Glava B2B Analysis Research practice

Desk research and surveys on B2B markets.

We write the terms of reference, set the method, do the analysis and build the argument the findings support. That is the work, and it is ours.

We have published studies with our partners, centres with the statistical apparatus and platforms with reach into a professional audience. In Brazil and Kazakhstan we are now running the whole cycle of work ourselves, using automation technology and our own analytical staff.

Three things we build

Working formats

The design work is the same in all three. What changes is where the raw numbers come from.

Format 01

Desk research on official statistics

Published state statistics rebuilt into a series nobody has assembled yet, such as splitting revenue growth into price effect and volume effect and holding that method constant across sectors and years.

Ours
Terms of reference, the cut, analysis, argument
Numbers
State statistics, our own or a partner's apparatus
Format 02

Recurring surveys of professional audiences

A defined population surveyed in waves. The same instrument repeated shows movement rather than a single snapshot, which is what earns a second and a third publication.

Ours
Instrument, wave design, reading of results
Numbers
Our own field, or a platform with reach into the population
Format 03

First count in a category that has none

When a market has no agreed size, definition or vocabulary, the first credible measurement sets the terms. The company that publishes it becomes the source the rest of the market cites.

Ours
Initiative, definition of the category, structure
Numbers
Platform data, client analytics, our own collection

Research register

2024 to 2026

Selected published studies, what each one measured, and whose dataset the numbers came from, since on these the collection was not ours.

Joint study Tebiz Group, analytical centre

Russian industry grew on prices, not on output

Across 2021 to 2025 the study separated revenue growth into price effect and volume effect for 14 industrial sectors. In 10 of the 14, passing costs into prices produced more than half of all revenue growth. Physical output barely moved over the five years, and revenue fell in 2026.

The study also names what held volume back, including capacity limits, labour shortage, equipment and credit costs, and logistics disruption.

What we did We wrote the terms of reference, designed the cut across sectors and held the method constant over the five years so the series would hold up. We did the reading of what the price and volume split means for industry. The Rosstat series ran through the statistical apparatus of Tebiz Group.
Method
Desk research
Data
Rosstat series
Period
2021 to 2025
Scope
14 industrial sectors
Study by
Glava and Tebiz Group
Published findings
Joint study PurpleSchool, developer education platform

The Russian IT labour market, in three waves

A recurring survey of working IT specialists through 2026, tracking employment, income and behaviour as the market tightened. In the first wave, 26% were still employed but expected to be cut, 34% were actively looking, and 22% had been dismissed during 2025.

The second wave found that 79% either already run a project of their own alongside work or intend to, 61% of them for skills and 59% for a second income. By the third wave one in ten had seen pay fall over the year, 57% had rewritten or reinforced their CV, and 13% had put it through an AI.

What we did We wrote the questionnaire, designed the wave structure so the three rounds stayed comparable, and analysed the responses. PurpleSchool fielded it inside its own community of developers, which is how the population was reached.
Method
Panel survey, 3 waves
Sample
1 038 in wave 2
Field
16 to 23 Apr 2026, wave 2
Population
Russian IT specialists
Study by
Glava and PurpleSchool
Published findings
Client dataset Doma.ai, housing management software

What digital management companies spend, and what residents get

Residents of apartment buildings are a large population that is rarely surveyed on the same questions twice, so almost nothing published about the sector shows movement rather than a single year.

Repeating the same instrument year on year gave the sector its first comparable series. Adoption of AI by management companies went from 13% in 2024 to 19% in 2025, the share of residents wanting work and study space in their building rose from 15% to 25% in a year, and a third of residents still cannot read their own utility bill.

What we did We set the schedule of recurring rounds, wrote the terms of reference for each one, and chose the cuts that had not been measured before. The resident surveys and the operating data from management companies came from Doma.ai.
Method
Annual resident surveys, recurring analytics
Sample
2 500 to 3 000 residents per wave
Sector
Housing and utilities
Period
2024 to 2026
Study by
Glava, data from Doma.ai
Published findings
Client dataset Mymeet.ai, AI meeting assistant

Five studies on how AI changes the way people work

How artificial intelligence changes a working day is argued about constantly and measured rarely. Five studies and a set of forecasts looked at meetings, hiring and retail, using anonymised platform data alongside surveys.

One counterintuitive result held up across cuts. Employees over fifty get through a call a quarter faster than their younger colleagues, 36 minutes against 48, which runs against the usual assumption about who slows a meeting down.

What we did We proposed the subjects, wrote the brief for each of the five studies and did the analysis. Anonymised product and usage data came from the Mymeet.ai team.
Method
5 studies and forecasts
Data
Product data and surveys
Subject
AI in work and hiring
Period
2025
Study by
Glava, data from Mymeet.ai
Published findings

Studies of our own, in progress

Brazil, Kazakhstan, Russia

The practice is now running its own desk research programme, built on national statistics and regulator filings rather than on a partner's dataset. First countries in work are Brazil and Kazakhstan, alongside continuing work on Russia. One rule holds across the programme. We do not study any market in which we hold a commercial interest of our own.

In work 01

Brazil

Three B2B markets to begin with, each one where public filings exist but nobody has assembled them into a series. Regulated crypto assets, insurance brokerage, and IT services, with adjacent markets to follow.

Basis
National statistics and regulator filings
Status
Terms of reference written, collection under way
In work 02

Kazakhstan

A market where B2B categories are growing faster than the published record of them, which is exactly the gap a first count is for.

Basis
National statistics, own collection
Status
Scoping
In work 03

Russia

Continuing coverage, with the industrial revenue series extended and new sectors added on the same method so the numbers stay comparable year to year.

Basis
Rosstat series
Status
Published and continuing

How a study comes together

Four stages

The order matters. The usual approach decides the headline and then looks for numbers to support it. We start from what can actually be measured, and take whatever the series says.

Stage 1

Find the gap in the record

What does the market argue about without data. We look for a question that has an answer somewhere in public statistics or in a reachable population, and that nobody has put together yet.

Stage 2

Write the terms of reference

The method is fixed in writing before any number is collected, so that later rounds stay comparable. Where the data is ours to gather we gather it, and where a partner already holds it we agree the same method with them.

Stage 3

Publish the method with the numbers

Findings are released with the sample, the field dates and the calculation open to inspection, so anyone who wants to check a figure can. Everyone who worked on the study is named.

Stage 4

Run it again

A single study is a data point. Repeating the same instrument on the same population is what turns it into a series, and a series is the only thing that shows direction.

How we work with editors

Disclosure

A study is only worth running if a desk can check it. These are the terms we work under, and they apply whether the study was commissioned by a client or started by us.

Who commissioned it

Every study names the company that commissioned it and whose data it runs on. Where a client funded the work, that is stated in the study itself and in the material sent to editors, not buried at the end.

Method on request

The questionnaire, the sample structure, the field dates and the calculation go to any editor who asks for them, before publication rather than after.

Raw data on request

Anonymised response level data is available to a desk that wants to run its own cut, including cuts we did not publish.

Findings as they came out

We report figures that do not suit whoever commissioned the study. A client can decide not to publish a study at all, and cannot ask us to change a number inside one.

Nothing bought

We do not pay for placement and do not offer payment for coverage. A study that is not interesting on its own merits does not run, and that is the correct outcome.

No market of our own

We do not study any market in which we hold a commercial interest, so nothing in the programme doubles as an argument for something we sell.

Start a study with us

Get in touch
Regina Rafikova
Research lead

Tell us the question your market argues about without data. If there is a defensible way to measure it, we will write the terms of reference and say what the study would take. If there is not, we will say that too.