Lift Analysis vs. Multi-Touch Attribution

Multi-touch attribution vs. lift analysis.

What do they do?

How are they different?

And in what situations and for what goals should you use one or the other?

Let’s explore.

What is lift and attribution?

Multi-touch attribution and lift analysis are two different approaches for measuring marketing. Together, they help cover each other’s blind spots.

Multi-touch attribution (MTA)?

Multi-touch attribution (abbreviated MTA) is a measurement method that tracks digital touchpoints across owned media and ad platforms and estimates their influence on revenue.

For example, in HockeyStack, we use MTA to track Linkedin ad impressions, website activity, clicks from ad platforms, engagement with live chat, outbound sales sequences, gifts received, and much more. Then we organize those touchpoints from first touch to closed/won in a chronological timeline called a “journey.”

From there you can use different attribution models to estimate the value of each touchpoint, analyze customer journeys independently, feed sales with buyer intelligence, and a whole lot more we’ll get to in just a second.

Lift analysis?

Lift analysis, on the other hand, estimates the incremental lift (results that would not have happened without a tactic) a campaign or marketing activity had on a variable like demos, sales, or subscriptions by comparing people who experienced the activity (treatment) vs. those who didn’t (control).

For example, want to know how much more likely subscribers are to book a demo than non-subscribers? Lift analysis compares people who booked a demo but didn’t subscribe (control) against people who booked a demo and did subscribe (treatment). The increase or decrease above or below the control group represents the incremental lift- either positive or negative.

Both multi-touch attribution and lift analysis have their benefits and limitations. But let’s start with attribution.

Benefits and limitations of attribution

Multi-touch attribution is outstanding at a few surgical tasks, namely owned media, paid media, and segmentation.

Zooming in on owned media

First, MTA is great for zooming in on owned media like content and creative- in real-time.

In fact, MTA zooms in on owned media better than most sophisticated algorithms.

For example:

  • How many organic blog entrances convert into customers a year later? Attribution will tell you

  • Who in the buying committee engages with what content pre-to-post-demo? Attribution will tell you

  • What website pages do closed/won accounts engage with most? Attribution will tell you

Even sophisticated statistical algorithms like marketing mix modeling can’t zoom in on owned media, in real-time, that effectively.

Tying paid media to revenue

Second, MTA is great at tracking paid media to revenue.

Since paid media is a tightly controlled environment, MTA can replace biased attribution models from ad platforms with first-party tracking and help you discover which keywords and campaigns drive qualified pipeline.

The result?

Less bloated metrics from ad platforms, more accurate attribution, better decisions.

Segmentation

Third, MTA makes segmentation and cohort analysis super simple.

When you’re tracking touchpoints, you can use those touchpoints as filters to create segments for more granular cohort-level analysis.

For example, want to analyze performance for people who received Linkedin impressions vs. those who didn’t? Or want to compare buying behavior for different products, regions, or price points? Or want to compare newsletter subscribers to non-subscribers?

With attribution, you can create those segments in minutes whereas it would take you hours (and a data team) to do it otherwise.

Limitations of MTA

What about the limitations of MTA?

Can’t track everything

First, MTA can’t track everything, which means it has a tendency to bias results favorable toward easy to track, bottom of funnel activity- which we know isn’t always true.

Can’t measure incrementality

Second, since MTA is only tracking a limited number of touchpoints, and since it’s not running any type of regression analysis over the data, it can’t measure incrementality or revenue contribution effectively for most of your marketing mix aside from paid ads.

In that sense, MTA is purely directional, not incremental.

Limited by privacy regulations

And last, MTA is subject to privacy laws and regulation since it tracks user-level data- which means a shrinking pool of collectable data as privacy restrictions tighten.

Benefits and limitations of lift analysis

Like MTA, lift has a few key benefits.

Incrementality

First, incrementality.

Incrementality refers to the additional impact a specific marketing activity has (or doesn’t have) on a goal beyond what would have happened naturally.

Lift analysis performs conversion lift, which is a specific form of incrementality measurement that focuses on the increase in conversion rates directly attributable to a particular marketing effort.

Unlike MTA, which can only track touchpoints, lift can estimate how much incremental lift those touchpoints or activities had on results.

Big difference.

Zooming in

Second, lift is also built for zooming in on campaigns, content, and creative, not zooming out on channels and aggregate data (that’s what marketing mix modeling is for).

Caveat: Lift won’t have the data to zoom in on a single blog post in most cases, but it can zoom in on categories of blog posts and tell you which types of blog had the most lift on conversions.

Limitations of lift analysis

What about the limitations of lift analysis?

Imperfect control

First, lift analysis operates in a world of imperfect control.

In other words, the control groups that lift uses to estimate incrementality are never perfect like they would be in a science lab.

For example, not everyone in the control vs. treatment group will be made up of exactly the same types of buyers.

Or not everyone will be perfectly randomized.

The result?

Biased or skewed answers on occasion.

Data requirements

Second, you may not have a sufficiently large enough control group to get statistically relevant results to begin with.

For example, you can’t use lift analysis to estimate how much more likely someone is to book a demo after viewing a blog post if that blog post only has 12 visitors. There’s just not enough data to work with.

Can’t go everywhere

Third, for the reasons mentioned above, you can’t use lift analysis to measure incrementality for everything, unless you want misleading results.

Lift works best for activities or touchpoints where data is abundant and where control groups and treatment groups are made up of the same people.

But it doesn’t work well where data is sparse, control groups vs. treatment groups are made up of wildly different people.

How do use lift and MTA together

Both MTA and lift have their strengths and weaknesses.

But how should we use each? And how should we use them together?

When to use MTA

In general, multi-touch attribution (MTA) is directional and real-time. Use it to draw preliminary insights, surface trends, and get a sense of whether or not you’re headed in the right direction.

It’s a great starting point for exploration and decision-making. And you can use it to gut check performance instantly- it doesn’t need historical data, experimentation, or data engineering to work today.

When to use lift

The goal of incremental lift analysis, on the other hand, is to estimate a cause-and-effect relationship between a marketing activity (like subscribing to a newsletter) and an observed change in a particular metric (like an increase in booked demos).

Use it to determine if a campaign or type of content had incremental lift on outcomes, or if the observed changes are due to other factors.

How to use them together

Whereas MTA is for observation, preliminary insights, trend spotting, and segmentation, lift is for validating those preliminary insights with more statistical rigor.

Lift can help you cross-verify whether or not the hypothesis you made based on MTA data is actually true.

For example, does it look like most of your buyers engage with a certain type of content on their path to purchase? You can use lift to categorize those types of content and estimate engagement on demos booked.

Also, you can use MTA to create custom segments for analyzing lift in the first place.

For example, want to estimate how likely someone is to book a demo if they subscribe to your newsletter first?

Since you’re tracking subscribers as a touchpoint with your MTA, now you can run a lift analysis in seconds to find your answer. Without MTA, it would take hours to produce a lift estimate in this scenario.

Wrapping up

Want to learn more about attribution and lift?

Schedule a demo (or reach out to your CS manager at HockeyStack) and we’ll walk you through our process.

Last updated