I Ran Both Attribution Models for a Quarter. Here's What Changed

By Ralph West  ·  August 19, 2026

Quick answer first, because you have a budget meeting this week. First-touch attribution tells you what got people to notice you. Last-touch tells you what closed them. If you run a business with a sales cycle longer than a few days, you need both, but you weight them differently depending on the decision you're making. Top-of-funnel spend decisions lean on first-touch. Budget allocation for the last mile leans on last-touch. Neither one alone will tell you the truth.

I learned this the hard way running growth on a DTC brand I scaled from $100K to $3M in annual revenue. We had one dashboard set to last-touch. It told us branded search and retargeting were our best channels. It was lying to us, sort of. So I ran both models in parallel for a full quarter, same data, same spend, two different lenses. What changed my mind is the story below.

What the problem actually looks like

Here's what you'll see if you're only running last-touch attribution:

• Retargeting and branded search look like your top performers, quarter after quarter
• Top-of-funnel channels like podcast ads, influencer content, or cold social show weak or negative ROAS
• You keep cutting the "underperforming" channels and shifting budget to retargeting
• Six months later, total revenue growth has flattened even though your last-touch ROAS numbers look great

If you're only running first-touch, you get the mirror image problem:

• Awareness channels get credit for every sale that eventually closes, inflating their apparent value
• Your bottom-funnel spend, the stuff actually closing the deal, looks unnecessary or redundant
• You underfund retargeting and email, then wonder why conversion rate drops

On the DTC brand, our last-touch model showed retargeting at a 6.2x ROAS and cold prospecting at 0.9x. Any reasonable operator looking at that dashboard cuts cold prospecting. We almost did. Then we ran first-touch on the same data and cold prospecting showed up as the first interaction for 41% of all customers who eventually purchased. Retargeting wasn't creating those customers. It was closing customers that cold prospecting had already found.

Why it happens

Attribution models are just rules for assigning credit across a sequence of touchpoints. Last-touch gives 100% of the credit to whatever happened right before the purchase. First-touch gives 100% of the credit to whatever happened first. Both are simplifications of a messier reality, which is that most purchases involve multiple touchpoints doing different jobs.

The mechanism that breaks last-touch: retargeting and branded search by definition sit at the end of the funnel. Someone already knows your brand, already has intent, and clicks the ad or searches your name right before buying. Last-touch attribution rewards proximity to the sale, not the actual work of creating the demand. It's structurally biased toward bottom-funnel channels, every time, in every business.

First-touch has the opposite bias. It rewards discovery, but discovery alone doesn't convert anyone. A podcast ad that introduces 10,000 people to your brand but converts none of them without a retargeting sequence and three emails still needed that retargeting sequence and those emails. First-touch will tell you the podcast ad "generated" all that revenue, which overstates its actual causal weight and understates the mechanics that closed the sale.

What we do about it

The procedure I use now, on every account I run:

• Pull both models on the same date range, same conversion events. Most ad platforms and tools like Triple Whale, Northbeam, or even GA4 let you toggle attribution windows and models. Run first-touch and last-touch reports side by side, not sequentially.
• Look at the delta, not either number in isolation. A channel that scores high on both models is a genuine performer. A channel that scores high on one and low on the other is playing a specific role in the funnel, not underperforming.
• Assign budget decisions to the right model. Use first-touch data when you're deciding what to test or scale for new customer acquisition. Use last-touch data when you're deciding how much retargeting or remarketing budget you actually need.
• Run a linear or position-based model as a sanity check once a quarter. It won't be your primary model, but it'll flag cases where both first and last-touch are misleading you about a mid-funnel channel like email nurture or a landing page test.
• Budget for the analysis time. This isn't free. Pulling and reconciling two attribution models properly takes a few hours a month minimum, more if your stack isn't unified. On the DTC brand this was about 6 hours a month of my time. Worth every minute.

The real constraint is data quality. If your tracking is broken, cookies are getting blocked, or your platforms don't talk to each other, both models will lie to you in different ways. Fix tracking before you fix attribution philosophy.

What it costs to ignore

On the infrastructure project I marketed, a $2.2B program with a $10M budget, we didn't have consumer purchases to attribute, but we had a version of this problem with lead sources for public engagement events. Relying on one model of "what worked" nearly got us to cut community outreach spend because it never showed as the "last touch" before someone showed up to a public hearing. Direct mail and paid search did. But outreach was what got people aware enough to respond to the mail in the first place. Cutting it would have starved the whole funnel a quarter later.

For the DTC brand, running last-touch alone for two more quarters would have meant cutting a channel responsible for 40% of new customer discovery. At $3M in annual revenue, losing that channel doesn't just cost you the direct revenue, it costs you the customers retargeting would have closed six weeks later. We estimated the mistake, had we made it, at roughly $400K in lost annual revenue once the funnel dried up.

The most common mistake

The most common mistake is picking one model and treating it as objective truth instead of a tool with a specific bias. Marketers want a single number. Attribution doesn't give you one, honestly. Anyone who tells you their attribution model is "the accurate one" is selling you something.

FAQ

Which model should I trust for reporting to leadership?

Report both, with the delta explained in plain language. Leadership doesn't need the raw model, they need to know which channels create demand and which ones close it. That framing lands better than either number alone.

Do I need multi-touch attribution software to do this properly?

Not to start. GA4 and most ad platforms let you switch attribution models manually. Multi-touch tools like Northbeam or Triple Whale make it faster and more precise, but you can run this comparison manually with spreadsheets for the first quarter to see if it's worth the investment.

Practical takeaway: don't ask which model is right. Ask what decision you're making, then pick the model built for that decision. First-touch for finding new demand. Last-touch for closing it. Run both, watch the gap between them, and you'll know more about your funnel than most of your competitors ever will.