Timothy Davis on Performance Marketing, Paid Growth, and Attribution
Timothy Davis, Head of Performance Marketing at Shopify, in conversation with Lenny Rachitsky. The episode covers when to start paid growth, channel selection, attribution philosophy, team structure, and the data-centric hiring approach Timothy developed from a lifelong love of baseball statistics.
Key ideas
- Paid is for everyone — but timing matters. Start paid immediately if you need fast results; if product-market fit is not yet proven, paid spend mostly annoys users and creates lasting negative brand associations. Sort operations first: currency support, checkout flow, and basic conversion are prerequisites.
- Signs of life before scaling. Before scaling spend, run 1% lookalike audiences on your best customers as a signal test. If a small budget can generate conversions at acceptable cost, the funnel is ready for investment. If not, the problem is product or operations, not media.
- Channel sequencing: Google Search → Meta → YouTube. Google Search captures existing demand. Meta builds brand in-feed. YouTube works when creative is available and refreshed regularly. LinkedIn is powerful for B2B account-based targeting but is roughly three times more expensive than other channels — warrant only for high-LTV products.
- Attribution: multi-touch, time-decay. Users forget where they first saw your brand. Time-decay attribution gives credit to recent interactions without erasing early-funnel activity. Incrementality testing (GeoX, Conversion Lift) supplements attribution to detect whether paid spend is actually causing conversions or merely accompanying organic intent.
- Hire signal-finders, not platform operators. The hard part of performance marketing is not knowing how to place ads — it is identifying the signal in vast amounts of noisy data. Interview candidates by presenting a pile of metrics and asking what they would optimise; the right answer starts with, “What is the goal of the campaign?”
Channel Strategy
Google Search is almost always the first channel: it captures users who already know they have a problem and are searching for a solution. The exception is a genuinely new category where no search volume exists yet — in that case, Google Display Network builds awareness first, then Search captures the demand it creates.
Meta (Facebook + Instagram) is the second channel. Creative quality drives performance; without strong imagery or video, Meta underperforms. Video is the highest-priority creative investment: YouTube ads that open with emotion — comedy, warmth, genuine human feeling — outperform feature-first ads consistently.
LinkedIn targeting by job title, company, and seniority is uniquely powerful for enterprise B2B. Timothy’s example: geo-fencing Coca-Cola’s LA office and targeting their decision-makers by LinkedIn profile with ads addressing their two specific objections (security and recency) — the prospect told the sales team they were “seeing it everywhere” before the next call.
Attribution and Measurement
Multi-touch time-decay is Timothy’s default model: acknowledge the full customer journey but weight recent interactions more heavily than initial awareness. The limitation of all attribution models is that they cannot answer whether the person clicking the ad would have converted anyway without it.
Incrementality testing closes the gap: eBay’s famous 2012 study found that most branded search spend was driving conversions that would have happened organically anyway. Running GeoX tests — holding out a geography from paid spend and comparing conversion rates — gives the best read on true incrementality.
Quality Score optimisation in Google Ads is systematically under-exploited. Ad strength ratings (Poor/Good/Excellent) and the three components — expected click-through rate, landing page experience, and ad relevance — give a direct roadmap for improvement. A single “Poor” ad in a brand campaign can depress account-level quality score; pausing it before measuring other changes is a quick win.