Marketers evaluate results by tracking specific metrics at every stage of the customer journey, then using analytics tools and statistical methods to determine which efforts actually drove revenue. The process ranges from simple cost-per-customer calculations to sophisticated attribution models that assign credit across dozens of touchpoints. Here’s how each layer of that evaluation works in practice.
The Core Metrics That Define Success
Every data-driven evaluation starts with choosing the right numbers to watch. While dozens of marketing KPIs exist, a handful form the backbone of most evaluations.
Customer acquisition cost (CAC) measures what you spent to gain each new customer. You calculate it by adding up all marketing and sales expenses over a set period and dividing by the number of new customers you acquired during that time. If you spent $50,000 in a quarter and gained 500 customers, your CAC is $100. Marketers compare CAC across channels to see where their budget works hardest. A paid search campaign with a $40 CAC is outperforming a social media campaign with a $120 CAC, assuming both bring in customers of similar value.
Customer lifetime value (LTV) estimates the total revenue a customer will generate over the entire relationship. Comparing LTV to CAC reveals whether your acquisition spending makes financial sense. If it costs $100 to acquire a customer who will spend $800 over three years, that’s a healthy ratio. If LTV barely exceeds CAC, your margins are too thin to sustain growth.
Return on ad spend (ROAS) zeroes in on advertising specifically. It divides the revenue generated by an ad campaign by its cost. A ROAS of 4:1 means every dollar spent on ads brought in four dollars of revenue. Marketers use this to compare individual campaigns, ad sets, or even single creatives against each other.
Other metrics fill in the picture depending on the goal. Conversion rate tells you what percentage of visitors take a desired action. Click-through rate measures how compelling an ad or email is. Retention rates (tracked at day 1, day 7, and day 30 after signup) reveal whether customers stick around or disappear. Each metric answers a different question, and marketers choose their KPIs based on whether the goal is awareness, engagement, conversion, or long-term loyalty.
Attribution: Figuring Out What Actually Worked
Most customers interact with multiple marketing messages before buying. Someone might see a social ad, read a blog post a week later, click an email, and finally convert through a search ad. Attribution models are the frameworks marketers use to decide which of those touchpoints deserves credit for the sale.
Single-source models keep things simple by giving all the credit to one moment. A first-touch model assigns 100% of the credit to the initial interaction, the one that introduced the customer to the brand. A last-touch model does the opposite, crediting only the final interaction before purchase. These models are easy to implement but ignore everything that happened in between.
Multi-source models spread credit across several touchpoints, which better reflects reality. A linear model divides credit equally among every interaction. If a customer had five touchpoints, each gets 20%. A position-based model gives 40% credit to the first touch, 40% to the last touch before conversion, and splits the remaining 20% among everything in the middle. A W-shaped model assigns 30% each to the first touch, the moment a lead was created, and the moment an opportunity was created, with the final 10% distributed elsewhere. A time-decay model gives the most credit to interactions closest to the purchase, on the logic that recent touchpoints had the strongest influence.
The model a marketer chooses shapes their entire evaluation. A first-touch model makes top-of-funnel channels like display ads and social media look powerful. A last-touch model favors bottom-of-funnel channels like branded search. Multi-source models provide a more balanced view but require more data and more sophisticated tracking. Most mature marketing teams test multiple models to see how conclusions shift.
Analytics Platforms That Aggregate the Data
Raw data is scattered across ad platforms, email tools, social networks, and websites. Analytics platforms pull it together into a single picture. Google Analytics, for example, tracks how customers interact across websites and apps throughout their lifecycle. It provides cross-platform attribution, meaning it can follow a user from a mobile app visit to a desktop purchase and connect those dots. Its machine learning features surface patterns that manual analysis would miss, like predicting which audience segments are most likely to convert.
CRM platforms (customer relationship management tools like Salesforce or HubSpot) play a complementary role. While web analytics tracks anonymous visitor behavior, a CRM ties activity to individual customer records. It logs every email opened, every sales call made, every deal closed. Marketers use CRM data to calculate metrics like CAC and LTV at the individual level and to see which marketing campaigns generated leads that actually turned into paying customers, not just clicks.
The integration between these systems matters. When your analytics platform and CRM share data, you can trace a path from a visitor’s first anonymous website visit through lead capture, nurturing emails, and final purchase. Without integration, marketers are stuck evaluating channels in isolation, which leads to misleading conclusions about what’s working.
Cohort Analysis for Long-Term Evaluation
Not all evaluation happens immediately after a campaign ends. Cohort analysis groups customers by shared characteristics or timeframes, then tracks their behavior over weeks or months. This lets marketers evaluate whether a campaign brought in customers who actually stuck around, not just customers who converted once and vanished.
Acquisition cohorts group users by when or how they arrived. You might compare everyone who signed up during a January ad campaign against those from a March campaign. By tracking retention, engagement, and spending for each group over time, you can see which campaign attracted higher-quality customers. A campaign with a lower initial conversion rate might actually perform better if its cohort retains at twice the rate six months later.
Behavioral cohorts group users by what they do after arriving. If customers who complete an onboarding tutorial retain at 60% after 30 days while those who skip it retain at 15%, that’s a clear signal about where to invest. Marketers use these insights to reshape both product experiences and marketing messages.
Predictive cohorts use machine learning to group users by what they’re likely to do next, such as churn risk, probability of purchase, or potential lifetime value. Marketers use these forward-looking groups to trigger timely campaigns. A user flagged as likely to churn might receive a special offer. A user approaching their highest-value moment might see a premium upsell. This turns evaluation data into proactive strategy rather than a backward-looking report.
Testing to Prove Cause and Effect
Correlation is not causation, and marketers know it. Seeing that revenue went up during a campaign doesn’t prove the campaign caused the increase. A/B testing and incrementality testing are the tools that establish a direct link between marketing activity and results.
In A/B testing, marketers show two versions of an ad, email, or landing page to randomly split audiences and measure which version performs better. The randomization controls for outside variables, so the difference in results can be attributed to the change being tested. Marketers run A/B tests on headlines, images, calls to action, pricing displays, email subject lines, and virtually every element a customer sees.
Incrementality testing goes further. It answers the question: “What would have happened if this campaign never existed?” By showing ads to one group and withholding them from a matched control group, marketers can measure the true lift in revenue or conversions caused by the campaign. This is especially valuable for channels like display advertising, where it’s hard to tell whether someone who saw an ad and later purchased was influenced by the ad or would have bought anyway.
Evaluating Results in a Privacy-First World
The way marketers collect and use data is shifting. Browser restrictions on third-party cookies, app tracking transparency features, and privacy regulations have made it harder to follow individual users across websites and platforms. This has real consequences for evaluation, because traditional attribution models relied heavily on tracking users across the web.
First-party data has become the foundation. This means data collected directly from your own website, app, email list, or purchase records. Because customers share this information through direct interactions with your brand, it remains accessible regardless of browser or platform restrictions. Marketers are investing more in building these direct relationships through loyalty programs, email signups, and gated content.
Statistical modeling and predictive analytics are filling the gaps left by lost tracking. Instead of following individual users across sites, these approaches use aggregate patterns and machine learning to estimate which channels drive results. Media mix modeling, a technique that uses historical data on spending and outcomes to estimate each channel’s contribution, has seen a resurgence. Incrementality testing has also gained importance, since it doesn’t depend on tracking individual users at all. It simply compares outcomes between groups that did and didn’t see a campaign.
The shift away from user-level tracking hasn’t eliminated data-driven evaluation. It has changed the methods, pushing marketers toward approaches that rely on patterns, experiments, and owned data rather than following individuals across the internet.