The answer is already in-house. No one has read it yet.
Most organisations don't need more information.
They need someone to make sense of what they already have.
That is where we start. We work with the material you have already paid for, add relevant published sources when they help, and turn the whole lot into an answer you can act on.
1,200 open-ended responses read and classified in two working days
Nine source groups and years of data turned into one clear view
Five practical actions for a problem whose causes had remained unclear
All figures are from completed client assignments. See what this looks like in practice →
The data is there. The direction isn't.
Reports tell you what happened. They rarely tell you why. Open-ended answers stay unread because reading them would take weeks. And when several sources need to be combined, the job is often too big to begin.
So decisions fall back on urgency and experience — even when the answer is already sitting in an archive, paid for and waiting to be found.
Same method. Different worlds.
For businesses
What are customers really saying beyond the NPS score?
Customer feedback, employee surveys, sales notes, support logs and exit interviews.
Years of evidence can explain why customers leave and why good employees move on — if someone reads it as one story.
One example: 1,200 open-ended responses analysed in two working days →
For the public sector and associations
What's really driving the problem — and what can be done about it?
Project reports, workshop materials, citizen feedback, member surveys and studies. The material has been collected and paid for. But once the project ends, it often goes on a shelf instead of into the next decision.
One example:
finding the root causes behind gaps in services for neurodivergent young people →
Researcher frames the question.
Models do the reading.
Process keeps it on track.
Tiivistämö (Finnish for "refinery") turns scattered sources into traceable, decision-ready insight.
First, map what's already there
We map what already exists around the question — inside and outside your organisation. Registers, research, statistics, public discussion and industry reports. The missing piece is often already there. You just need to know where to look.
Senior researcher
Asks the right question, draws the boundaries and keeps the analysis focused on the decision that needs to be made.
Language models and agents
Read at scale. Different models take on different tasks, and parallel runs reveal where they agree — and where they don't.
Curated process
Keeps every claim tied to its source. If a conclusion cannot be found in the original material, it does not make the cut.
Tiivistämö, step by step
Researchers stay in charge. Models do the heavy reading. Agents challenge the output. A human is accountable for the result.
03
Read in parallel
Analyse each source separately. Use multiple models in parallel and check the results before synthesis.
02
Frame the question
Define the analytical structure, context and causal relationships.
01
Source material
Bring together the sources that matter and map the links, dependencies and gaps between them.
04
Synthesize and verify
Combine the findings, link them back to their sources and verify every conclusion against the original material.
05
Put the answer to work
Bring hard-to-use and previously hidden evidence into the decision.
Can't we just ask a language model?
Sometimes, yes. For simpler tasks, that may be all you need.
The difference is everything around the prompt: who frames the question, spots bias in the source material and checks that every conclusion leads back to the evidence.
Without that work, a confident answer can still be confidently wrong.
Good analysis also knows where to stop.
Every source answers some questions and leaves others open. Feedback reflects the people who chose to respond. Existing sources tell us what has happened, not how people would react to something that does not yet exist.
We make those limits visible in every assignment. Better decisions start with knowing both what the evidence says — and what it doesn't.
Three ways to get started
START
PROJECT
ONGOING
One question. One set of source material. One clear finding. A low-risk way to see what Tiivistämö can do with your own material.
A broader piece of work that brings several sources into one view and turns the findings into recommended actions.
Add a workshop when it helps move the work forward.
Run the same scope and process at regular intervals. The results become comparable, and change becomes visible.
The first conversation is always free. If your own material is not enough, we find relevant external sources. We do not send you off to collect more data without a good reason.
Three ways to get started
START
PROJECT
ONGOING
One question. One set of source material. One clear finding. A low-risk way to see what Tiivistämö can do with your own material.
A broader piece of work that brings several sources into one view and turns the findings into recommended actions.
Add a workshop when it helps move the work forward.
Run the same scope and process at regular intervals. The results become comparable, and change becomes visible.
The first conversation is always free. If your own material is not enough, we find relevant external sources. We do not send you off to collect more data without a good reason.
Three ways to get started
START
One question. One set of source material. One clear finding. A low-risk way to see what Tiivistämö can do with your own material.
PROJECT
A broader piece of work that brings several sources into one view and turns the findings into recommended actions. Add a workshop when it helps move the work forward.
ONGOING
Run the same scope and process at regular intervals. The results become comparable, and change becomes visible.
The first conversation is always free. If your own material is not enough, we find relevant external sources. We do not send you off to collect more data without a good reason.
Making sense of complex information since 1997.
Blue Communications Oy has worked with research, analysis and participatory development for nearly three decades. Our clients include Finnish ministries, wellbeing services counties, cities and listed companies in Finland and abroad — organisations where the evidence is extensive, sensitive and tied to real decisions.
Our team combines more than 30 years of quantitative and qualitative research experience with over a decade in machine learning and language models. We stay small on purpose: the people you meet are the people who do the work.
