Pangolin
Login
Welcome back! Please enter your details.
The DTC Brand's MMM Glossary
This glossary is written for DTC brands evaluating, onboarding, or actively using a Marketing Mix Model. Every definition connects to a decision you will actually face — where to spend, what to measure, and how to tell whether your model is doing its job.
Some terms you will encounter in vendor materials are proprietary inventions rather than established industry concepts. We have included two of the most common - miROAS and causal MMM - with context on how they differ from the standard terms they claim to replace.
Measurement Foundations
The baseline vocabulary of marketing attribution, experiments, and statistical constraints.
How credit for a conversion is assigned to marketing touchpoints across a customer's journey.
DTC Context: Most DTC brands start here, but click-based attribution alone cannot account for offline, brand, or organic effects.
Pangolin uses this to: combine click attribution signals with high-level MMM outputs for a single, unified view.
The true additional impact of a marketing activity beyond what would have happened anyway without that specific spend.
DTC Context: The gold standard question — did this specific Facebook spend create new revenue, or did it just capture demand that already existed?
A digital measurement rule that assigns 100% of conversion credit to the very final touchpoint before a purchase occurs.
DTC Context: Severely over-rewards bottom-funnel channels (branded paid search, retargeting) and under-rewards critical awareness spend.
A digital tracking framework that attempts to distribute conversion credit across multiple digital touchpoints using rules or machine learning.
DTC Context: Better than last-click, but fundamentally limited to trackable digital pathways — it misses TV, podcasts, influencers, and word of mouth.
An experiment where you deliberately suppress marketing spend in a specific region or audience segment to measure incremental lift.
DTC Context: The most reliable, absolute way to validate whether your MMM predictions match physical reality.
Pangolin uses this to: calibrate model priors and validate channel-level spend recommendations.
The proportion of true, meaningful pattern in a dataset relative to random variations or missing data fields.
DTC Context: Channels with low signal-to-noise (very small budgets, highly infrequent conversions) are harder for any statistical model to measure precisely.
Marketing Mix Modelling (MMM)
The mathematical concepts behind top-down aggregate statistical attribution.
A statistical model that estimates the contribution of each marketing channel to a business outcome, using historical aggregate data rather than individual user-level tracking.
DTC Context: The only measurement approach that naturally works across all channels, including offline, catalog, and privacy-shielded environments.
Pangolin uses this to: build a comprehensive full-funnel view of channel performance without relying on third-party cookies or fragile platform pixels.
An analytical approach that expresses results as probability ranges and distributions rather than rigid, single-point estimates.
DTC Context: Tells you not just 'Meta drove exactly £120k' but 'there is a 90% probability Meta drove between £95k and £145k.'
A Marketing Mix Model that uses Bayesian inference — combining prior human/experimental knowledge with observed data streams to produce realistic posterior probability distributions.
DTC Context: Far more transparent than black-box alternatives; you can easily audit what the model assumed and how the new data updated those assumptions.
Pangolin uses this to: provide full posterior distributions for every channel, ensuring you see clear safety boundaries rather than fake certainty.
In Bayesian statistics, the initial mathematical belief or parameter constraint set for a channel before observing the current data.
DTC Context: Priors encode what you already know — e.g., that Meta probably drives a higher ROAS baseline than an unproven channel you started testing yesterday.
Pangolin uses this to: incorporate industry benchmarks and your own historical lift-test data so the model starts from a sensible, physically possible baseline.
The final updated probability distribution for a parameter after combining the prior assumption with new observed transaction data.
DTC Context: The posterior is the actual answer — it tells you what the data, combined with prior safety parameters, now suggests about each channel's true contribution.
The lingering, decayed effect of advertising exposure over time. It represents the reality that an ad seen today can influence a purchase days or weeks later.
DTC Context: Critical for understanding high-influence channels like TV, podcasts, and YouTube, where the target consumer rarely purchases immediately.
Pangolin uses this to: model the custom time-delayed and cumulative impact curves of each individual channel.
Response Curves & Diminishing Returns
Understanding scale limits and how budgets behave as you ramp spend.
The mathematical speed at which the memory or impact of an ad run fades from the target audience's mind.
DTC Context: A high decay rate means the channel's impact is very short-lived (e.g., Google Paid Search); a low decay rate means it lingers much longer (e.g., TV).
The curve representing the relationship between spend and return for a channel, illustrating how incremental returns diminish as spend increases.
DTC Context: Every channel eventually saturates — the saturation curve tells your team exactly where they sit on the curve relative to optimal volume.
The point on a channel's saturation curve where additional budget dollars yield progressively smaller incremental returns.
DTC Context: The primary reason to reallocate budgets — you may be spending way past the point of diminishing returns on your largest channel without realizing it.
Pangolin uses this to: automatically identify the optimal spend level for each channel before return lines completely flatten.
When marketing spend on one specific channel indirectly lifts the performance of another, untracked channel.
DTC Context: A classic example: brand-building spend on TV or YouTube consistently lifts branded organic search clicks and direct website entries.
The mathematical equation that describes how spend in a given channel translates into incremental sales, incorporating adstock, saturation, and diminishing returns.
DTC Context: This is the central mathematical engine of what an MMM estimates — the physical shape of your channel curves.
DTC Commercial Terms
Aligning media efficiency metrics directly with P&L profit realities.
A channel can show strong ROAS on paper while destroying actual cash margins if the specific products it sells have unusually high return rates or razor-thin contribution margins.
Revenue minus all variable costs directly tied to producing, shipping, and processing the product.
DTC Context: The single most important unit-economics metric for any high-growth DTC brand.
Gross sales revenue minus the cost of goods sold (COGS) only.
DTC Context: Tells you whether the product itself is economically viable before any marketing, shipping, or storage costs are factored.
CM1 minus product fulfillment, payment processing, and shipping costs.
DTC Context: The true 'landed' margin — what you have left over to fund marketing experiments and operational overheads.
CM2 minus direct marketing/customer acquisition costs.
DTC Context: The ultimate bottom line on whether your marketing-driven sales are generating real profit for the business.
Pangolin uses this to: optimise budget allocation against CM3 rather than top-line revenue, ensuring every recommendation is fundamentally profit-aware.
The total marketing cost required to acquire a single customer, calculated as total campaign spend divided by new customers acquired.
DTC Context: Only meaningful when calculated strictly against new customers, not repeat buyers — customer retention efforts should not dilute CAC.
The total net profit a customer generates over their entire lifetime relationship with your brand.
DTC Context: The strategic counterpart to CAC — a brand can afford a higher acquisition CAC if their LTV curves are mature and predictable.
Pangolin uses this to: factor long-term LTV patterns into budget allocation so high-LTV acquisition sources are not starved.
The percentage of sold products that are returned by customers for a refund.
DTC Context: A major hidden margin killer — a channel that drives massive top-line revenue but high return rates may be highly unprofitable.
Budget Optimisation & Execution
How models transition from passive observation to active budget changes.
The strategic distribution of marketing funds across channels to maximize a defined objective like CM3 or revenue.
DTC Context: The ultimate output of any valuable MMM — transitioning from passive measurement into a concrete spend blueprint.
Pangolin uses this to: generate weekly or monthly spend recommendations optimised for maximum CM3.
The additional incremental outcome generated by the very next pound spent in a channel.
DTC Context: The primary input for allocation — smart systems move spend from low-marginal-return channels into high-marginal-return channels.
Total revenue generated divided by unit ad spend.
DTC Context: The most common digital marketing metric, but highly dangerous in isolation because it completely ignores variable product costs and returns.
A proprietary metric introduced by certain MMM vendors trying to estimate incremental return at the margin.
DTC Context: Still a revenue-focused metric, not a profit-focused one — it can still easily recommend bad allocations if product margins differ.
The automated, hands-off implementation of model-recommended budget changes directly across ad platforms.
DTC Context: Closes the gap between analytic insight and active campaign execution.
Pangolin uses this to: push optimised budgets directly into Meta, Google, and TikTok APIs without manual spreadsheet adjustments.
The statistical range of values within which the true parameter value is highly likely to fall under a specified probability.
DTC Context: When a model says 'Meta drove £120k ± £25k,' that ±£25k is the range of confidence.
Pangolin uses this to: surface uncertainty thresholds explicitly so you can easily separate high-confidence opportunities from speculative ones.