Challenges in Marketing Mix Modeling

What’s up everyone, today we have the pleasure of sitting down with Constantine Yurevich, CEO and Co-Founder at SegmentStream.

Summary: Multi-touch attribution is a beautifully crafted illusion we all pretend to believe in while knowing deep down it's flawed. The work is mysterious, but is it important? The big ad platforms sell us sophisticated solutions they don't even trust for their own internal decisions. Is it time we accept marketing causation is a thing we can’t measure? Visitor behavior scoring is a really interesting alternative or extra ingredient to consider. Often thought of as a tool for lead management to help prioritize your SDR’s time, the team at SegmentStream started using the same scoring methodology, but with an attribution application. Enter synthetic conversions. Instead of just tracking conversions, track meaningful visits like  time spent, pages explored, comparisons made. This allows you to connect upper-funnel campaigns to real behavior patterns rather than just looking at who converted in a single session.

About Constantine/SegmentStream

• SegmentStream was founded in 2018 in London
• Feb 2022 raised a first funding round of 2.7M
• SegmentStream is now trusted by more than 100 leading customers across the globe including L’Oreal, KitchenAid, Synthesia, Carshop, InstaHeadShots, and many others

The Messy Truth About B2B vs B2C Attribution Models

Price tags and decision timeframes obliterate the B2B/B2C attribution divide faster than most marketers realize. Constantine shatters conventional wisdom by showing how his team leverages their own attribution tools to measure website engagement because enterprise software purchases rarely follow predictable patterns. "Trusting last click is impossible," he explains, "because it takes too much time before conversion happens."

You've likely noticed this pattern in your own marketing stack. A $2,000 direct-to-consumer exercise bike creates the same multi-touch, 60-day consideration journey as many supposedly "straightforward" B2B software purchases. Meanwhile, those $30/month SaaS tools targeting small businesses convert with the immediacy of consumer products. Constantine points out how this pricing reality creates measurement challenges that transcend business categories:

High-ticket B2C products demand extended 30-60 day consideration windows
SMB-focused B2B subscriptions ($20-30/month) behave like impulse purchases
Enterprise B2B sales cycles stretch beyond a year with critical offline components

The offline measurement void plagues marketers everywhere. Constantine admits many of his most valuable marketing activities resist quantification. "I write a lot of LinkedIn posts, newsletters, we do podcasts. Some of these activities are very hard to measure unless you explicitly ask someone, 'How did you hear about us?'" Your gut tightens reading this because you've felt this same tension between attribution models and marketing reality.

Scale transforms your attribution approach more dramatically than business classification ever could. Small operations handling 100 monthly leads can simply ask each prospect about their discovery journey. Large enterprises processing thousands of conversions require sophisticated multi-touch models regardless of whether they sell to businesses or consumers. Constantine explains this convergence clearly: "When we talk about larger B2B businesses with thousands of leads and purchases, it becomes more similar to B2C with a long sales cycle plus an offline component."

The unmeasurable brand-building activities require a leap of faith that makes data-driven marketers squirm. Constantine embraces this uncertainty with refreshing honesty: "When you post on LinkedIn, build your personal brand, share content—that's really hard to measure and I don't even want to go there." His team focuses on delivering value through content, trusting that results will materialize. "You just share your content and eventually you see how it plays off." This pragmatic acceptance of attribution limitations feels like cool water in the desert of measurement obsession.

Key takeaway: Match your attribution model to purchase complexity rather than business category. Implement multi-touch attribution with lead scoring for high-consideration purchases across both B2B and B2C, while accepting that valuable brand-building work often exists beyond the reach of your measurement tools.

Why Marketing Attribution Still Matters Despite Its Flaws

Attribution chaos continues to haunt marketers drowning in competing methodologies and high-priced solutions. Constantine blasts through the measurement fog with brutal practicality when tackling the Multi-Touch Attribution (MTA) debate. While many have written MTA's obituary due to its diminishing visibility into customer journeys, his take might surprise you.

The attribution landscape brims with alternatives that look impressive in PowerPoint presentations but crumble under real business conditions:

Geo holdout testing sounds brilliant: Turn off ads in half your markets, keep them running in others, measure the difference. Simple! Except it'll cost you millions in lost revenue during testing. Constantine points out the brutal math: "For some businesses, this is like losing 1 million, $2 million during the test. Would you be willing to run a test that's gonna cost you $1 million?" These tests require a minimum 5% revenue contribution from the channel to even register effects, making them impractical for anything but your biggest channels.
MMM promises statistical rigor: But demands absurd amounts of data covering everything from your competitors' moves to presidential elections and global conflicts. Good luck collecting that comprehensive dataset spanning 2-3 years, then validating whether the TV attribution your fancy model spits out actually reflects reality.

> "Mathematically, everything works fine, but when you apply it in reality, there is no way to test it. You just see some numbers and there is no way to test it."

For scrappy D2C brands, SaaS startups, and lead gen businesses, Constantine argues MTA still delivers more practical value than its supposedly superior alternatives. You won't achieve perfect attribution, but you can compare campaigns at the same funnel stage against each other. Your lower-funnel campaigns can be measured against other lower-funnel efforts. Mid-funnel initiatives can compete with similar tactics.

Constantine drops a bombshell observation that should make you question the industry's MMM evangelism: "If Google and Facebook so willingly open-source different MMM technologies and they really believe in this technology, why wouldn't they implement it into their own product?" These data behemoths with unparalleled user visibility still rely on variations of touch-based attribution internally. Something doesn't add up.

Key takeaway: Stop chasing perfect attribution unicorns. MTA delivers practical campaign comparisons within funnel stages despite its flaws. For most businesses, sophisticated alternatives cost more than they're worth in lost revenue during testing or impossible data requirements. Compare apples to apples (lower-funnel to lower-funnel campaigns) with MTA, test different creatives, and focus on relative performance improvement. The big platforms themselves don't fully trust their publicly promoted alternatives - why should you bet your marketing budget on them?

Simplified MMM is a Measurement Fantasy You're Being Sold

Marketing Mix Modeling has roared back into fashion as third-party cookies crumble and marketers scramble for measurement alternatives. Constantine cuts through the hype with brutal clarity. Traditional MMM demands...