📊 MMM Interactive Dashboard — Where Every Dollar Works

Meta Robyn dt_simulated_weekly  ·  208 weeks  ·  Nov 2015–Nov 2019  ·  6 Media Channels

1 · Revenue Timeline
2 · ROAS & Paradox
3 · Decomposition
4 · Saturation
5 · Week Clusters
6 · Budget Simulator
★ Unexpected Insights
$1.82M
Avg Weekly Revenue
208 weeks
$3.83M
Peak Weekly Revenue
Q4 holiday week
+75%
Q4 vs Q1–Q3
★ Unexpected finding

Weekly Revenue Timeline

Yellow dots = holiday event weeks (+35% avg). Q4 averages +75% above Q1–Q3 — far beyond the event-flag effect alone.

★ Q4 Revenue Cliff (+75% above Q1–Q3)
Holiday flags explain +35%. The remaining 40-point gap is structural seasonality — a Q4-specific campaign cadence could unlock additional upside the model currently misses.
Interactive tip
Filter by Year to compare Q4 activation patterns across 2016–2019. Notice revenue spikes are consistent but scale with Facebook and Search activation — not just TV spend.
93.4x
Facebook ROAS
Only 2.3% of budget
44.7x
Search ROAS
Only 3.4% of budget
5.8x
Newsletter ROAS
Gets 24.3% of budget

ROAS by Channel (OLS Model Coefficients)

Revenue generated per $1 of adstocked spend. All 6 channels are statistically significant (p<0.05).

★ Budget vs Revenue Share — The Paradox

Dark = budget share. Bright = revenue share. They should match. They don't.

★ The Budget-Efficiency Inversion
Facebook gets 2.3% of budget but drives 21.2% of media revenue (93x). Newsletter gets 24.3% — the largest allocation — but drives only 13.8% of media revenue (5.8x). The budget is approximately the inverse of the efficiency ranking. This was invisible before decomposition analysis.

Avg Weekly Revenue Contribution ($K)

How much does each channel contribute per average week? Filter by year to see how the mix shifts.

Attribution Mix (% of media revenue)

Channel share of total media-driven revenue. Baseline organic demand = $864K/week (not shown).

Holiday Event Premium: $1,072K per event week (p<0.001)
Holiday events add over $1M per event week — the most statistically reliable signal in the entire dataset. Baseline organic demand accounts for $864K/week (47% of total explained revenue). Search drives the highest channel contribution at $264K/week.
14%
TV Saturation
★ Unsaturated yet inefficient
20%
Facebook Saturation
Room to grow + 93x ROAS
36%
Search Saturation
Most saturated, still 44.7x

Channel Saturation at Current Spend (Hill Function)

Green = unsaturated (room to invest more). Yellow = moderate. Red = approaching ceiling. Read alongside ROAS — not independently.

★ The TV Paradox — Low Saturation, Low Efficiency
TV is only 14% saturated — mathematically, there is room to spend more. Yet TV delivers only 14.9x ROAS vs Facebook's 93.4x. The Hill curve for TV has a shallow shape (alpha=1.67), meaning each marginal $1 of TV generates less revenue than $1 of Facebook at every spend level. Saturation is NOT the reason for TV's low returns.
Budget Action from Saturation + ROAS Combined
Invest more: Facebook (20% sat, 93x ROAS), Search (36% sat, 44.7x ROAS)

Reduce 20%: TV (14% sat, but only 14.9x ROAS) — not because saturated, but inefficient

Exit: OOH (10% sat, 1.6x ROAS) — lowest on both dimensions

PCA Cluster Map — Each Dot = 1 Week

K-Means k=3 on 7 features (6 channel spends + revenue). Select a cluster to highlight those specific weeks.

Avg Spend by Cluster ($K/week)

What separates High from Low-Performance weeks? Digital co-activation — not total spend volume.

★ Digital Co-Activation is the Differentiator
High-Performance weeks are defined by simultaneous Facebook + Search activation alongside TV. Low-Performance weeks have TV spend but lack digital activation. This pattern holds consistently across all 4 years. Never run TV-only weeks — always pair with Facebook and Search.
$1,947K
Current Fitted Revenue/wk
Model baseline
+0.0%
Scenario Revenue Lift
vs current allocation
+$0.0M
Annual Uplift (52 wks)
Modelled estimate

Budget Sliders — Zero New Spend

Drag sliders to shift budget allocation. Revenue KPIs update in real time based on OLS model coefficients.

Current vs Scenario Budget Share

Dark bars = current. Bright bars = your scenario.

Recommended: Scenario B — Set TV=-30%, Print=-20%, Facebook=+25%, Search=+25%
This is the highest-return reallocation validated by both ROAS ranking and saturation analysis. Zero new budget required. Estimated annual uplift: positive. Validate with a geo-holdout test before full deployment.
★ Finding 1: The TV Paradox
Pre-analysis assumption: TV underperforms because it is saturated.
What the data showed: TV is only 14% saturated — room exists to spend more. Yet TV earns only 14.9x ROAS vs Facebook’s 93.4x. The Hill curve for TV has a shallow shape (alpha=1.67), meaning each marginal $1 of TV generates less revenue than $1 of Facebook at every spend level — saturation is not the explanation.
TV gets 16.1% of budget but generates 23.3% of media revenue. Facebook gets 2.3% of budget and generates 21.2% of media revenue. The budget is the inverse of efficiency.
★ Finding 2: The Newsletter Overinvestment
Pre-analysis assumption: Newsletter is low-cost and acceptable to maintain.
What the data showed: Newsletter receives 24.3% of total budget — the LARGEST single-channel allocation. It contributes only 13.8% of media revenue with ROAS of 5.8x (second-lowest). The 10.5-point gap between budget weight and revenue weight is the largest misalignment in the portfolio — completely invisible before decomposition analysis.
★ Finding 3: The Q4 Revenue Cliff
Pre-analysis assumption: Holiday event flags capture the Q4 revenue premium.
What the data showed: Event flags explain +35% above normal weeks (p<0.001). But Q4 as a whole averages +75% above Q1–Q3 — a 40-point gap the flag doesn’t capture. This structural seasonal effect reflects consumer behaviour and media interaction effects not visible in a binary indicator. Activating Facebook and Search 4 weeks before Q4 exploits adstock carryover (FB θ=0.35, Search θ=0.12) to unlock additional revenue the model currently misses.