219: A Safety Net for When I Guess Wrong

Within the span of several weeks, I’ve advertised not one but two complete liquidations of antique stores (for different clients in different states). Both had quality vintage items. Both had catalogs brimming with a large number of lots. But only one became a prime example of why I prefer to advertise with a safety net.

I launched the ads for the first antique store auction on a Thursday afternoon. When I checked my Facebook analytics the next morning, I did a double take. One of the ads was getting results at $1.10 more per click than the other ads in the campaign. I screen captured the stats and emailed my client. I asked if they’d like me to shift some of its budget to the other ads or if it was worth paying extra for these prospects’ clicks. I rarely hear how the estate and antique auctions I advertise fare, and often big-fish buyers in other asset categories are worth the extra cost to get them on the line. (Also, I didn’t have a lot of personal property experience in that part of their state.) So, I leaned on the client’s experience.

The answer came back to let the ads run as they were originally budgeted. When the campaign closed, that ad finished with a cost per click 19.5 times that of the other ads and a cost per landing page view of 20.3 times all of the other ads. The click-through rate finished at a meager 0.66%, while the other ads averaged 7.65%.

So, who were we targeting with that expensive, inefficient ad?

People with occupations connected to antiques. The line of thinking: the prospects most likely to purchase a store full of high-end antiques are people who already own or work in another antique store. We guessed wrong. Thankfully, the other ads carried the day; and the campaign averaged 7¢ per click, even with the expensive ad in the mix. In fact, the most efficient ad for that auction didn’t target any interest categories related to antiques or collecting.

My client wouldn’t have guessed that. But Facebook’s artificial intelligence engine would. And did.

Robots to the Rescue

The longer I advertise on Facebook, the less I trust my strategic instincts and the more I lean on data. I get surprised just about every week by which ads do better than others in the same campaign. Same photos. Same headlines. Just different audiences. Audiences I expected to jump on ads don’t. Prospects that would make sense to most people with any marketing experience prove themselves indifferent to the quality items shown in professional pictures. I can’t tell you how many times I’ve A/B tested images and been shocked by what pictures proved the best bait. Hint: they weren’t the “money shot,” the brochure cover photo, or the main image on the client’s website.

A global artificial intelligence engine, fed by tens of thousands of impressions and thousands of clicks per auction, regularly surpasses my educated guesses. Three days of machine learning can outperform strategy informed by almost 8,000 auctions advertised over 20 years across 49 states. It’s humbling to be proven wrong. At the same time, it’s exciting. That means we as marketers can now adapt our advertising in real time to fickle realities and head-scratching trends. 

Robot Consultants for Print Media

I still have clients who refuse to leverage paid social media advertising. Others only throw spare budget at it. They either don’t believe it can bring them buyers or have yet to see a buyer come from it. I can’t argue against their experience or results, but I offer this counter thought: what if you used Facebook to learn what kind of audiences respond to your ads, what kind of headlines work best, and what images get the most people to your website? What if you then adapted your other media to what’s working in your Facebook ads? Your print media, signs, and email marketing could grow more effective because of the immediate feedback that social media ads provide.

Recently, a client asked for my opinion on a change to a tiny plat map in several small newspaper ads. I told her that I wouldn’t recommend using plat maps at all—particularly small ones—in newsprint. I’ve learned from Facebook ads that people typically respond to ground photos, drone shots, and oblique imagery at higher rates than plat maps, orthogonal aerials, and multi-parcel outlines. (Tract layouts are tertiary information for buyers; they don’t need to know boundary shapes until they decide they are interested in what’s inside them.)

It’s just opinion until you have data. 

Train Your Own Robot

I’m not telling you to avoid using aerials. I’m suggesting that you A/B test them yourself. Test photos vs videos, collages vs slideshows, ads vs promoted posts. Test headlines. Test spending more in the first half of a campaign vs the second half. Test whether you should advertise to past web traffic or past Facebook interactors or neither.

This pragmatism most helps the auctioneers who focus on the same one or two asset categories. It’s easier to see trends when most of your auctions are like-kind. If that’s not you—if you are a generalist or your specialty is a geographic area more than an asset type—I highly recommend you team up with a specialist. That can be a broker, a dealer, a journalist, a collector, a vendor, or another auction company. The added expense on your current auction could give you tremendous insight for future auctions. It might even pay for itself on the auction at hand because guessing wrong can be expensive. 

Thankfully, Facebook’s robots can make mistakes less costly—if we let them do their work. If we’re humble enough, we can have a safety net for the high wire act of advertising auctions. 

—

Stock images purchased from iStockPhoto.com