Findings you can trace
Open a finding and follow it back to the verified quote, cell or data row that supports it. Anything that cannot be checked stays visible as unverified and does not enter the totals.
Product — by Polaris Closed Beta
Most operational drag is scattered across waiting, handoffs, rework and small workarounds. Friktion brings those traces together, shows what they cost, and gives you a practical order for fixing them. Then it returns to the same evidence to see what changed.
A Polaris product, created for the kind of operational work we do with clients.
01 — How it works
We begin with the traces a workflow already leaves behind. From there, Friktion helps a team understand the drag, choose a response, shape the work and return to the evidence after the change.
Reconstruct the workflow from records, documents and the people who know it.
Stage 01 · ClarityStart with the structure: remove a step, simplify it, or make ownership clear.
Stage 02 · ArchitectureCreate the workflow, specification, checklist or automation the team can put into practice.
Stage 03 · IntelligenceUse the same yardstick to understand what moved and what still needs attention.
Stage 04 · PartnershipA Friction Audit you can revisit, compare and act on. See the Polaris model
02 — What goes in, what comes out
03 — The profile
Friktion looks at a named flow of work from its trigger to its outcome—quote to order, for example, or new-hire onboarding. Nine dimensions show where the drag appears and how it behaves.
The dimensions stay visible; they are never flattened into one score.
Separates active work from the time cases spend waiting.
Tracks where work changes hands and where something gets lost along the way.
Finds cases that return to an earlier step, reopen or need correction.
Surfaces work or information that the team creates more than once.
Shows where people re-key or carry information between systems.
Looks at the delay created by approvals and whether they alter the outcome.
Identifies work that stops while someone searches or waits for missing input.
Measures cases that leave the usual path and the extra cost they carry.
Compares the paths cases take with each other and with the documented process.
When the evidence cannot support a dimension, the profile says “not measured” and explains the gap.
04 — Why it's different
Open a finding and follow it back to the verified quote, cell or data row that supports it. Anything that cannot be checked stays visible as unverified and does not enter the totals.
Versioned code calculates the published numbers from stored observations. Models help read the material; they do not calculate, adjust or round the result.
Measured, documented, stated and unverified evidence remain distinct, so a confident-looking total cannot hide a weak foundation.
Friktion looks for a step that can be removed, merged or clarified before it considers automation or AI.
The team records what it expects to move, how it will be measured and when the follow-up starts before the change goes live.
A reviewer resolves uncertain evidence and accepts the workflow maps, recommendations and blueprints that move forward.
05 — Evidence
Friktion keeps the written process, the operational record and people's account of the work separate. A model can propose what a source means, but code still has to find the supporting quote or cell. The gap between the process on paper and the process in practice often reveals the most.
Code derives the figure from timestamps, status changes, owners, outcomes or time logs.
An SOP, checklist or other source describes how the work is meant to run, and the quoted text has been verified.
A person describes the work in an interview, email or message. A headline uses it only when another independent source agrees.
The source could not be checked. It remains visible for review but stays out of the figures.
06 — Deciding what to change
A recommendation begins with a named friction event and the evidence behind it. Friktion works up from small structural changes to integration, automation and AI, stopping as soon as there is a credible answer.
07 — Did it work?
Before a change goes live, the team records what it expects to move and how the follow-up will be run. The result then carries the strongest claim the evidence can support. For most SME engagements, that will be one of the first two levels.
| Level | What we can say | What it requires |
|---|---|---|
| T0 · Descriptive | In this period, on these sources, the workflow waited this long. | A frozen protocol, with coverage and basis reported. |
| T1 · Associative | The measure moved. Other changes are not separated out. | Two or more cycles on an identical protocol. |
| T2 · Controlled | It moved more than comparable workflows we left alone. | Pre-registration, plus unchanged control workflows or a staggered rollout. |
| T3 · Causal | The intervention caused the change. | Randomised or staggered assignment, pre-registration and replication. |
The expected movement is written down before the change goes live. A cycle with no movement, or movement the wrong way, is reported in full. No part of the measurement fee depends on measured improvement.
08 — Honest by design
Findings remain attached to a named flow of work and the evidence behind it. Friktion does not collapse an organisation into a single score.
The first cycle establishes a repeatable baseline. Later cycles show whether the measured friction moved; they do not turn assumptions into a promise of savings.
A written process shows how work is meant to run. Friktion marks the limits of that evidence until records or independent accounts show what happens in practice.
The goal is to improve the work. Sometimes that calls for automation or AI; often the better answer is to remove a step, clarify a decision or fix the source data.
09 — With Kontext
For organisations using both products, a planned read-only connection can place Kontext's brand and AI-discovery signals beside the operational workflows that may help explain them.
The connection is optional and off by default. It offers hypotheses for investigation, while Friktion's own measurements remain unchanged. Planned
Friktion is currently in closed beta.
10 — FAQ
It is the avoidable waiting, rework, duplication and effort between the start of a workflow and its outcome. Friktion compares a workflow with its own demonstrated best and connects every finding to a possible intervention.
It will show the delay and effort visible in the evidence, along with scenarios for recoverable time. Those scenarios are assumptions, not promises, and every figure says whether it was measured, documented or stated.
Friktion can work with process documents, exports from systems such as ticketing, CRM, ERP or order tools, time logs, and interviews or messages that explain how the work really happens. It can begin with documents alone, but the profile will make that more limited basis clear.
AI helps Friktion read and organise the source material, but it does not produce the official numbers or make the final decision. The method starts with the workflow itself and considers AI only where it is a better answer than removing or simplifying a step.
Friktion is being built for small and medium-sized organisations that know a workflow is harder than it should be and want evidence before deciding what to change.
Next step
We will help you work out where the friction sits and whether Friktion is the right way to investigate it.