insights / February 11, 2021

You likely don’t know who your target customer is. And that’s fine.

Kevin Harwood

The speed at which marketing technology is accelerating has no doubt left many with severe whiplash as the tools and processes continue to change how customers are found online.

Even my head is spinning (and I’ve been working with online ads for ten years). Why? Because of the sheer power of ‘unattended’ machine learning to drive better results. That’s why we launched Apparatus: to leverage its power to generate leads, conversions, and revenue for clients.

One of the most common conversations that we have while onboarding our clients is when they tell us “we really know the demographic profile of our most profitable customers, can you make sure we target them?”. It’s a very well meaning conversation because the client is telling you exactly who to go after. And historically that’s been of great value.

But the last few years have yielded a consolidated strategy and best practice known as Look-Alike Modelling (LAL) for finding your target audience online. After five years of applying LAL, I’m now unconvinced that a client knows who they want to target; only who they were targeting before.

My experience with a large ‘safety-oriented’ Canadian SaaS company provides a recent, relevant example. Their direct and outbound marketing success over the years had been based exclusively on call centres and direct mail campaigns. Like so many other companies they were ‘once burned, twice shy’ when they dipped their feet in the digital waters to abysmal results. Meanwhile, their market share gradually declined year-over-year as new players entered their space. That fact – plus it was getting too expensive to produce new leads for their team using non-digital means – prompted them to overcome their shyness and try digital again.

During our onboarding call, they provided a detailed list of their customers they’d acquired over the years along with their customers’ industries including construction, extraction, electrical, plumbing and even resource exploration. Naturally we aggressively began targeting those markets across multiple channels.

The early results were fine though very much on par with their offline marketing experience. So after the first month of exploring the market through digital advertising, I suggested that we target more broadly and include users we may have missed or assumed didn’t exist. More specifically, I suggested we experiment with a LAL look-alike model of their existing customer base and find users who were similar to them, regardless of their industry.

Everyone that works in digital marketing will likely endorse my view that look-alike match technology is one of the most consequential tactics to hit the market over the last decade. I can’t overstate how amazing it is, nor can I understate my passion for it.

In the below chart you can see (in blue) the week-to-week results of this engagement with a complete focus on look-alike audience technology over demographic data.

Look-alike chart

Apart from a major dip during the Christmas season when everyone was disengaged from work, there’s been a consistent climb to new all-time highs on a weekly basis. Happily we now find ourselves in a situation whereby the client conversation has evolved from “what’s the minimum we can spend online to get results?” to “what’s the maximum amount of budget we can give you?” A big swing in a really short time!

There were two additional interesting outcomes in this engagement. First, our client accounts have been used by LAL for targeting users who have a background in “safety” with job titles like “environmental safety compliance officer” rather than just involvement in broad career or industry groups that we were originally targeting. Second and even more interesting, we found there was a complete change in the demographic makeup of our ideal lead, who is now 10 years younger than our original assumption.

It’s all upside, right?

For clients, yes. For me not so much. I confess that the campaign architecture and design is almost too simple. It challenges my preference for greater sophistication and strategic acumen. Almost every one of this client’s accounts looks similar to the simplicity of what you see below.

Look-alike accounts

That graphic is the actual design framework we leveraged for the SaaS client. Now the crazy part is that the “1% LAL” audience above has no other modifiers or audience/demographic parameters and it’s delivering the best CPA for the entire account. These “naked LALs” ad-sets have become the bane of my life, because I can’t beat them with better targeting or more sophisticated inclusions.

Below you can see how many instances we’re running this design in. Believe me, I am trying to find a more sophisticated architecture. But it just keeps winning!

Look-alike ad sets

I can’t quite articulate the humour in this, but there’s a certain irony of being dissatisfied with the aesthetic design of your account ‘build’, despite the client’s resounding praise for the results. I don’t know — maybe at the age of 30 I’ve become cantankerous. But designing new creative architecture builds is a passion for me. So the effectiveness of such a simple strategy feels offensive to me.

So I’ll live with it, but I don’t have to like it. Happily though, our clients like it a lot.