A Media Effectiveness Case Study
Venus Beauty — a large beauty retailer — came to us with a real data-sharing opportunity. They gave us everything: three years of weekly media spend across 10 paid channels, PR placement logs and impressions, customer acquisition data broken down by new versus reactivated customers and by ecommerce versus in-store, revenue, and brand sentiment.
The core question: where are people actually paying attention to Venus Beauty, and is the current media investment allocated accordingly?
Story Point 2
PR measurement has always been the awkward cousin of paid media measurement. Paid search gives you clean attribution — click, session, conversion. PR doesn't work that way. Earned media is diffuse. It shapes perception and drives behavior over time, and that signal gets buried in the noise of everything else happening in the business.
Barcelona Principles — AMEC, 2010 (updated 2025)
Principle 5: AVEs are not the value of communication. The industry explicitly rejects Ad Value Equivalency as a valid measurement approach.
What the Principles call for is measuring outcomes, not just outputs — not "how much coverage did we get" but "what happened in the business as a result."
OLS regression measuring PR alongside paid media, using adstock and saturation transforms for carryover effects and diminishing returns.
Quasi-experimental comparison of business outcomes in high-PR weeks versus low-PR weeks across 55 weeks of overlap data.
Story Point 3
The first thing you notice is the variability. Some weeks, Venus Beauty has a massive earned media moment. Other weeks, it's quiet. That variability is actually what makes measurement possible.
"If PR activity were constant, we'd have no basis for comparison."
Story Point 4
Paid search dominates. Google PLA, Google Non-PLA, and Bing together account for the bulk of Venus Beauty's paid media spend. That's typical for a beauty retailer with strong transactional intent.
Story Point 5
A simple plot of PR impressions vs. sales would only show correlation. Marketing Mix Modeling goes further: it isolates each channel's marginal contribution while controlling for trend, seasonality, competitor activity, and the diminishing returns of ad spend. This lets us compute a defensible Return on Ad Spend (ROAS) for every channel — including PR — in the same framework. That's what makes apples-to-apples comparison possible.
R² of 0.15 sounds low — but ~96% of weekly sales are driven by baseline factors (brand equity, seasonality, macroeconomics). Marketing is operating on the margin. The model is explaining the piece it's supposed to explain.
Story Point 6
Marketing is a thin layer on top of a massive baseline — and that's exactly what you'd expect for a business this size. But that thin layer is where all the optimization lives.
"~96% of sales are baseline. Marketing is the thin layer on top — but it's the part Venus Beauty can actually move."
Story Point 7 — The Money Chart
This is the chart the whole analysis builds toward. Each channel plotted by its Return on Ad Spend — how much revenue the model attributes per dollar invested.
Highest of any channel in the portfolio. On 2% of total spend.
Negative ROAS on Meta and Bing Non-PLA are almost certainly multicollinearity artifacts, not evidence these channels are destroying value. PR's 59.6× deserves the same skepticism — that's why we did the lift analysis.
Story Point 8
Bubble size = total spend. X-axis = spend share (%). Y-axis = ROAS. PR sits in the upper-left: tiny spend, massive return.
"PR delivers the highest return on the smallest investment — the definition of an under-allocated channel."
Story Point 9
Saturation curves show where each channel sits on its diminishing returns curve. Channels near the flat part are maxed out. Channels still on the steep part have room to grow.
PR and Pinterest sit on the steepest parts of their curves — meaning additional spend in these channels would generate proportionally more return than channels like Google Non-PLA or Meta that have already hit diminishing returns. These are the primary targets for incremental budget allocation.
Story Point 10 — Lift Analysis
For the lift analysis, we split the 55-week overlap period into quartiles by PR impressions, then compared the top 25% (high-PR) against the bottom 25% (low-PR).
Story Point 11 — The Headline Finding
"PR drives digital customer acquisition, not foot traffic. Earned media prompts someone to search — they go online, they find Venus Beauty, they buy. The path runs through digital."
Story Point 12
The statistical significance is concentrated in the ecommerce acquisition metrics. When significance clusters exactly where your hypothesis predicts it should — digital acquisition, not store — that's evidence.
Story Point 13
Not all PR weeks are created equal. We analyzed 1,170 placements across 10 campaign families to find which ones correlated with the strongest business outcomes.
p < 0.001 during its 6 active weeks — the strongest campaign-level signal in the dataset
of organic placements
Highest of any coverage category
with 74% affiliate link inclusion
Top 5 impression weeks
| Week | Impressions | Campaigns | New Ecomm | Reactivated |
|---|
Hair Care (464 placements, 20.9B impressions) and Nails (237 placements, 18.4B impressions) drive the most total reach.
But Skincare and Hot Tools have the highest quality metrics — favorability, earned rate, and affiliate conversion.
Story Point 14
Two independent methods. Different assumptions, different mechanics, different data slices. Same conclusion: PR punches far above its weight at 2% of spend.
Story Point 15
PR is the most efficient channel per dollar in the portfolio. The data supports a serious conversation about whether the current 2% budget allocation reflects that efficiency.
The industry default of measuring PR against total sales is too blunt. The right KPI for PR is ecommerce customer acquisition — new and reactivated. That's where the signal lives.
A Bayesian MMM extension using PyMC-Marketing would allow explicit priors on coefficients, better uncertainty quantification, and credible intervals rather than point estimates.
Which campaigns, placements, and narrative angles drive the strongest ecommerce signal? Disciplined tracking will sharpen every future PR brief.
This was a project about measurement rigor. But what it points toward is a clearer, more confident investment story for earned media — built on evidence, not intuition.
Gabe Schneier · Praytell Strategy / University of Michigan MADS