ENRUSR
Back to cases
Still from the recording: two speakers in armchairs on a studio set lit purple and red
Yandex · 2023—2024

How demand was mapped in a category nobody talks about

The opening talk of a conference on adult products. Search data was not enough, and the survey team turned me down first.

942
usable answers out of 1,004 collected
3
data sources instead of search alone
3
weeks from request to clean data, on budget and on time
13,000+
views of the talk recording

A breakdown of the method, not the findings: the market figures belong to Yandex. Slides come from the public recording of the talk.

FormatOpening talk of an online conference
Date23 January 2024
AudienceShops, manufacturers and agencies in the category
My partThe analytics behind the talk. A speaker presented it, not me
01 · Task

Here search data is precise about products and tactless about people

An ordinary brief: what demand looks like, who buys, what an advertiser should do. The category was not ordinary.

Difficulty 1 · data
Accurate ≠ usable
The “who searches for what” cut is detailed and real. Read out from a stage, it turns the audience into test subjects.
Difficulty 2 · audience
Insiders
These people live in the category every day. They would spot a polished answer in the first minute.
02 · Sources

Three sources: each covers the blind spot of the other two

Search tells you what people look for. It does not tell you why they hesitate, where they buy, or what would make them buy more.

01 · Search demand
  • size and structure of the category
  • seasonality
  • motive and barrier
2.5 years of history, each quarter against the same quarter a year earlier. Geography in three layers: the two capitals, cities over a million, everything else.
02 · Ad restrictions
  • what the category may not say
  • what to stop trying
  • anything about buyers
A review of rejections and moderation. In a restricted category, moderation shapes the advertising before strategy does.
03 · Our own survey
  • motive and barrier
  • purchase channel
  • willingness to talk
  • market size in money
Our own questionnaire, fieldwork across the country, the capital regions broken out separately.
03 · Refusal

I answered “no” with a finished questionnaire — the next day

The survey team said no: the request was too big, the deadline too short. Instead of escalating, I came back with specifics.

4 DecemberRequest to the survey team
5 DecemberRefused, sent on to the strategists
6 DecemberBuilt the questionnaire myself: 16 questions
11 DecemberThe team joined in, questionnaire refined
25 December~900 interviews over a weekend
26 DecemberFieldwork closed
Why the second attempt worked
Broad request
  • “we need a survey on the category”
  • scope of work unclear
  • answer: no
→
Specific request
  • a finished questionnaire on my own account
  • cost, fieldwork dates, reach
  • answer: they joined in
04 · Survey

Embarrassment was not asked about head-on

Asked directly, nobody admits to being embarrassed. People name a general taboo readily, as long as it is not about them.

The barrier question is projective
“What holds you back?”
“What do you think might hold others back?”
This question produced the conclusion the whole talk was built around.
How the questionnaire was built
11 + demographics
purchase experiencefrequencychannelcategoriesspendbarriersanonymityreviewswhat is missing
Cleaning the sample
Collected in the field1004
Kept for analysis942
Removed: those who declined a survey on this topic, those who gave no clear answer on whether they had ever bought, and a handful of fraudulent responses.
Every chart had its own base
Frequency and spend

Asked only of those who had bought before.

First-purchase channel

Asked only of those who had not bought yet but planned to.

The audience could see what each number was made of.
05 · Slide

The precise slide came off the stage. The conclusion stayed

The deck had a search cut: which gender and age searches for which product. I proposed replacing it with survey data.

Before
A cut of search demand
  • gender × age × specific product
  • real data, nothing to fault
  • from a stage — indelicate
After
The same conclusion from the survey
  • people said themselves what they buy
  • in general terms, with no item-by-item split
  • two weeks later the PR review asked for exactly this
Stage: the survey versionHandouts: the detailed split, marked confidential

The detailed split was not deleted: precision for those who need it, restraint for the stage.

06 · Outcome

The talk opened the conference, and the recording is still being watched

Still from the recording: two speakers in armchairs on a studio set Watch the talk on YouTube
The video is 18+ — YouTube will ask you to confirm your age
Slide from the talk — 1
Slide from the talk — 2
Slide from the talk — 3
Slide from the talk — 4
Slide from the talk — 5
Slide from the talk — 6
Slide from the talk — 7
Slide from the talk — 8

Eight slides from the deck. I assembled the deck; the design team styled it.

In a sensitive category, accuracy and fitness for the stage are two different exams. Passing the first does not mean passing the second.

My part
Survey

The questionnaire after the refusal, fieldwork, cleaning the sample down to 942 answers.

Ad restrictions

What moderation rejects and why — the practical part of the talk.

Building the deck

Data from three sources turned into the speaker's slides; the indelicate cut replaced.

survey designprojective questionssensitive categoryad and moderation restrictionssearch demandwork under NDA
What I would do differently
Ask for the watch-through data

The second-by-second curve sits with the channel owner. I should have asked while anyone still cared.

Start fieldwork earlier

A spare week would have allowed a pilot of the barrier question wording, and everything else depended on it.

Write the method down right away

Rebuilding this page took a day of digging through two-year-old correspondence.

The conference recording is public and the event page is still live. The research behind the talk is not: the figures belong to Yandex, and this page only describes how the work was done. 13,000+ is the cumulative view count of the recording on YouTube.

Conference page →  ·  ← Back to cases