What is lift in marketing?
How is lift calculated?
Take two comparable groups. One sees the ads. The other does not. Measure the same outcome in both, such as purchases. The gap is the lift.
The same gap has two descriptions. Extra purchases is the absolute lift. The percentage over the control group is the relative lift. Ask for both. A big percentage on a tiny base can be a small number of orders.
Always against what?
A lift figure is only meaningful against a named control. Lift against a group that did not see the ads is one thing. Lift against last month is another. The second mixes in everything else that changed. The control side is your baseline.
This is why we avoid the word "uplift". It is also used in conversion rate work, where it usually means the gain of one page version over another.
Does "no lift" mean the ads do not work?
Not necessarily. A research chapter on online advertising experiments reports that measuring the effect of ads can need very large samples, and that results can come with wide intervals2. A test with too little data can show no lift when there is one. Check the sample size and the uncertainty before reading a flat result as "no effect".
Worked example
Easily confused with
Lift vs Incrementality
The concept: what the ads caused. Lift is how one test measured it.
Lift vs Brand lift
The change in a brand metric, such as ad recall or awareness, usually measured by survey. It is not the same as a lift in conversions.
Lift vs iROAS
Incremental revenue divided by ad spend. Lift gives you the incremental part.
Lift vs Conversion rate change
A shift in conversion rate over time. It has no control group, so it is not lift.
Platform experiments that report lift are covered in conversion lift study. The general method is a holdout test.
FAQ
Is lift the same as incrementality?
Not quite. Incrementality is the idea of extra results caused by the activity. Lift is the number a test gives you for it.
Is a higher lift always better?
Not on its own. Look at what the lift is worth after margin, and how certain it is.
Sources
Let’s find what’s actually working.
Next step. Check what your ROAS looks like on explicit assumptions. Use the Is your ROAS real? calculator.