What is a cold, warm and hot audience?

These are the stages of task awareness. Cold - the person has not yet realized that he has a problem, or is not looking for a solution. Warm - he has realized the problem and is comparing options. Hot - has chosen a solution method and selects a performer or seller. Temperature is not about familiarity with the brand, but about readiness to act.

What is LAL and how does it work?

Look-alike, similar audience: the system takes your segment and finds people with similar behavior. The quality of the result depends entirely on the quality of the source - the algorithm looks for people similar to those you provided, and does not know how to distinguish a good client from a random one.

From which segment should we build look-alike?

From the narrowest quality: customers with repeat purchases, with a high receipt, who have reached the desired stage. Look-alike from all customers blurs the sign and makes it look like no one. A thousand correct entries are better than twenty thousand random ones.

What is a custom audience?

A segment collected from your data: a loaded phone or email database, website visitors who interacted with advertising. Unlike interest targeting, here you know exactly who these people are.

Why do custom audiences stop working?

They are getting old. Contacts change, interests fade, some people have already bought. The database, loaded a year ago and never updated, is gradually turning into a way to show advertising to those who do not need it, and spoils the campaign statistics.

What is an ICP and how is it different from an audience profile?

ICP, ideal customer profile - a description of the company or person for whom the product suits best and who brings in the most money. The audience profile describes who buys in general; ICP answers the question of who to focus on. The first is broader, the second is more useful for budget decisions.

Is it possible to run one campaign for all temperatures?

Technically, yes, in practice this ends with the algorithm optimizing for the most conversion part - the hot part - and stops purchasing the top of the funnel. After a month, the hot one runs out, because there is no one else to replenish it.

All articles

PerformanceSeptember 5, 202610 min

Audiences in advertising: cold, warm, hot, LAL and custom

I analyze the temperature of the audience in a human way: how cold differs from warm, what is look-alike and what segment to build it from, when custom audiences work, and when they burn the budget. With a table of offer temperature correspondence.

The words “cold” and “warm” audiences are thrown around in every discussion of advertising and almost always mean different things to different people. Most often they are confused with the degree of familiarity with the brand - and this is where the whole structure breaks down.

Temperature is about the stage of the task, not about getting to know you. A person can know your brand for ten years and be icy until they have a problem.

1. Three stages and what to show them

Stage	Where is the man	What works	What doesn't work
Cold	Didn’t understand the problem or isn’t looking for a solution	Analysis of the problem and its costs, benefits, observation	Buy offer, price, comparison with competitors
Warm	I understand the problem and compare methods	Comparisons, cases, selection criteria	Explanation that the problem exists
hot	The method chosen, chooses the performer	Conditions, guarantees, withdrawal of the last objection	Long warm-up, useful “in general”

The right column explains most of the dips. A cold audience is shown the price and is surprised at the zero conversion. They lead the hot item into a seven-day warm-up and lose it on the third letter, because the person was ready to buy on the first day.

2. Why do stages live in different campaigns?

The temptation to combine is clear: fewer campaigns, easier to manage, more data for the algorithm. But the algorithm optimizes for what converts better, and the hot part always converts better.

After a month, the entire budget goes to the hot audience, and impressions at the top of the funnel stop. After another month, the hot supply runs out - no one replenishes it anymore. This is the classic story of “at first everything was great, then it suddenly went downhill,” and the reason is the combination of stages.

3. Look-alike: the quality of the source is everything

LAL takes your segment and looks for people with similar behavior. The key here is: the algorithm looks for people similar to those you gave. He doesn't know how to differentiate between your best customer and a random customer who returned an item.

Hence the rule that is most often violated: the source is the narrowest quality segment, not the entire base.

Source for LAL	What will happen
All site visitors	Similar to everyone, that is, to no one
All buyers	It’s already better, but the sign is blurred by one-time purchases
Purchased twice or more	Workable option for most
Top third by revenue	Narrowest and usually most accurate

Bottom limit: too small a source produces noise. A reasonable guideline is from a thousand high-quality records; the algorithm below has nothing to learn from.

4. Custom audiences and their shelf life

Custom audience is a segment from your own data: loaded contact database, site visitors, those who interacted with advertising. Its advantage over interest targeting is that you know exactly who these people are.

The disadvantage is that it deteriorates. Contacts change, interests fade, part of the audience has already bought and its shows are now meaningless. A database that was loaded a year ago and has never been updated is gradually becoming an expensive way to show ads to the wrong people.

The practice is simple: update once a quarter and subtract those who have already become clients. The second is rarely done, but in vain - this is pure budget savings.

5. ICP and portrait: not synonymous

The audience portrait answers the question “who buys from us.” ICP - ideal customer profile - answers another: “who should be focused on.”

The difference is practical. The portrait is usually broad and therefore does not help make decisions: half the market fits it. ICP is narrow, describes those who bring the most money and the least problems, and it is he who determines where the budget goes and from whom to build a look-alike.

Definition and boundaries of the term - in ICP card, and the segmentation of an existing database by behavior is in the analysis RFM segmentation.

6. How to check that the stage is determined correctly

A simple test: Look at the ad and ask what the person must already know for it to make sense.

If understanding requires knowing that he has a problem, this is not a cold audience. If you need to know how your method differs from the neighboring one, it is no longer warm, but hot. Mismatch here is worth more than any creativity.

7. Summary

Temperature is a task stage, not an introduction to the brand. Each stage lives in its own campaign with its own goal, otherwise the algorithm eats up the top of the funnel. Look-alike is built from the narrowest quality segment. Custom audiences are updated once a quarter and are cleared of those who have already purchased.