label 0: 1234
Custom_Label 0: 1111
Custom_Label 0: 1213
Another interesting way to apply a custom label is to group products by their price point, or average order value. For example, you could apply values of “High,” “Medium,” or “Low” to products based on where they belong in the overall product mix. In your campaign, you could then segment out Product Groups based on these values, and utilize these levers to bid more or less aggressively on products based on their anticipated return and volume.
Categorizing products by how (or whether or not) they sell according to time of year can be a wise consideration depending on the nature of your business. Indicating these designations in the product feed allows you to increase or decrease bids across a range of products classified by time of year.
So, in the final days of summer, you could increase bids across all seasonal products and then, right after, immediately bid down on those after the heat breaks.
Custom_Label 1: Winter
Custom_Label 1: Summer
Custom_Label 1: Fall
These are just a few ways to apply custom labels—and give your campaigns levers to bid and segment on—that are completely specific to your account. The main takeaway is to use them.
Brainstorm what makes sense to be able to dictate volume by (whether or not a product is in peak season, whether or not it brings a high yield per order, etc.), then apply the values in the feed and create the defined product group in AdWords. Custom labels are the most flexible of the attributes available, so test them out and see what helps shift the needle in your overall Shopping performance.
Dynamic Ads enable you to automatically promote your entire catalog across devices. With Dynamic Ads, you have full control over the products you advertise, ensuring you’re reaching audiences who’ve expressed high intent to purchase with the most relevant products. Advanced tactics are also available—cross-sell, upsell, and even prospecting.
For more tips to make the most of your holiday ad campaigns across social media, download our Social Advertiser’s Holiday Guide.
SEM has long been considered bottom-funnel advertising, but Facebook’s new and improved direct-response features mean you can turn the platform into a revenue machine over the coming weeks. Here are five essential direct response (DR) features available on Facebook—and how to use them to your advantage.
With Q4 bringing a high saturation of ad demand—especially ecommerce and retail for the holiday season—carousel ads are a great way to stand out. With these ads, you have the ability to tell a story about your brand or product. You can also showcase the different benefits of a signature product, or a range of products or product lines. More real estate means more opportunity to grab attention and gain potential new customers.
If your focus is on lead gen, you should be using lead gen ads. The benefit here is that the user doesn’t need to leave Facebook to fill out your form. There are tons of fields to choose from—Facebook will try to fill them in automatically based on information from the user’s Facebook profile (which makes it a lot easier for someone to complete the form). You also have the option of putting up to three custom questions on the form that are specific to your company.
Wondering what creative works well? You can use carousel ads with this campaign objective to help tell your company’s story, list different benefits or services, etc. (Yes, we’re fond of carousel ads.)
If you’re worried about spending a ton during a highly competitive season (remember, with Facebook ads, you’re competing for space, not keywords), we recommend establishing a test budget to figure out which messaging works, then open the budget spigot in Q1.
If your focus is on app installs, Facebook now has a great new type of bidding to use for app install ads. You can now bid on in-app events such as a registration or purchase. Your cost per install may rise with this, but you’ll be targeting people who are more likely to complete the in-app event that’s most important to you.
Trying to figure out who to target this season? Lookalike audiences should definitely be in the plan. With lookalike audiences, you can target from a 1% lookalike all the way to a 10% lookalike. A 1% lookalike will be the percentage most closely like your original seed audience, but once you find a seed audience that works, you can expand the percentage in order to scale more. Fee seed lists we see work well on Facebook are:
To go even deeper into lookalike audiences, make sure you caught Marin’s detailed article.
Last but not least, make sure you’re confident in your pixel implementation! Once the pixel is placed, make sure you go through the user flow on your site a few times and test the pixel. You can use the Facebook Pixel Helper Tool extension on Google Chrome. With this tool, you’re able to see if the right event is firing on the page you want it to. Since you’re likely bidding oCPM for conversions (in order to maximize the number of conversions you can get), you want to ensure you’re optimizing on a conversion that’s firing correctly.
Make sure you use all of the above features throughout the year, not just in Q4. But, to compete on Facebook and show some return for your dollar from now through mid-December, it’s imperative to use what’s available to you. These five features are a great start.
Lookalike audiences are thee Facebook feature when it comes to audience targeting. If you’re looking to scale your campaigns and more, it’s a must-consider option.
You can use lookalike targeting to find similar users to your core audience based on interests, click behavior, and conversion habits. The smaller the percentage of your core audience, the more similar your lookalike audience will be.
You have several options from which to generate lookalike audiences:
As you can see, you have a lot of choices to test different audience types and associated performance. The key challenge is to segment and structure the audiences to avoid overlaps and achieve the best delivery.
Something to note: Since frequency caps limit the daily number of times you can deliver an ad to a user, lookalike audiences won’t increase your overall reach. And, you’ll have less predictability when it comes to which ad wins each auction.
There’s a way to overcome these challenges, however. Make sure your strategy includes nested lookalikes and smart exclusions. Let’s go into more detail.
Let’s start with an example, where we exclude the next-highest percentage audience from our targeted lookalike audience. So, if you’re targeting lookalike 3% and lookalike 5%, then exclude the 3% audience from the campaign that’s targeting the 5% one.
With smart exclusions, we exclude the targeted audiences that we’re already using in other live campaigns. For example, if you’re running campaigns with 1% lookalike and 3% lookalike and want to launch a broader targeting campaign, then exclude the 3% lookalike.
When you’re planning your targeting strategy, make sure you’re segmenting your lookalike thresholds according to the value of the user, and excluding the targeted audiences from campaigns to avoid overlap. This’ll allow you to use lookalike audiences from different sources, increasing the overall reach and scalability of your campaigns.
For example, if you’re running a retargeting campaign based on a Website Custom Audience of all your site visitors, exclude this campaign from all of your acquisition initiatives, along with the associated lookalike audiences.
Here’s another scenario. Suppose you’re a travel website and the user funnel includes two conversions—registration and booking. You would segment the audiences based on your goals—perhaps based on the custom audience of the previous month’s bookers, conversion pixel data, and Website Custom Audience of people who registered but didn’t book. Your segmentation would look like this:
Custom audience segmentation:
You can use all of these audiences for your acquisition campaigns, along with interest-based and other targeting options.
Here’s the final campaign planning structure for this example. This takes into account that retargeting campaigns are running based on your Website Custom Audiences.
Fine-tuned campaign planning structure:
Creating effective lookalike audiences takes a bit of cunning and patience, but it’s not rocket science. With continued practice, refinement, and measurement, you can scale your campaigns to ensure you’re targeting audiences with the most relevant ads at the most relevant time, in a way that works the best for your business. If you haven’t yet implemented this feature, we strongly recommend you get started today!