ATTEX
A head of insight at her desk by a window, typing a question into a laptop, a notebook beside it
ProductsAsk it from your assistant

The questions you ask every week, answered from the panel.

Five questions, asked in plain words, answered from the panel in the first week of November 2026. Every number below came back from the ATTEX server as it is shown here.

Works with
  • Claude
  • ChatGPT
  • Copilot

Early access opens to the waiting list first. Leave a work address and Emelie will be in touch before the server opens.

§ 01Five questions, five answers

In your words, not in a query language.

ATTEX is also an MCP server: connect it once to Claude, ChatGPT or Copilot and ask what you would have asked an analyst. Below, five questions put to it over mobile impressions in Sweden, Norway, Denmark and Finland, 1 to 8 November 2026, the week before the last two days that were still filling in.

Each rate carries the server’s margin of error. Rows returned at ±5 points or less, which the server calls a strong or a good basis, are drawn; rows it returned as indications, ±10 or wider, are named under the chart and left out. Beside each rate: the margin and the eye-tracked impressions it is a share of.

  1. 01You ask

    Which apps held attention most often in the first week of November?

    Attention by app · attention rate± margin · eye-tracked impressions
    Instagram60%±3 · 13,871Chrome45%±5 · 30,490Facebook43%±5 · 55,323Snapchat31%±5 · 9,146

    Instagram ads were looked at on 60% of eye-tracked impressions, Chrome and Facebook on 45% and 43%, which the margins do not separate, and Snapchat on 31%.

    TikTok (48%, ±20), Google, Reddit, YouTube and Temu (±10) came back as indications in the first week and are not drawn.

    Source: ATTEX, measured mobile attention, SE, NO, DK, FI, 2026-11-01 to 2026-11-08run 6a347b29

  2. 02You ask

    On Instagram, are reels looked at more often than feed posts?

    Format on Instagram · attention rate± margin · eye-tracked impressions
    Reels72%±5 · 3,537Feed53%±3 · 7,230

    Yes. Reels were looked at on 72% of eye-tracked impressions, feed posts on 53%, a gap of 18 points against a margin of 5.

    Stories came back at 70% as an indication (±10) and are not drawn.

    Source: ATTEX, measured mobile attention, SE, NO, DK, FI, 2026-11-01 to 2026-11-08run c5277b15

  3. 03You ask

    Do men and women look at mobile ads differently?

    Men and women · attention rate± margin · eye-tracked impressions
    39%Men±5 · 95,8871.3 s when looked at39%Women±5 · 90,5791.3 s when looked at

    No. Men looked at 39% of eye-tracked ads and women at 39%; the half-point gap is inside the margin, and when they looked, both rested on the ad for 1.3 seconds.

    Source: ATTEX, measured mobile attention, SE, NO, DK, FI, 2026-11-01 to 2026-11-08run 2d71c9b7

  4. 04You ask

    When in the day do we get the impressions, and when the attention?

    Hour of the day · attention rate
    impressionsattention rate, ±5 or less45%30%0003060912151825,047212336%42%

    Swedish time, UTC+1. Bars are impressions, the gold dots the attention rate.

    The impressions come in the evening, 25,047 in the hour from 21; the attention rate moves less, 42% at 20 and 36% at noon, a gap just outside the margin.

    The attention rate is drawn for the hours where its margin is ±5; the night and the early morning came back as indications.

    Source: ATTEX, measured mobile attention, SE, NO, DK, FI, 2026-11-01 to 2026-11-08run 7910d399

  5. 05You ask

    Which industries got the most attention last week?

    Industries · attention rate± margin · eye-tracked impressions
    Events and performances46%±5 · 3,556Clothing and accessories45%±5 · 4,742Consumer packaged goods45%±5 · 11,259Health and medical services45%±5 · 3,017Food and beverage services44%±5 · 3,152Vehicles43%±5 · 5,223Education and careers42%±5 · 2,515Travel and tourism42%±5 · 4,392

    Events and performances lead at 46%, clothing and packaged goods at 45%; the eight sit within four points, so no neighbour is apart from the next.

    Retail, the largest industry by volume, came back at 37% as an indication (±10) and is not drawn.

    Source: ATTEX, measured mobile attention, SE, NO, DK, FI, 2026-11-01 to 2026-11-08run b369a431

§ 02Questions about the assistant

What it does, and what it will not.

The data is the same as the export, read in aggregates. The answers below are what makes a number fit to put in front of a client, and where the assistant stops.

01
Does it ever show me a person, or one impression?
No. Your assistant gets aggregate numbers, never an individual impression or panelist, and a row that would rest on fewer than five panelists is left out before it reaches you.
02
How sure is a number?
Every rate and average comes with its margin of error, computed with the person as the unit, so many impressions from a few people never look more certain than they are. ±5 or less is a good basis; ±10 or wider is an indication, and the server says so in the answer.
03
How many people are behind a row?
Every row states what share of the panel in its markets and period saw an ad in it, so a striking number is read with the people behind it.
04
Why is a total over a breakdown smaller than the plain total?
Because the rows under five panelists were left out of the breakdown. The plain total still counts them.
05
Can I put an answer in front of a client?
Yes. Every answer carries its source line and the id of the run that produced it, as the plates above print them, so a number can be traced to the question and the day it was asked.
06
Can I ask about yesterday?
Ask about last week or last month. Phones upload late and the brand reading runs after the day, so the last two days are still filling in.
07
What can I ask about?
Mobile ad impressions in Sweden, Norway, Denmark and Finland, by day, hour, app, website, brand, industry, creative, ad format, age, gender and region, up to 300 questions a day.
08
Can I get a heatmap?
No. The record says whether an ad was looked at and for how long, not where on the creative the eyes rested.
09
Can I get the rows themselves?
Not through the assistant. One row per ad on screen is the API's product; the assistant answers in aggregates over the same record.
10
Does it cover desktop?
No. The assistant answers over the mobile record, phones in apps and on the web. Desktop impressions come through the export.
§ 03How to connect

One address, pasted once.

ATTEX invites the people on your account. Each gets a personal connection address that works as a key: you paste it into your assistant once, there is no password, and ATTEX mails you a short notice each time it connects an assistant. You sign in with a code sent to your invited address when the connection goes through a shared organisation address, or when your personal address is more than a week old; the code works for a week, and the assistant stays signed in for 30 days.

The steps for Claude, ChatGPT and Microsoft Copilot Studio are in the Docs; your own address appears there when you open the page from your invitation mail: How to connect

Emelie Löfdahl

Questions about the assistant, or an account for your team, go to Emelie Löfdahl, Co-founder and CEO: emelie@attexdata.ai. She replies within one business day.

Talk to Emelie