I got asked last month to clarify attribution for a company running simultaneous campaigns in Moscow and San Francisco, and that conversation showed me how broken most attribution approaches are.
The standard question was simple enough: “Which touchpoint is actually driving conversions?” But the answer was impossible because our data was all over the place. Influencer traffic was tagged one way, paid ads another way, organic was a black hole. And across markets, the setup was completely different.
I started thinking about it differently. Instead of trying to build one Perfect Attribution Model (which doesn’t exist), I decided to build frameworks for different questions:
-
Awareness-to-decision path: Who gets the first impression? Usually ads or organic. What’s the last touchpoint before purchase? Often an influencer or email.
-
Market-specific patterns: In Russia, we found that influencer partnerships were often the last touch, meaning they got credit for conversions that started with ads weeks earlier. In the US, it was messier—more multi-touch attribution was necessary.
-
Content-driven signals: Which content formats actually moved people closer to purchase? A product tutorial performs differently than a lifestyle testimonial.
I started working with some frameworks that US-based marketers had shared—they were built for multi-channel environments and actually accounted for the fact that different markets move at different speeds.
The real win was admitting that one attribution model won’t work for everything. Now we use:
- First-touch for awareness campaigns (to evaluate top-of-funnel creators)
- Last-touch for conversion campaigns (to evaluate bottom-of-funnel partnerships)
- Multi-touch (using a time-decay model) for everything else
It’s not perfect, but it’s honest about what we know and don’t know.
How are you folks approaching attribution when campaigns span multiple markets and channels? What’s actually worked for you?
You’re approaching this right by acknowledging that one model doesn’t fit all situations. That’s actually the maturity level most brands aren’t at yet.
However, I’d push you toward something more sophisticated: build a probabilistic touch attribution model instead of relying purely on time-decay. What matters isn’t just the order of touches, but the conditional probability that a user converts when they encounter a specific touchpoint.
For example: if a user sees an influencer ad, what’s the probability they convert? 2%. If they see that ad AND visit your site organically the next day, what’s the probability they convert? Maybe 15%. That incremental lift is what you should be crediting to that influencer touchpoint.
For cross-market work, you’ll need separate models for each market because the conversion patterns are genuinely different. Russian users often research longer before purchasing, while US users tend to be more impulse-driven in certain categories.
One question: are you controlling for seasonality and market-specific events in your attribution? A campaign running during Russian New Year will have completely different conversion patterns than the same campaign in summer.
I love that you’re moving away from single-attribution thinking. In my analysis work, every time we tried to force one rule across the entire campaign, we missed important patterns.
What I’ve found is that you need to validate your attribution assumptions constantly. We built a model, then we looked at actual customer journey data (when we could get it), and the model was completely off in some segments.
For example, our model said influencers were driving 30% of conversions. But when we surveyed customers, only 5% had even heard of the influencer. What was happening? Influencer content was creating brand familiarity that made our ads more effective. But our attribution model wasn’t capturing that.
So now we use a slightly different approach: we measure influencer impact on ad performance, not just direct conversions. Influencers lift our paid media ROI. That changes how we evaluate influencer partnerships entirely.
Have you been able to measure that kind of halo effect, or are you mostly looking at direct conversions?
The way you’re breaking this down by question type (not just by model) is smart. Honestly, as someone who works on the partnership side, I see how attribution confusion kills collaborations.
Creators always want to know: “How many sales did I drive?” And brands can never give a straight answer because they don’t know themselves. Your approach means at least you can have an honest conversation.
I wonder if sharing this framework with creators upfront would help set better expectations. Like, “For this awareness campaign, we’re measuring first-touch attribution. For this conversion campaign, we’re looking at last-touch.”
That would actually make partnerships clearer, because everyone knows the rules of the ROI game.
Do you share your attribution methodology with creators, or is that kept internal?
This is exactly the problem we’ve been having. We run ads and partner with influencers on the same products, and we can never figure out which was actually effective.
What’s been happening in our data: sometimes an influencer posts about our product and sales spike, but the spike is mostly from people who already saw our ads. So is the influencer driving sales, or are they just reaching the same people the ads already warmed up?
Your multi-touch approach makes sense, but I’m wondering about the practical side: how do you actually implement this if your tracking isn’t perfect? We use a mix of UTM parameters, affiliate links, and manual tracking, and there are definitely gaps.
How do you handle the “unknown” portion of conversions that you can’t clearly attribute?
This is what separates predictable account management from guesswork. We’ve been pushing our clients toward more sophisticated attribution for years, but the friction is real—most platforms suck at cross-domain tracking.
What we’ve actually found works: build separate attribution models for paid traffic and organic/influencer traffic, then reconcile at the aggregate level. Paid traffic is trackable. Organic and influencer are not (at least not cleanly). So measure them differently.
For influencer work specifically, we’ve moved toward branded search lift as a proxy for influencer impact. When an influencer campaign runs, does branded search volume increase? That’s a leading indicator that the influencer is moving brand perception.
Cross-market becomes even trickier because Google Trends data quality varies by region, and search behavior is different. In Russia, Yandex is more relevant than Google.
Have you built region-specific attribution models, or are you trying to use the same framework across markets?
Okay I’m reading this as a creator and basically understanding that brands will never really know if I drove sales or not, which is kind of the reality, right?
But what I’m realizing from your post is that brands SHOULD be measuring influencer impact differently depending on the goal. If they want awareness, stop asking me to prove direct sales. If they want conversions, pick creators whose audiences are actively buying.
I think the honest version of this is: some of us are awareness creators, some of us are conversion creators, and brands need to know the difference before they brief us.
From my side, I’d way rather have a creator that’s transparent about what type of impact they drive than one that claims they can do everything.