While working at an ecommerce company, we ran Google Search Ads that seemed modestly profitable but paled in comparison to our Smart Shopping campaigns.

To test the true impact of Search Ads, we turned them off completely. Analytics showed customers typically took about 14 days from first visit to purchase, so we knew that by doing this, there was a chance we wouldn’t notice any impact for the next two weeks.

As expected, sales remained steady – people already in our funnel continued converting. But at the two-week mark, overall sales dropped significantly and stayed down.

After a week of lower performance, we reactivated the Search Ads. Two weeks later, sales returned to previous levels. While direct attribution reports had shown minimal impact from these ads prior to our test, the experiment revealed they were actually driving substantial revenue that wasn’t being properly attributed.

As seasoned digital marketers, we went into this test knowing the likely outcome. To someone who’s only just heard enough about attribution to be able to spell it, you’d be veering off into the dark.

Used badly, attribution becomes a source of vanity metrics, circular debates, and finger-pointing over whether Meta, Google (or even Shopping vs Search) gets the win. Used well, it gives you something rare in marketing: a nudge in the right direction.

But here’s the key point most performance marketers forget: Attribution only works when you know when to care, and when not to bother.

Attribution is a Model, Not a Mirror

Every attribution model is just that: a model. A simplified story about what caused someone to act.

Here’s how the most common ones break down:

• Last-click attribution gives all the credit to the final touchpoint before a conversion. Simple, but reductive.
• First-click attribution credits the first interaction, even if the conversion happens weeks later. Good for top-of-funnel insight.
• Linear attribution splits credit evenly across all touchpoints. Sounds fair, but often dilutes insight.
• Time-decay attribution gives more weight to recent touchpoints. More realistic, but still oversimplified.
• Data-driven attribution (often Google’s default) uses machine learning to assign credit based on historical patterns. Powerful, but a black box.

Each model tells a version of “what worked,” but none of them tell the whole story.

They’re not reporting facts.

They’re reporting theories.

Most attribution is just an educated guess – one that often gets taken way too literally.

The More Complex the Journey, the Worse Attribution Gets

Let’s say someone:

• Hears about your brand on a podcast
• Sees your logo on a mate’s hoodie
• Watches a TikTok review
• And Googles you three weeks later

Google will take credit. Meta might too.

But neither created the demand.

Most attribution tools only measure what’s trackable. That means:

• Offline influence is invisible
• Word-of-mouth gets ignored
• Brand exposure gets zero credit

And yet, those are often the most powerful drivers in your funnel.

So When Should You Ignore Attribution?

Perhaps “ignore” isn’t the right word. You should always be aware of what your attribution is telling you. But that doesn’t mean you need to act on everything you see.

Attribution can become misleading when:

• Customer journeys span multiple non-digital touchpoints
• Campaigns are largely impression-based, such as brand awareness campaigns that aren’t designed to drive site traffic
• You have small sample sizes
• You’re trying to measure offline or delayed conversions
• You’re spending next to nothing on marketing,
• You’re only running one marketing channel

In these cases, attribution doesn’t give insight, it gives false confidence. Reports, not reality.

But Can’t You Measure Brand Campaigns?

Yes. Brand campaigns on Meta or Google do come with metrics: impressions, reach, even view-through conversions.

There’s even one called “Brand Recall Rate” that largely comes off as a made-up number, and Meta’s description of what it does confirms that.

But brand awareness isn’t about what someone saw. It’s about what they remember when it counts.

Attribution stops measuring the moment someone converts. Brand-building often starts working before that moment.

If you judge brand work by attribution alone, you’ll kill the campaigns that make future sales possible, because the report won’t show their true value.

An Attributable Problem

It’s one of the biggest problems in digital marketing today.

Because branding and traditional campaigns don’t offer an attributable Return on Investment, those campaigns are canned quickly in favour of conversion-driven campaigns that do.

The problem with this is that only about 5% of your audience is ready to convert at any given time. The remaining 95% may not have even realised you exist, let alone realised a need for your product.

Conversion-driven campaigns won’t help you here, no matter what your attribution modelling is telling you, so we end up with bland BUY IT! BUY IT NOW! ads that fade into noise.

It’s marketing by desperation – there’s no creativity, no story, no differentiation. Everything has a CTA and promotes urgency, but if you hide the name of the brand, the ads are largely interchangeable with any other business.

Interchangeable as in ‘forgettable’.

So businesses double down on the conversion-driven campaigns by spending more.

Your attribution modelling tells you this is working, but your bank account is telling you it isn’t. So you scale back your budgets or end some campaigns, and guess what? Your sales die on their arse.

All because your attribution didn’t take one thing into account:

“Not everyone that knows your name will convert, but everyone that converts knows your name.”

So When Is Attribution Actually Useful?

• When you’re working with high-volume, short-lag campaigns,
• For longer-term campaigns (between 30-90 days), provided you understand the longer the data is stored, the more holes that will start to appear (e.g. cookie length, browser history clearing, different devices, etc.)
• When you’re comparing performance between clear, distinct campaigns,
• When you want to optimise a specific lever, not justify your whole strategy.

The Real Risk: Chasing What’s Measurable Instead of What Works

If you only trust what attribution tells you, you’ll double down on short-term channels, and slowly starve the things that build demand.

That’s how you get campaigns with great ROAS… and stagnant revenue.

Attribution tells you what you harvested, not what you planted.

Want to Know What Actually Works? Measure Incrementality

Incrementality asks: What changed because of our marketing?

It compares:

• Group A who saw the campaign
• Group B who didn’t

And then looks for a difference.

Just like in the example from the intro:

• Direct attribution reports showed minimal impact from Search Ads, but turning them off revealed they actually drove substantial revenue
• There was a clear two-week lag between ad visibility and conversion impact, demonstrating how attribution can miss delayed effects

It can be more effort and higher risk. But it gets you closer to causality, not just correlation.

Want to Zoom Out? Try Media Mix Modelling

Attribution looks at user-level actions.

Media Mix Modelling (MMM) looks at channel spend and outcomes over time.

You collect:

• Weekly or monthly spend per channel
• Sales or conversions over the same period
• Context like promotions or seasonal changes

Then you look for patterns:

• “Every time we spent more on Meta, total sales increased.”
• “When we paused YouTube, site traffic dropped.”

You don’t need expensive software. A spreadsheet with 12–26 weeks of data is enough to spot signals.

MMM won’t tell you who clicked. But it will tell you what’s driving growth.

Don’t Have the Patience for MMM? Try The Marketing Efficiency Ratio (MER) Instead

The Marketing Efficiency Ratio (MER) is a key metric that measures the overall effectiveness of a business’s marketing efforts by comparing total revenue generated to the total marketing spend.

The formula is simple: MER = Total Revenue / Total Marketing Spend

For example, if you spent $10,000 on marketing in June and generated $50,000 in revenue, your MER would be 5.0.

The beauty of MER is its simplicity and big-picture focus:

• It cuts through attribution disputes between channels
• It forces you to look at marketing as a holistic system
• It prevents channel-specific optimisation that hurts overall performance

While MER won’t tell you which specific tactics are working, it will tell you if your marketing engine as a whole is becoming more or less efficient over time.

If your MER is healthy and is largely growing month-on-month, then you don’t need to pull apart your marketing with attribution modelling – the combination of what you’re doing is working. Enjoy it.

But if your MER is stagnant, shrinking, or makes your eyes water, then attribution can help you identify what part of your marketing is working, and what part isn’t.

Smart Marketers Don’t Just Optimise, They Think Critically

They use:

• Attribution to test tactical & strategy changes
• Incrementality to validate channel impact
• MMM to guide budget
• Brand tracking to measure mental availability

And they know that not everything worth doing shows up in a dashboard.

Attribution Is a Tool, Not Your Whole Strategy

Attribution has its place. But it’s not a source of truth. It’s just one lens.

If you rely on it too much, you’ll optimise for short-term wins and lose long-term impact.

Use it when it helps. Ignore it when it doesn’t.

And remember: The real skill in attribution isn’t picking the right model. It’s knowing when to bother at all.