What is marketing attribution? Ask ten marketers and you'll get ten different half-answers, usually involving the word "ROAS" said with a shrug. Here's the plain version: attribution is the process of figuring out which touchpoints, which ads, emails, or posts, actually deserve credit for a sale.

That's it. Sounds simple. It isn't, because most Shopify and GA4 dashboards default to "last-click" thinking: whatever the customer clicked right before checkout gets 100% of the credit. Everything before that click, the stuff that actually built the desire to buy, gets nothing.

Picture this path: a customer sees a TikTok ad on Monday. Three days later, she clicks a Meta retargeting ad but doesn't buy. On Friday, she googles your brand name and converts. Three touchpoints, one sale. Last-click attribution hands the entire win to branded search, which is basically the internet's version of taking credit for someone else's setup.

Why Attribution Matters More Once You're Spending Across 3+ Channels

Run one channel and attribution barely matters. Run TikTok, Meta, and Google at the same time, and the model you use starts deciding where next month's budget goes.

Without attribution, brands keep funding the channel that happens to close the sale, usually branded search or retargeting, and starve the channels that actually created the demand. TikTok and Meta prospecting look expensive because they rarely get the last click. So the budget shifts away from the exact channels doing the heavy lifting, and growth stalls without anyone knowing why.

It's gotten worse since iOS 14.5. Platform-reported ROAS from Meta and Google is increasingly a number the platform wants you to believe, not a number you can act on alone [VERIFY current post-iOS14.5 platform ROAS reliability claims before publishing]. Signal loss means the ad manager dashboards are optimistic by design.

None of this means attribution gives you the truth. It doesn't. What it does is cut down the guesswork enough that you can make a budget call with some confidence instead of flying blind. That's the actual bar to clear, not perfection.

The Main Attribution Models, Explained Simply

Every model is a different set of rules for splitting credit. None of them are "correct," they're just more or less useful depending on what you're trying to answer.

Last-touch

  • What it measures: Credits the final click before purchase
  • Good for: Simplicity, it's the default in most tools
  • Bad for: Anything upper-funnel, it systematically undervalues TikTok and Meta prospecting

First-touch

  • What it measures: Credits the very first interaction in the journey
  • Good for: Measuring discovery and awareness channels
  • Bad for: Ignoring every touchpoint that came after, including the one that closed the sale

Linear and time-decay

  • What it measures: Splits credit across every touchpoint, either evenly (linear) or weighted toward touchpoints closer to purchase (time-decay)
  • Good for: A more balanced view than single-touch models
  • Bad for: Still a rule you're imposing on the data, not something the data proved

Multi-touch attribution (MTA)

  • What it measures: Stitches together the full click path using pixels and UTMs across platforms
  • Good for: Granular, journey-level detail
  • Bad for: Breaking down the moment a user switches devices or blocks cookies

Media Mix Modeling (MMM)

  • What it measures: Statistical relationships between spend and outcomes, no user-level tracking required
  • Good for: Privacy-restricted environments where pixels and cookies keep failing
  • Bad for: Speed, it needs a lot of historical data and updates slower than MTA

How Attribution Data Actually Gets Collected

Before any model runs, data has to get collected, and this is where most of the "attribution is broken" complaints actually originate.

The sources: UTM parameters on links, pixel and cookie tracking on your site, platform APIs pulling reported performance from Meta, Google, and TikTok, GA4 event data, and post-purchase surveys asking "how did you hear about us."

Every one of those has holes. Cross-device journeys (ad on phone, purchase on laptop) break the thread. Ad blockers strip out pixel data before it fires. iOS privacy restrictions mean Meta can't see what happens after the click half the time. Delayed conversions, someone clicks today and buys three weeks later, fall outside most tools' default lookback windows entirely.

Here's the point people skip: attribution accuracy is a data-pipeline problem first, a modeling problem second. A brilliant model built on gappy, mismatched data still produces a wrong answer, just a confident-looking wrong answer. Most tools jump straight to modeling because that's the sellable part, and quietly skip the pipeline work that would make the model worth trusting. Getting GA4 event tracking actually clean is unglamorous, and it's the part that determines whether anything downstream is usable.

Where Most Attribution Tools Fall Short (Including Northbeam)

Northbeam built its reputation on multi-touch attribution for paid social and DTC brands specifically, and for that narrow job, it's a known name in the space.

But the complaints that follow that category are consistent: a high price point, a scope that stops at attribution, and limited visibility once you need to tie ad spend back to inventory, margin, or Amazon and marketplace data [VERIFY current Northbeam pricing/feature set before publishing]. You get a clean-looking dashboard of channel credit with no way to answer the next question, which is whether that channel is actually profitable.

That's the real gap. Attribution numbers in isolation tell you which channel drove the click. They don't tell you whether that sale made money once you account for margin, discount codes, shipping cost, and returns. A channel can win every attribution model and still lose money on every order. If you want the full side-by-side on how Northbeam, Polar, and Trivas differ on this, we've broken it down here.

How Trivas Approaches Attribution Differently

Trivas doesn't treat attribution as a standalone bolt-on. Amazon, Shopify, Meta and Google ad data, and GA4 funnel data all land in one Redshift-backed warehouse, which means attribution numbers sit next to margin and cost data instead of living in a separate tab you have to reconcile by hand.

On top of that sits Wingman, the AI layer that surfaces which channels and campaigns are actually driving profitable conversions, not just whichever one happened to get last-click credit. The difference matters: a campaign can look like a top performer in a standalone attribution tool and still be a net drag once fulfillment costs and margin come off the top.

Attribution also isn't just a rearview mirror. What worked last quarter, by channel and by campaign, is the input that feeds forecasting: knowing what drove profitable growth historically is what makes next quarter's budget allocation a calculated bet instead of a hunch.

Getting Started With Attribution the Right Way

Don't start by shopping for a modeling tool. Start with UTM discipline, tag every campaign consistently, every time, and get GA4 event tracking actually firing correctly. Skip this step and any tool you buy afterward is just applying math to bad inputs.

Then pick one model, time-decay or MTA are the two reasonable starting points, and stick with it for at least a full sales cycle. Model-hopping every few weeks feels like optimization, but it mostly just creates false signal: the numbers move because the ruler changed, not because performance did.

If you're a marketing leader trying to get past attribution-only dashboards and into a view that ties spend to actual profitability, start with what BI reporting looks like for teams like yours, built for marketing leaders who need the margin math included, not bolted on later. When you're ready to see it against your own data, start a trial.