From Clicks to Conversions: Marketing Use Cases Powered by Model Builder

Read the first article in this series on Sales Use Cases, which can improve productivity of sales teams and help achieve targets faster .

Read this article to understand how to tackle below models:

• Most likely to buy / Propensity to buy prediction
• Email campaign target prediction
• Improve media spend
• Increase advertising ROI
• Increase conversion of leads

Note: A ready-reference of this blog is available as a pdf in our revamped Help article section. Here is the direct link.

The State of Marketing Industry

As the New year began, marketers began strategizing for the months ahead—planning their campaign calendar, allocating budgets, and implementing the lessons from the holiday season. Marketing success will rely on crafting data-driven, impactful campaigns that strike the right balance between relevance and engagement, while optimizing marketing spend to drive impactful results.

Marketers’ Top Priorities

According to Salesforce’s “State of Marketing, Ninth edition“ report, top priorities for marketers include:

• Maximizing ROI for the costly ad budgets — 20% of the entire marketing expenses is just Ad spend.
• Overcoming challenges in gaining actionable insights
• Focusing on the accumulation of first-party data to build stronger data foundations

Data cloud + Model Builder = Marketing insights

Data Cloud can assist in addressing all of these priorities to:

1. collect first-party data and building stronger foundations — the base for building a model.
2. get insights on data quickly with an enriched customer 360 profile — predictive models can provide with not just scores but actionable insights
3. invest on the right audiences and optimizing spends — all solvable with the right predictive model

The report also identifies AI as both the top priority and challenge for marketers. 54% of marketers use predictive AI to unlock next-level customer experiences. Marketers still need to rely heavily on predictive AI for getting their predictions, segmenting right audiences based on customer behavior. And Model Builder is the right capability to help with building models on top of these enriched data to drive better marketing metrics such as conversions.

How can predictive models power better campaigns and conversions?

Before diving into the use cases, let’s first get a sense on the marketing data that is required to build Predictive models for marketers to derive value.

What data can Marketers use to build Predictive models:

First-party data, transactional / non-transactional data are all the right data that we need. The below image highlights all the data collected from the Customer 360 foundations that they can use for building models . (Source: Salesforce’s “State of Marketing, Ninth edition“ report).

What problems can Predictive AI solve for marketers?

Predictive AI can help marketers precisely target the right audiences, help improve engagement, optimize media spend for better cost efficiency.

Here is a consolidated list that can help marketers turn insights into impact and campaigns into conversions:

1. Most likely to buy / Propensity to buy prediction
2. Email campaign target prediction
3. Improve media spend
4. Increase advertising ROI
5. Increase conversion of leads

Impact of B2B vs B2C on the use cases
Propensity to buy is focused on B2C conversions, so focus on adding individual demographics, geo data, and engagement. The lead conversion model is for B2B, so use account demographics, firmographics, geo data, and engagement. Prepare your data accordingly.

Now let’s dive into each of the use cases.

1. Most likely to buy / Propensity to buy prediction model

Use the propensity to buy model to predict which leads or customers are most likely to make a purchase. This allows marketing teams to focus on high-value prospects, delivering personalized offers and content that increase conversion rates and drive sales.

Training data preparation:

To get started with building a predictive propensity to buy prediction model using Data Cloud’s Model Builder, you’ll need some historical data:

• Gather lead demographics such as age, geographic details like city, customer segment details and so on.
• Compile interaction history with previous campaigns, including email opens, clicks, click-through rates (CTR), and engagement across channels.
• Include details on past purchase behavior, such as order frequency, recency of transactions, and average purchase value.
• Analyze website interaction data, such as session duration, pages visited, downloads, and time spent on key sections of the site.
• Add insights into service data on customer interactions, support cases.
• Integrate third-party intent data, capturing signals like competitor research, industry topics of interest, and related online activity.

Data structuring and target definition:

To prepare your training data:

• Start by compiling data that includes the available attributes along with the target variable (or goal) into a single DMO.
• Aggregate the data at the lead level using functions like SUM, COUNT, or AVERAGE as appropriate.
• Create the goal column Purchase and tag as True for customers with prior purchases (or exceeding a threshold) and False otherwise

Sample training data:

Using the propensity to buy model as an example, here we illustrate the training DMO that will be used for training a model in Model Builder.

Note: In this case, the target variable is Purchase (highlighted in purple along with a icon). The variable Purchase is True if the lead had prior purchases (exceeding a threshold, optionally) and False otherwise

Recommended Model Builder setup:

With this data in hand, you can set up a binary classification model aimed at maximizing propensity to buy. This model will help identify top prospects and prioritize their outreach.

Model type: Binary Classification
Goal: Maximize Purchase = True

A sample image of the Training metrics for a Propensity to buy model

How to use the model outputs in business:

With your propensity-to-buy model in place and predictions generated on live data, the next step is seamlessly integrating these insights into your marketing workflow. Here’s how you can take action:

• Prioritize and segment leads based on their likelihood to buy. You can use predictive outputs in a segment as a filter as shown below.

• Personalize offers and content tailored to high-propensity buyers, enhancing engagement and boosting conversions.
• Adjust campaign strategies by targeting audiences based on their predicted purchasing potential, ensuring your efforts drive maximum ROI.

Outcomes driven with this model:

By implementing a robust propensity to buy model with the Model builder, businesses can achieve below outcomes:

• Higher conversion rates by focusing on the top prospects
• More efficient budget allocation to focus their outreach programs and campaigns
• Increased revenue
• Improved customer retention

2. Email campaign target model

Use the email campaign targeting model to predict which leads or customers are most likely to engage with email campaigns. This can help marketers tailor efforts to the right audience, delivering personalized content that drives higher engagement rates and conversions.

Training data preparation:

For this model, you need the historical data on email campaigns mainly.

• Gather lead demographics such as their job title, experience, geographic details and so on.
• Add the Account’s firmographic data on Industry, No of employees, Revenue ranges, etc. and its geographic details as well (applicable for only B2B)
• Compile interaction history with previous campaigns, including email opens, clicks, click-through rates (CTR), and engagement across channels.
• Include details on past purchase behavior, such as order frequency, recency of transactions, and average purchase value to include the products a lead is interested in.
• Analyze website interaction data, such as session duration, pages visited, downloads, and time spent on key sections of the site to include the topics or products a lead is interested in.

Data structuring and target definition:

To prepare your training data:

• Compile data that includes the available attributes along with the target variable (or goal) into a single DMO.
• Aggregate the training data at the granularity of lead level using functions like COUNT, SUM, or AVERAGE as appropriate.
• Create the goal column EmailEngagement, tagging as True for leads with prior engagement (or exceeding a threshold) and False otherwise.

Recommended Model Builder setup:

With this data, you can set up a binary classification model aimed at maximizing email engagement. This model will help marketing teams to tailor their efforts to the right audience, delivering personalized and timely content that drives higher engagement rates and conversions.

Model type: Binary Classification
Goal: Maximize EmailEngagement = High

How to use the model outputs in business:

Here are some key actions you can take once your model is up and running:

• Target the most engaged leads for upcoming campaigns
• Personalize content based on individual behaviors and preferences
• Segment audience based on predicted likelihood to open, click, convert to improve engagement rates
• Refine targeting strategy based on insights from model predictions

Outcomes driven with this model:

By implementing this model with the Model builder, businesses can achieve below outcomes:

• Improved email conversion rates
• Improved customer retention and lifetime value
• More efficient budget allocation
• Increased brand loyalty

3. Improve media spend prediction model

Use the media spend model to predict the most cost-effective media channels and campaigns.

Training data preparation:

To predict media spend, you will mainly need historical spend data and a few more:

• Gather data on the historical media spend for each channel.
• Collect channel specific performance metrics (CTR, CPC, conversions)
• Gather media channel performance data like Reach, Frequency, Cost per click (CPC), Cost per thousand impressions (CPM), Return on ad spend (ROAS)
• Add customer engagement data for individual channels like Engagement rate, Impressions, Likes/Shares/Reposts, Views, Bounce rates and so on.
• Add any aggregated audience or segment-level demographic and behavior data from first part or third-party audience insights and data

Data structuring and target definition:

To prepare your training data:

• Compile data that includes the available attributes along with the target variable (or goal) into a single DMO.
• Aggregate the training data at the Channel + time period like month / quarter / year to evaluate channel-level spend effectiveness.
• Use functions like COUNT, SUM, or AVERAGE as appropriate.
• Create the goal column MediaSpend comprising of total media expenses for each channel and time period assumed earlier.

Recommended Model Builder Setup:

With this data in hand, you can set up a regression model aimed at minimizing ROI on Media spend. This approach will allow marketing teams to allocate budgets more efficiently, driving higher returns on ad spend and improving the overall effectiveness of media investments.

Model type: Regression
Goal: Minimize MediaSpend

How to use the model outputs in business:

Here are some key actions you can take once your model is up and running:

• Adjust campaign budgets to reflect media strategies
• Allocate optimal media budget across high-performing channels
• Forecast impact of varying spend on conversion rates
• Refine media spend allocation to optimize returns

Outcomes driven with this model:

By implementing this model with the Model builder, businesses can achieve below outcomes:

• More efficient allocation of marketing budget
• Maximized marketing impact
• Increased profitability
• Accelerated business growth

4. Increase advertising ROI model

Use the advertising ROI increase model to predict and enhance the return on investment for advertising campaigns.

Training data preparation:

To increase ROI from advertising, the main data required is on historical advertising budgets, spends and its details at a campaign level.

• Gather lead demographics such as their job title, experience, geographic details and so on.
• Add the Account’s firmographic data on Industry, No of employees, Revenue ranges, etc. and its geographic details as well (applicable for only B2B)
• Collect campaign-specific metadata like type, channel and so on.
• Add campaign-specific performance metrics (Revenue, Spend, engagement data like impressions, CTRs, CPA, RoAS and so on)
• Add any aggregated audience or segment-level demographic and behavior data from first part or third-party audience insights and data

Data structuring and target definition:

To prepare your training data:

• Compile data that includes the available attributes along with the target variable (or goal) into a single DMO.
• Aggregate the training data at the Channel + Campaign + time period level to assess ROI and effectiveness of ad campaigns.
• Use functions like COUNT, SUM, or AVERAGE to aggregate data
• Create the goal column AdvertisingROI, and derive using (Revenue generated – Media spend) ÷ Media spend.

Recommended Model Builder Setup:

With this data in hand, you can set up a regression model aimed at maximizing advertising ROI. This model will help focus on more targeted audience segmentation. and optimized spend allocation across channels.

Model type: Regression
Goal: Maximize AdvertisingROI

How to use the model outputs in business:

Here are some key actions you can take once your model is up and running:

• Adjust targeting to improve audience quality and relevance
• Optimize budget allocation toward more effective channels and campaigns
• Forecast impact of varying spends and targeting on ROI

Outcomes driven with this model:

By implementing this model with the Model builder, businesses can achieve below outcomes:

• Higher profitability from ad spends
• Maximized marketing impact
• Marketing aligned with customer interests
• Improved long-term growth

5. Increase conversion of leads

Use the lead scoring model to automatically score and rank leads based on their likelihood to convert.

Training data preparation:

To get started with building this model, you need data on the leads and their past performance.

• Gather lead demographics such as their job title, experience, role in the deal cycle (buyer, user, decision-maker), geographic details and so on.
• Add the account’s firmographic data on Industry, No of employees, Revenue ranges, etc. and its geographic details as well (applicable for B2B)
• Compile interaction history with previous campaigns, including email opens, clicks, click-through rates (CTR), and engagement across channels.
• Include details on past purchase behavior, such as order frequency, recency of transactions, and average purchase value.
• Analyze website interaction data, such as session duration, pages visited, downloads, and time spent on key sections of the site.
• Add insights into product usage patterns, such as recency and frequency of usage, key features engaged, and active users count (DAU or MAU).
• Integrate third-party intent data, capturing signals like competitor research, industry topics of interest, and related online activity.

Data structuring and target definition:

To prepare your training data:

• Compile data that includes the available attributes along with the target variable (or goal) into a single DMO.
• Aggregate the training data at the lead level using functions like SUM, COUNT, or AVERAGE as appropriate.
• Create the goal column LeadConvertedToCustomer, tagging as True for customers with prior purchases (or exceeding a threshold) and False otherwise

Recommended Model Builder Setup:

With this data in hand, you can set up a binary classification model aimed at improving conversion rates and expect to accelerate sales cycle by providing the best leads. Marketing teams can prioritize efforts on high-value leads, deliver more personalized and targeted campaigns, and efficiently allocate resources.

Model type: Binary Classification
Goal: Maximize LeadConvertedToCustomer = True

How to use the model outputs in business:

Here are some key actions you can take once your model is up and running:

• Prioritize most promising leads
• Route leads to sales teams
• Personalize outreach
• Optimize resource allocation

Outcomes driven with this model:

By implementing this model, businesses can achieve below outcomes:

• Improved lead conversion rate
• Optimized marketing resources
• Accelerated sales cycle
• Aligned Marketing & Sales

Conclusion

In this article, we explored practical Marketing use cases that can be built with Model Builder aimed at improving conversions, ad budget usage and getting better ROI. By exploring these use cases and building models tailored to your needs, you can unlock marketing value, by precisely targeting the right audiences, improving engagement, optimizing media spend to stay ahead in the market.

Let’s take your marketing to the next level—smarter strategies, stronger impact, and faster results!

Special thanks to Bobby Brill for his review.

For more resources on Model Builder:

• Blogs:
  • Build an AI Model with Clicks in Data Cloud
  • Build Forecasting models using Einstein Studio’s Model Builder
  • From Data to Decisions: Integrating Predictive AI for Key Business Workflows
  • First article in this series of Use cases – Unlocking Sales Success with Model Builder: A Practical list of Sales Use Cases
• Salesforce Help articles:
  • Create AI Models
  • Build Your Own Predictive AI Models
  • Get Predictions
  • Actionable variables
• YouTube: Model Builder: Ask Me Anything with Salesforce Developers
• Podcast: What are the key features of Salesforce’s Model Builder?

Written by