Alex Sofronas

Chris Shupe is a Senior Marketing Analytics Leader who built his own plain English query (NLP) interface to enable the marketing to ask questions like, "What's the ROAS on x campaign in California?" He connected to Snowflake, using Virtual Studio and Claude Code, and the hardest part wasn't the build. It was making the AI trustworthy enough to actually use. He documented every table, every field, every quirk in the data (i.e. dollars stored in one place, cents in another ) and fed all of it into a context file the agent could reason from. Without it, the queries would be wrong just often enough to erode trust before adoption ever took hold. The result: marketing stakeholders at DoorDash could self-serve supply and spend data by market, in real time, without waiting on a backlogged analyst. The AI didn't replace the analyst. It ran on the analyst's institutional knowledge

12

1 Comment

*** ATTENTION BRAND & PERFORMANCE MARKETERS *** Your Incrementality Report has a Conflict of Interest The Situation: [🦊foxes are counting chickens🐔] > Somehow every platform and every campaign shows incremental lift. Every time. Even if you target the same audiences across platforms. The Struggle: [🎙️platform measurement bias⚖️] > Complete data and methodology transparency from platforms and normalization across platforms is unlikely. The Solution: [🌲persistent, deterministic, omnichannel identifier🪪] [1] a hold out you control [2] identity matching based on data you own [3] a cross-channel measurement baseline The above content is a summary derived from Postie's most recent weekly newsletter. See the links in 1st comment to subscribe or view the newsletter.

5

1 Comment

How much evidence is needed to prompt clients to take action against programmatic fraud? "A report released Wednesday by curation startup CleanTap claims to have proof that CTV and programmatic ad tech platforms served ads to spoofed devices pretending to be connected TV sets." https://lnkd.in/gRh5-233

2

Data teams are rebuilding marketing mix models as privacy-safe measurement becomes mission-critical. 📊 AMA Chicago's new blog takes a deep dive into how your team can leverage open-source tools and Gen AI to create a transparent MMM that won't lock you into a long term commitment with a legacy vendor. 🤖 Check it out here: https://bit.ly/4rQh8bY

4

You are donating to ad fraud. Willingly. I'll say the quiet part loud. AI is generating fake articles and MFA websites at industrial scale. These slip into programmatic supply chains undetected. Your DSP bids. A bot clicks. You report "reach." The fraud economy gets paid. $84 billion lost to ad fraud globally in 2023. That's 22% of every dollar spent on digital advertising that year. Gone. And we're still debating viewability benchmarks. MFA sites alone account for 21% of all programmatic ad impressions — $13 billion in fraudulent ad revenue annually according to the ANA. The same AI being celebrated in every marketing webinar is the exact tool being used to industrialise this at an unprecedented scale. India's programmatic supply chain makes this worse. Multiple SSPs, resellers, sub-resellers, zero accountability. The more hops in the chain, the more places fraud hides. If you've never asked for an IVT audit — you are funding this. If you've never pressure-tested your supply path — you are funding this. If you're buying reach without questioning its source — you are funding this. The fraud economy and the ad economy are running on the same rails. Fix the supply chain. Or keep funding it. There's no third option. When was the last time you audited where your ad rupees actually went?

8

5 Comments