Marketing attribution is broken — here's what to do instead

If you run B2C marketing at any meaningful scale, you've experienced the attribution problem. You spend £100k on marketing in a month. You get 5,000 new customers. When you add up what each channel claims to have driven, the total is 8,000. Every platform is taking credit for the same conversions.

This isn't a bug—it's a feature. Ad platforms are designed to make themselves look good. Google counts a conversion if someone clicked an ad in the last 30 days. Meta counts if they saw an ad in the last 7 days. Both claim the same customer. Your internal last-click model gives credit to whichever touchpoint happened to come last. None of these are "right."

So what should you actually do about it?

Why last-click attribution is misleading

Last-click is the default for most companies because it's simple. The last thing someone clicked before converting gets 100% of the credit. Easy.

The problem is that it systematically undervalues channels that create demand and overvalues channels that capture it. Brand advertising, social media, content marketing, and PR often introduce people to your product—but they rarely get credit because someone Googles your name and clicks a brand search ad before signing up.

The result: you keep cutting the channels that create demand (because they look expensive on a last-click basis) and keep investing in the channels that capture it (because they look cheap). Over time, you hollow out your top-of-funnel and wonder why your bottom-of-funnel is getting more expensive.

Why multi-touch attribution is complicated

The "solution" to last-click is usually multi-touch attribution (MTA). Instead of giving all credit to the last touchpoint, you distribute it across all touchpoints in the customer journey.

In theory, this is better. In practice, it's messy:

• You can't track everything. Cross-device journeys, ad blockers, iOS privacy changes, and cookie restrictions mean you're missing significant chunks of the customer journey.
• The models are arbitrary. Linear attribution (equal credit to all touchpoints), time-decay, position-based—they all make assumptions about which touchpoints matter, and those assumptions are often wrong.
• It creates a false sense of precision. A dashboard showing "23.7% of this conversion came from paid social" implies an accuracy that doesn't exist.

MTA is better than last-click, but it's not the answer on its own.

A practical approach: triangulation

Instead of relying on any single attribution model, the best approach is to triangulate using multiple methods. No single source of truth—but several directional signals that, together, give you a clearer picture.

1. Platform-reported data (with a discount)

Each ad platform tells you what it thinks it drove. These numbers are inflated, but they're not useless. Track them consistently over time. If Meta says it drove 1,000 conversions this month vs. 800 last month, the absolute number may be wrong but the trend is probably real.

Apply a discount. In my experience, platform-reported conversions are typically 20-50% overstated for most B2C companies. The exact number varies—calibrate it by comparing platform claims to your actual total conversions.

2. Incrementality testing

The gold standard. Turn a channel off (or reduce spend) in a controlled way and measure the impact on total conversions. Geo holdout tests are the most practical: stop running ads in one region, keep everything else the same, and compare conversion rates.

This tells you the true incremental impact of a channel, not what it claims to drive. It's not always easy to run—you need enough volume and a clean test design—but even occasional incrementality tests will dramatically improve your understanding.

3. Media mix modelling (MMM)

Statistical modelling that uses historical data to estimate the impact of each channel on total outcomes. Unlike MTA, it doesn't need user-level tracking—it works with aggregate data, making it privacy-friendly and resistant to tracking limitations.

MMM used to require expensive consultants and months of work. Modern tools (Google's Meridian, Meta's Robyn) have made it more accessible, though you still need a data analyst who understands the methodology.

The limitation: it works best with significant historical data and meaningful variation in spend levels. If you've been running the same budget split for 12 months, there's not much signal to model.

4. Qualitative signals

Don't underestimate the simple "how did you hear about us?" question. It's imperfect—people forget, misattribute, and simplify—but it captures channels that digital attribution misses entirely: word of mouth, podcast mentions, social media browsing, PR coverage.

If 30% of your new signups say "a friend told me about you," that's a signal about the power of your product experience and word-of-mouth engine, even if no attribution model captures it.

Making decisions with imperfect data

The goal isn't perfect attribution—it's making better decisions with the information you have. Here's how I approach it:

• Use ranges, not point estimates. "Paid social drove between 800 and 1,200 conversions this month" is more honest and more useful than "paid social drove 1,043 conversions."
• Focus on big decisions, not small ones. Should you invest £500k in this channel or not? Attribution can inform that. Should you shift £5k from one campaign to another? Don't overthink it—just test and see.
• Track total efficiency alongside channel metrics. If your total CAC is improving as you scale a channel, that's a good sign—even if the channel-level attribution is fuzzy.
• Run incrementality tests on your biggest spends. If you're spending £50k/month or more on a single channel, invest in a proper incrementality test at least once a year. The insight is worth far more than the cost of the test.

What good looks like

The best B2C marketing teams I've worked with don't obsess over getting attribution "right." They accept that it's inherently imperfect and build a measurement framework that combines multiple signals:

• Platform-reported metrics for day-to-day optimisation and trend tracking
• An internal attribution model (even last-click) as a consistent baseline
• Periodic incrementality tests for their top 2-3 channels
• Quarterly or annual media mix modelling for strategic budget allocation
• Qualitative surveys and customer conversations to fill the gaps

No single method is right. Together, they tell a story that's close enough to act on with confidence. And that's all you need.