How MMM works in Dema

​

How Dema approaches MMM

Dema’s Marketing Mix Modeling (MMM) is designed to help you optimize your marketing investments by focusing on profitability rather than just sales. By modeling Net Gross Profit 2 (GP2) (and extending insights to Net Gross Profit 3, or GP3), Dema empowers you to allocate your budgets more effectively and identify the point at which further spending yields diminishing (and potentially unprofitable) returns.

Key Idea: Traditional MMM tools often optimize for sales or ROAS. Dema’s model goes a step further by identifying where additional spend isn’t profitable, making it easier to maximize profit instead of just revenue.

---

​

Funnel-based modeling for accuracy

During onboarding, you’ll define funnels that reflect your marketing strategy. These funnels let you group campaigns by their purpose or goal. For example:

• Lower funnel: Meta campaigns containing “lower” in their name.
• Upper funnel: Google Display or awareness-focused campaigns.

Using funnels ensures the MMM model accurately reflects how you manage your marketing channels. Dema’s modeling draws on these funnel groupings to capture the unique contribution of each channel to profitability.

---

​

Profit-first modeling

Dema’s MMM centers on profitability curves:

• Diminishing returns: Each channel’s impact on profit decreases as you spend more, represented by curves where returns eventually plateau or become negative.
• Profitability threshold: Dema pinpoints the moment where an additional dollar of spend no longer generates positive GP2 (and GP3). By focusing on profit rather than sales, you can avoid overspending in pursuit of revenue alone.

This approach helps you discover exactly where your marketing budget delivers the best possible profit impact—not just top-line sales.

---

​

How the modeling works

1. Data ingestion

   • Dema taps into your historical spend, traffic, conversions, and GP2/GP3 data (including marketing costs), all of which is already cleaned and enriched within the platform.
   • Minimum one year of data is required to capture seasonality and establish a robust baseline.
   • Relevant control variables (e.g., holidays, discounts, inventory levels) are also integrated.

2. Time-varying baseline

   • Dema accounts for factors like organic traffic and brand strength to build a baseline model for both sales and profit.
   • Seasonality and brand growth are automatically included.

3. Funnel-based modeling

   • Your predefined funnels guide the model in grouping campaign activities.
   • Dema uses advanced statistical approaches (e.g., Bayesian methods under the hood) to create separate models per market, updated weekly.

4. Control variables

   • Events like Black Friday, product launches, or discount periods can be included.
   • For additional control variables not automatically tracked (e.g., offline activities), you can provide this data to Dema’s MMM specialists for integration.

---

​

Model validation & diagnostic checks

Dema uses hold-out validation and goodness-of-fit measures to ensure each MMM is reliable and accurate:

• Hold-Out Validation: Part of your historical data is withheld from the training process. The model’s performance is then measured against this “unseen” data.
• Goodness-of-Fit Metrics: Indicators like R-squared or MAPE show how closely the model’s predictions align with actual performance.

---

​

What makes Dema’s MMM different?

​

1. GP2 (and GP3) focus

While many tools optimize for sales or ROAS, Dema’s focus on net profit ensures every spend decision ties directly to your profitability objectives.

​

2. Weekly insights

Dema’s MMM runs every week, delivering fresh insights for each market and funnel group. This granularity makes it easier to course-correct and respond to shifting market conditions in near-real time.

​

3. Real-world constraints

• Spend change limits: By default, Dema caps weekly spend adjustments at ±30% to avoid disrupting ad algorithms (e.g., Meta, Google).
• Funnel/channel spend thresholds: You can configure spend limits or minimum budgets at a channel or funnel level, ensuring alignment with your strategic goals (e.g., invest a minimum of 10% in new channels).

---

​

Key outputs from MMM

​

Optimal spend per channel

Based on the model’s diminishing returns curves, you’ll see recommended spend levels that maximize profit (rather than just revenue). This includes respecting the ±30% adjustment rule by default.

​

Channel contribution

Dema quantifies each funnel’s contribution to performance, including synergy effects across channels. You’ll learn which channels truly drive incremental profit.

​

Profit potential

The model calculates potential profit gains—showing you the value of reallocating spend. By highlighting missed profit, Dema helps quantify what’s at stake if budgets aren’t optimized.

---

​

Implementation tips

Adjusting real-world budgets to match Dema’s recommended spend can be both an art and a science. Here are some best practices:

1. Gradual budget changes

   • By default, Dema suggests a ±30% maximum weekly adjustment, but if that still feels too aggressive, consider smaller increments (e.g., ±10–15%) and monitor performance metrics closely.

2. Syncing with ad platforms

   • Google Ads: Align Dema’s recommended ROAS target with your campaign settings (e.g., Target ROAS bidding). For daily budgets, break down the weekly recommendation accordingly.
   • Meta: If optimizing by CPA or conversion goals, adjust bids or budget caps gradually to avoid shocks to the algorithm.

3. Track early signals

   • Keep an eye on early KPIs (e.g., CPC, CTR) after making changes. If you spot unusual spikes or dips, consult Dema’s model to verify whether the shift matches your new budget strategy.

4. Ongoing monitoring

   • Dema’s weekly MMM updates make it easy to course-correct quickly. Regularly review performance and consider adjusting constraints if a channel consistently overshoots or undershoots profit targets.

---

​

Handling channels with very small spend

For channels that receive minimal spend and show little measurable impact:

1. Funnel consolidation

   • Group multiple small channels with similar objectives into a single funnel to increase data volume and improve model accuracy.

2. Minimum spend constraints

   • Force a small but consistent spend (e.g., 5–10% of the budget) so that the model can detect any incremental profit impact over time.

3. Incremental testing

   • Gradually boost spend to see if the channel responds positively. If not, consider reducing it further or removing it from optimization.