> ## Documentation Index > Fetch the complete documentation index at: https://docs.within.ai/llms.txt > Use this file to discover all available pages before exploring further. # How Advisor uses your process attributes > Which attributes Advisor does arithmetic with, how much precision your decision actually needs, and which fields carry the case you are making. export const NeedHelp = () => Need help? Use the in-app chat — click the chat bubble in the bottom-right corner of Within (staffed 24×5) — or email support@within.ai. ; Advisor reasons over your Process Index, so the attributes on your process nodes are the quantitative substrate under everything it produces. This page covers which fields it does arithmetic with, how much precision the decision in front of you actually needs, and which attributes carry each kind of value case. For what the attributes are, how their values get set, and how to correct one, see [Maintaining your process attributes](/user-docs/structure/process-attributes). **The rule this page is built around.** You don't need an accurate number — you need a number accurate enough for the decision in front of you, and you need to know which one you have. Choosing where to look first tolerates being 50% out. A funded business case doesn't. Most of the effort people spend chasing precision goes on questions that never needed it. ## What Advisor does with them Attributes feed Advisor in six distinct ways, and they are not equally load-bearing: * **Active Work Time × Annual Volume** is the effort arithmetic behind time-savings estimates and automation assessments. This is the pair to get right, and both halves matter equally — a careful Active Work Time against a guessed volume gives you a precise-looking wrong answer. * **Cycle Time − Active Work Time** is where wait time and rework surface. Advisor uses that gap to point at queues, approvals and handoffs — the waste that never appears in an effort total. See [What Cycle Time is actually for](/user-docs/structure/process-attributes#what-cycle-time-is-actually-for). * **Key Control** shapes what Advisor will recommend automating. A step that prevents or detects error is treated differently from a routine one. * **Systems Used** and **Teams Involved** drive scoping and hand-off analysis. * **Numbered stage folders** tell Advisor the order your stages run in — see [Naming conventions and taxonomy hygiene](/user-docs/structure/naming-conventions-and-taxonomy-hygiene). * **A pinned standard** turns attributes into a standard-versus-reality comparison — see [Set your process standard (UDS)](/user-docs/advanced/using-a-user-defined-standard). **In short:** Advisor answers whether or not the attribute is populated — it reasons around gaps rather than refusing. So ask it to show its work. Adding *"Include any assumptions or gaps"* to your prompt makes it name the processes where it had no figure, which is the fastest audit of your own data you can run. ## How much accuracy you actually need Work out which decision you're making first. The precision each one needs is very different, and so is the effort to get there. | What you're deciding | Precision that's enough | Evidence that gets you there | | - | - | - | | Where to look first — which processes carry the mass | Rank order. Being 50% out is fine. | Companion, with whatever it has already gathered. Spend nothing more here. | | The improvement case for a value stream | ±25% on Active Work Time, and the spread matters more than the point. Cycle Time directional only. | Companion across a full cycle, plus one interview on each of the top few processes | | A funded case — headcount, budget, an SLA commitment | ±10% on Active Work Time and Annual Volume — these two carry the number — with provenance you can show someone | A manually set value with its basis in the change note, signed off by the process owner | | A compliance or audit narrative | Exact, and stable over time | A human-approved figure pinned as a standard, with the observed version kept alongside | The most common mistake here isn't inaccuracy — it's spending audit-grade effort on a prioritisation-grade question. Decide which row you're in before you start gathering. ### The confidence ladder The capture paths on [Maintaining your process attributes](/user-docs/structure/process-attributes#where-the-values-come-from) are ranked by confidence, highest first. This ladder runs the opposite way on purpose: it ranks them by **effort**, cheapest first — so it starts at the lowest-confidence end, and each rung asks the platform to deduce less and tells it more. 1. **Not populated** — Advisor is reasoning without the field at all. 2. **One Companion session** *(low confidence)* — a single pass, possibly a partial one. 3. **Companion across a full cycle** *(low confidence)* — still deduced, but the range now covers the hard weeks. 4. **A recording or document** *(medium confidence)* — a figure the material states, or one observed pass. 5. **A Subject Matter Expert interview** *(high confidence)* — with the basis stated on the record. 6. **A manual edit** *(high confidence)* — with its basis in the change note. 7. **A human-approved figure pinned as a standard** *(high confidence, and held)* — with the observed version kept alongside it. You rarely need to climb past step 3 — low confidence is perfectly adequate for deciding where to look. Steps 6 and 7 are for figures someone will be asked to defend. ## What kind of value are you proving? The table above says how *precise* you need to be. This says *which fields* — because the attributes that carry your argument depend entirely on what you're trying to prove. Almost every case falls into one of four. Hours given back (soft) and money not spent (hard) — headcount, overtime, contractor and vendor spend. Whether the work lands on time and lands right — SLA commitments, error and rework rates, and what it feels like to be on the receiving end. Where process speed turns into cash — DSO, collections, order-to-cash, renewals, expansion and retention. Controls, compliance and audit defensibility — what prevents, detects or corrects error, fraud and non-compliance. ### Where to spend your effort You don't need high confidence on everything — that's the point of having a scale at all. Pick the case you're arguing, put the human effort into the two or three fields that actually carry it, and leave the rest to Companion. **The fields that carry it:** `Active Work Time` × `Annual Volume`, then `Roles and Responsibilities` The first two are the effort total. The third gives you the role — and the role is what turns hours into a blended rate, which is what turns soft dollars into hard ones. The rate itself isn't a Within attribute; you supply that. Sanity-check `Annual Volume` against `Frequency` before you trust it. **The fields that carry it:** `Cycle Time`, then `Exception Handling` This is the case where the field you need is the field you're told to distrust. Cycle Time is directional by default — so if you're committing to an SLA, it has to be stamped by a person rather than inferred. `Exception Handling` is where SLAs actually break, and it's usually more useful than the headline number. **The fields that carry it:** `Cycle Time` and `Scope` In order-to-cash or collections, wait time isn't just waste — it's working capital. DSO is elapsed time, not effort, so Cycle Time is the field that carries it. `Scope` matters just as much: the trigger and end condition are what tell you where the clock starts and stops. Get the boundaries wrong and the number is meaningless however precise it is. **The fields that carry it:** `Key Control`, `Roles and Responsibilities`, `Exception Handling` Never leave `Key Control` inferred — it's binary and high-consequence. Roles carries decision authority and segregation of duties; `Exception Handling` carries the escalation path. `Systems Used` matters here too as the system of record. This is also the case where you pin the result as a standard rather than letting captures move it. **The rule of thumb worth remembering.** Only cost cases live on Active Work Time. SLA, revenue and risk all lean on Cycle Time or the narrative fields — which is why "only stamp Cycle Time when confidence is high" matters more than it first appears. If what you're proving isn't a pure cost case, you are going to need a human-set Cycle Time sooner or later. ## Related articles What the fields are, where their values come from, and how to raise confidence in one. Get an analysis running against the processes you have captured. Take the output from process-level findings to a portfolio view. Pin an approved figure so captures cannot drift it. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.