Marketing Program Multipliers: Using Association Analysis to Find Winning Combinations

By Pete Furseth

As a modern marketer, your team is likely rich with data. You have a marketing automation platform containing all of your leads and the programs you ran against them. Hopefully, it also tracks outcomes in terms of MQLs, SQLs, and wins.

The problem many of us face is not having too little data. It is not knowing what to do with all the data we have. How do we use it to improve our marketing and drive more revenue?

This post identifies an advanced analytics technique that will help you determine what combinations of marketing programs lead to superior returns. The technique is called association analysis, and it reveals the marketing mix modeling insights that are hiding in your data.

Complements and Substitutes

Think back to your first economics class. Complements are things that go together: peanut butter and jelly, printers and ink cartridges. Substitutes are interchangeable alternatives: coffee and tea, butter and margarine.

Your marketing programs follow the same pattern.

Complement programs produce higher returns when run together than separately. A webinar preceded by a targeted email nurture might convert at 3x the rate of either program alone. The email builds awareness and interest, then the webinar converts that interest into a qualified lead. Substitute programs compete for the same audience or solve the same need. Running two webinars on the same topic in the same month splits your audience and reduces attendance for both. Running a paid search campaign and an organic content campaign targeting the same keyword may cannibalize each other.

You probably have a gut feeling about which of your programs are complements and which are substitutes. But how do you back up those feelings with data?

Association Analysis: The Technique

Association analysis is a data mining technique that large retailers use to uncover patterns in transaction data. When a supermarket discovers that customers who buy peanut butter and jelly are also likely to buy bread, that is association analysis at work.

The technique creates IF/THEN rules from your data:

> IF customers purchase peanut butter AND jelly, THEN they are likely to buy bread

You can apply the same technique to your marketing database.

Applying It to Marketing Data

Think of each lead in your database as a transaction at a grocery store. The marketing programs they participated in are the "items purchased" (the antecedent). The resulting outcome (MQL, SQL, or Win) is the consequence.

For example:

> IF "Product Demo Webinar" AND "ROI Calculator Email" programs are run on a lead, THEN the outcome is a Win

Your marketing database contains thousands of these implicit rules. Association analysis makes them explicit and quantifiable.

Measuring Rule Quality: The Big Three Metrics

Your database will contain many possible rules. Some represent genuine patterns. Others are statistical flukes. Three metrics help you separate signal from noise.

Support

Support is the frequency of a rule in your dataset. It tells you how often a specific combination of programs leads to a specific outcome.

Example: "Product Demo Webinar" and "ROI Calculator Email" resulted in a Win 25 times out of a database of 1,000 leads.

> Support = (25 / 1,000) x 100% = 2.5%

Higher support means the pattern occurs more frequently and is more likely to be meaningful.

Confidence

Confidence measures the reliability of a rule. It compares the number of times the program combination produced the desired outcome to the number of times it did not.

Example: The same two programs resulted in 25 Wins, but there were 5 leads with those same programs that did not win.

> Confidence = (25 / (25 + 5)) x 100% = 83.3%

High confidence indicates a strong, reliable pattern. An 83.3% confidence means that when these two programs are run together, they produce a win 83.3% of the time.

Lift Ratio

Lift ratio compares a rule's confidence against a baseline. The baseline is the overall win rate across all leads.

Example: There are 100 total wins in the database of 1,000 leads. The baseline win rate is 10%.

> Benchmark Confidence = (100 / 1,000) x 100% = 10% > > Lift Ratio = 83.3% / 10% = 8.33

A lift ratio of 8.33 means this program combination is 8.33 times more effective than the average. The higher the lift ratio, the more valuable the combination.

Interpreting Lift: - Lift = 1.0: The combination performs at the baseline rate. No advantage. - Lift > 1.0: The combination outperforms the baseline. This is a complement relationship worth exploiting. - Lift < 1.0: The combination underperforms the baseline. This suggests a substitute relationship. Running both programs together actually hurts performance.

From Rules to Strategy

Once you have identified your high-support, high-confidence, high-lift program combinations, you can use them to make strategic decisions:

Sequence Complement Programs

If your data shows that email nurture followed by a webinar produces a lift ratio of 5x, build that sequence into your marketing calendar. Do not run the webinar in isolation. Always precede it with the nurture sequence that amplifies its effectiveness.

Separate Substitute Programs

If two programs have a lift ratio below 1.0 when run together, schedule them in different months or quarters. They compete for the same audience, and running them simultaneously reduces the effectiveness of both.

Allocate Budget to High-Lift Combinations

When planning your marketing budget, allocate more to program combinations with high lift ratios. A combination with 8x lift deserves more investment than a program with 2x lift, all else being equal.

Identify New Combinations to Test

Association analysis may reveal unexpected patterns. Perhaps a whitepaper download followed by a LinkedIn ad retarget produces wins at a high rate, even though your team never intentionally designed that sequence. These discoveries are where the biggest ROI improvements often hide.

Practical Implementation

You do not need a team of data scientists to run association analysis. Here is a simplified approach:

1. Export your lead data including all program memberships and outcomes (MQL, SQL, Win) for the past 12-24 months.

2. Create a binary matrix where each row is a lead and each column is a program. Mark 1 if the lead participated, 0 if not. Add a column for the outcome.

3. Count combinations. For each pair and triple of programs, count how many leads participated in that combination and how many resulted in the desired outcome.

4. Calculate support, confidence, and lift for each combination.

5. Rank by lift ratio and focus on the top 10-20 combinations for your next planning cycle.

For larger datasets, tools like Python's mlxtend library or R's arules package can automate the process. For simpler datasets, even Excel pivot tables can surface the most common winning combinations.

The Bottom Line

Your marketing database is full of hidden patterns. Association analysis surfaces them so you can build program sequences that exploit complement relationships and avoid substitute conflicts.

The multiplier effect is real. Teams that identify and leverage program synergies consistently outperform teams that evaluate each program in isolation. Every program in your portfolio interacts with others. Understanding those interactions is the difference between a good marketing plan and an optimized one.

Start by looking at your top 10 program pairs by win rate. Calculate the lift ratio for each. The results will likely surprise you, and they will definitely improve your next marketing plan.