Getting started / intake and upload

This is the first few minutes with the product, and it is built to ask for almost nothing before showing something real. A customer brings the documents they already have, in whatever formats they already exist in, and is told plainly, before uploading anything, what will and will not happen to those files. Nothing about intake requires cleanup, conversion, or a setup project. A visitor who has not decided to buy yet can still see a full sample report first, so the depth of the product is visible before you are ever asked to buy anything.

• A plain-language starting screen. The intake screen opens with "Let's find the untapped value buried in your business" and one sentence explaining what happens next: "Upload what your business already created. ValueAtlas saves it to a secure analysis under your organization, then finds what's worth fixing." There is no setup wizard, no configuration screen, and no integration step before you can start.
• Bring your documents as they already exist. You upload the operational documents your business already has: SOPs, policy PDFs, spreadsheets of cycle-time or audit data, BPMN process diagrams, email threads, meeting transcripts, and informal notes. In the demo analysis, 26 files were uploaded in one batch, including .txt, .docx, .xlsx, .pdf, .bpmn, .csv, and .xml files side by side, with no need to convert or clean anything first.
• A wide, published list of accepted file types. The product accepts PDF, Word, Excel, PowerPoint, images, and text and data files, with an explicit extension list (.pdf, .docx, .xlsx, .xlsm, .pptx, .pptm, .jpg, .jpeg, .png, .txt, .md, .csv, .json, .xml, .html, .log, .yaml, and more), so you rarely have to wonder whether a file will be accepted.
• A running file list with sizes, and easy removal. As you add documents, each one appears in a list with its file size and an individual remove control, and the running count at the top ("0 files selected," then "26 files selected") always tells you exactly what is queued for the analysis before you run it.
• A visibility choice made explicit at the moment you choose it. A toggle labeled "Visible to your organization" with the plain description "Everyone in the organization can open this analysis" lets you decide, before you even upload, whether this analysis is shared with your whole team or kept private to you. You can change this later per analysis.
• A firm, plainly stated data-handling promise at the point of upload. The upload screen states directly: "Read-only. Your files are never moved or changed." You are told this upfront, not buried in a privacy policy you have to go find.
• A way to preview the product before uploading anything of your own. A "View a sample report (demo data)" link lets a new visitor see a full example report before they ever upload a real document, so you can judge the depth of the output before you upload your own files.
• A single "Begin analysis" action. Once you have added documents, one button starts the process. There is no multi-page setup form standing between your upload and your first real result.
• Recognizing a document you've already given it. If you upload the exact same file again on a later run (byte-for-byte identical, for example re-uploading the same spreadsheet), ValueAtlas recognizes it and reuses what it already extracted from that file rather than re-processing it, which saves you both time and the cost of re-reading it.

Scope and process selection

Before ValueAtlas starts analyzing anything, it tells the customer what it actually found in their documents and lets them make the call. If a batch of files turns out to describe several different processes, that is surfaced honestly rather than blended into one muddled result, and the customer picks exactly one to focus on. What follows is a real negotiation over scope: which documents count, who is involved, what is being left out, all restated in plain language and open to correction before the paid work starts. The documents a customer excludes are never read at all, not just hidden from the output, which is what makes this stage a genuine boundary rather than a formality.

• Automatic detection of more than one process in your documents. If your uploaded documents describe several different business processes, ValueAtlas detects each one separately and shows them to you as candidate processes to analyze, each with a short, specific summary of what stood out in that process (for example, a summary naming a specific policy non-compliance rate and a specific approval bottleneck) rather than a generic label.
• You choose exactly one process to analyze at a time. ValueAtlas explains this directly: "These documents describe 3 different processes. I analyze one at a time so the findings stay focused. Choose the one to analyze now; the others are set aside for a later run." This keeps every analysis sharply focused on one process instead of producing a muddled result that tries to cover everything at once.
• A confirmable document checklist per process. Below the process choice, you see exactly which of your uploaded documents ValueAtlas proposes to use for this analysis (for example "8 of 26 will be analyzed"), and you can check or uncheck individual documents before you confirm.
• Documents you exclude are genuinely never read for this analysis. If you leave a document unchecked, it is not sent to the underlying analysis at all, not merely hidden from the output. You get a real guarantee that your other, unrelated documents were never looked at, not just a promise that they were filtered out afterward.
• A second scope screen that restates the boundary in plain language. Once you pick a process, ValueAtlas shows you what it understands the process to include (for example, the dollar bands and approval tiers it found), who is involved (tagged names and roles pulled from your documents, such as specific approvers, departments, and committees), and what it is deliberately leaving out, so you can catch a misunderstanding before the real analysis runs.
• The scope decisions update when you change your process choice. If you switch which process to analyze, the "what's inside," "people involved," and "left out" summaries regenerate for the newly chosen process instead of staying stuck describing whichever process the system originally guessed was primary.
• An optional free-text scope adjustment. A note field lets you tell ValueAtlas anything it should adjust about the boundary before it runs (for example, "start the clock at submission, not at manager approval"), and you can review your adjustment before confirming.
• A way to add more documents mid-scope. If you realize a relevant document is missing once you see the proposed scope, you can add it right there without restarting the whole intake, and only the newly added files are checked against the process you already chose (not a full re-classification of everything you uploaded).
• A clear "Confirm scope & analyze" checkpoint. Nothing runs against your data until you explicitly confirm the scope, giving you one last chance to change the process, adjust which documents are in or out, or add a note before the real work (and any cost) begins.

The interview: a short set of targeted questions

Documents rarely say everything, so ValueAtlas closes the gap with a short, bounded interview instead of asking the customer to write a brief from scratch. Every question states plainly why it matters, comes with an editable best guess already filled in, and can be skipped without derailing the rest of the process. A visible sense of progress, capped at a handful of questions, keeps this from ever feeling like an open-ended form.

• A guided interview that fills in what your documents don't say. Before the full analysis runs, ValueAtlas asks you a short series of targeted questions (a handful, capped at eight) about the specific gaps it found in your documents, for example who the designated backup approver is when the primary is unavailable, or what event exactly starts an SLA clock.
• Every question explains why it's being asked. Each question card includes a "Why I'm Asking" note that states, in plain language, why the answer matters to the analysis (for example, "without a documented backup owner, PO's sit in the queue indefinitely when the VP is out, which is the single largest documented driver of cycle time"), so you understand the stakes of each answer rather than being asked to fill in a form blind.
• An editable, pre-drafted best guess for every question. For each question, ValueAtlas shows you its own best guess at the answer based on everything it read in your documents (clearly labeled "Use my best guess"), which you can accept as-is, edit, or overwrite entirely with your own answer in your own words. You are never starting from a blank box.
• Skip any question you can't answer. Every question has a Skip control, so an unanswered gap does not block the rest of the interview or the analysis; ValueAtlas simply proceeds with what it has and notes what it could not confirm.
• A visible progress indicator through the interview. A row of dots shows how many questions remain and how far along you are, and the last question replaces "Next" with "That's everything," so you always know exactly where you are in a short, bounded process rather than an open-ended questionnaire.

The analysis run: live, honest narration of real work

Once the customer confirms scope and answers what they can, the actual analysis runs in view rather than behind a spinner. What appears on screen is a real, running account of the work as it happens, with specific counts and findings from the customer's own data filled in live, not a canned animation standing in for progress. A customer can cancel a run they no longer want, and the final dollar figure arrives as a deliberate, visible moment rather than a page simply refreshing into a finished report.

• A four-stage progress bar for the whole intake-to-report journey. Select, Scope, Answer, and Analyze are shown as a persistent top-of-screen progress bar throughout the entire flow, so you always know which phase of the process you're in.
• Real-time narration of what the analysis is actually doing, in plain language, as it happens. Rather than a generic spinner, the Analyze stage shows a running list of milestones as they complete: "Reading your 10 documents," "Identifying every step and who owns each step: 19 steps across 7 people," "Following how the work actually flows, hand to hand," "Finding where time and effort are lost," "Where is the process constrained? At 'VP Finance Approves'," "Tracing each problem back to what really causes it," "Redesigning the process around what the evidence shows," and "Pricing the recoverable value." Each line reflects a real, completed stage of the underlying work, not a scripted animation, and specific counts (documents read, steps identified, people involved, the actual name of the bottleneck step) are filled in live from your own data as they're discovered.
• A cancel option while the analysis is running. A "Cancel analysis" link lets you stop a run in progress if you need to; you don't have to wait out a run you no longer want.
• A visible, animated reveal of the headline number. The report doesn't just appear; the final dollar figure counts up on screen (starting from "$0/year") as the analysis concludes, giving you a clear, deliberate moment where the final number is delivered rather than a page simply refreshing.

Summary of Initial Findings and Observations

This is the Summary of Initial Findings and Observations, the section that opens the report before either the Executive Summary or the Comprehensive Analysis is opened. It gives a reader what a decision requires in seconds: one dollar range, one paragraph explaining what drives it and who owns fixing it, and a stated basis for how solid that figure is. Every deliverable can be downloaded separately and in the format the reader actually needs, and small conveniences, like naming the report's cover, keep this section from reading like a cover page and nothing more.

• A single, plain-language headline finding. At the very top of the report, before any detail, you get one dollar range per year and a one-paragraph plain-English summary of what drives it, who owns fixing it as the process owner, and who is accountable overall (for example naming the specific person who owns the process and the specific person accountable for the fix). This is a decision an executive can read in seconds.
• A stated basis for every headline figure. Every top-line number carries a caption such as "Based on your numbers plus reasonable benchmarks," so you know upfront whether a figure is drawn purely from your own data or is leaning on an industry benchmark to fill a gap.
• Two report sections you can open independently. "Executive Summary & Highlights" ("The value, the diagnosis, the few changes that matter") and "Comprehensive Analysis & Findings" ("The full deep dive: process, waste, root causes, plan") are shown as two collapsible panels, so a reader who wants the short version and a reader who wants the full backup material are both served from the same report without either having to scroll past the other.
• Independent downloads for every deliverable, in two formats. You can download the Executive summary, the Full written report, and the Process maps separately, each as either PDF or Word, so you can drop the right level of detail into whatever downstream document or presentation you need without editing a single combined file.
• Report cover customization. If no company name is set on the report cover, the report surfaces this directly with an "Add one" link, so you can put your own organization's name on the document you're about to hand to executives or a board.

Executive Summary & Highlights

This is the section built for someone who has a few minutes and a decision to make. It states the diagnosis in plain language, shows exactly which numbers move and by how much, separates free quick wins from the paid recommendations that carry the real value, and lays out a two-phase plan a team could act on immediately. Nothing here pretends to more certainty than it has: figures that rest on a benchmark instead of the customer's own data say so, and value that cannot yet be priced is still named rather than dropped.

1:09

Executive Summary

The value, the diagnosis, and the few changes that matter, in the length of a coffee break.

This video is shown unedited, so the runtime reflects real processing time. Feel free to drag the progress bar to skip ahead. Total length appears in the lower left when playback starts, and updates to show time remaining as the video plays.

Download the sample PDFExecutive Summary · 19 KB

• A plain-language Executive Overview paragraph. A few sentences state the core diagnosis in ordinary business language (for example, naming the specific SLA breach rate, the specific single point of failure, and the specific off-system workaround driving the problem), with no jargon and no methodology terminology.
• A before/after measures table. A short table lists each measure that matters (for example average cycle time, percentage of non-compliant email approvals, overall policy compliance, percentage of purchase orders to unregistered vendors, percentage above delegated authority), each row showing today's value, the target value if the recommendations are adopted, and the size of the change, so you can see exactly what moves and by how much.
• An explicit disclaimer box on any estimated figure, naming exactly what data is missing. When a figure can't be fully itemized because your documents didn't state a specific timing or cost, the report says so directly: "This figure is an estimate. Your documents did not state the timing of any step this redesign frees, so an itemized figure could not be counted. Provide the step timings and the analysis re-prices." You are always told when a number is a benchmark-based estimate rather than a figure computed straight from your own data, and told exactly what would upgrade it.
• Quick Wins: cheap, immediate actions called out separately from the paid recommendations. A small set of low-effort, this-month actions (for example enabling a hidden simplified submission role, or building a status report from data you already have in your existing system) are shown up front, each tagged "LOW EFFORT," described in one line each. These carry no dollar value of their own (the report is explicit that "value belongs to the pool and is claimed once, through the recommendations") but they unblock the real recommendations and make the constraint visible before any bigger change is made.
• Recommendations grouped into value pools, not scattered as a flat list. Every recommendation belongs to one of a small number of named value pools (for example "Faster approval throughput at the constraint" or "Less rework from incomplete and misrouted submissions"), each pool carrying its own dollar range per year. Recommendations within a pool are ordered by the sequence you'd actually do them in.
• Every recommendation states its value, what it removes, and its risk in the same structure. Each recommendation card is written the same way: the value it creates in plain language, the specific obstacle it removes, the risk of making this change, and how that risk is mitigated (for example, "risk: a named delegate is registered but never actually used, so the queue still forms around one person; mitigation: VP Finance owns a two-person standing roster with an automatic routing rule").
• Recommendations are tagged by type. Each recommendation carries a plain tag such as "QUICK WIN" or "STRATEGIC MOVE," so you can tell at a glance whether something is a fast, low-risk fix or a bigger structural change worth planning for.
• Value pools that can't yet be priced are shown honestly, not hidden. When a whole value pool depends on a cost your documents never stated (for example the per-incident cost of an out-of-policy approval), the report shows it as its own pool labeled "Not yet priced" rather than guessing or omitting it, and states plainly what number is needed to price it.
• Savings the analysis found but could not price are listed by name, not dropped. A dedicated list ("Savings we found but could not price") names every specific step that the recommended process frees up time on, tagged with who owns that step, even when the documents don't state how long the step takes, so nothing discovered is silently thrown away for lack of a number.
• A two-phase implementation timeline. "0-30 days" and "1-3 months" columns each show exactly which recommendations they deliver and a short bullet list of concrete actions, so the report reads as a plan you could hand to a team, not just a diagnosis.
• A directional "Months 4 to 6 Outlook." Beyond the 90-day plan, the report gives a forward-looking narrative about what to check on later (verifying the fix actually stuck, completing training, extending reporting) rather than inventing dated tasks that far out, and explicitly says this section is thematic rather than a scheduled action list because the next analysis is the real checkpoint.
• A stated list of what the next analysis will re-measure. The report tells you in advance which specific numbers a follow-up analysis will check again (for example the cycle time against the SLA, the email-approval share, the rework rate), so you know what "success" will be measured against next time, and can plan when to come back.

Comprehensive Analysis & Findings

This is the section built for someone who wants to check the work, not just read the conclusion. It walks the full current and future process step by step, with swimlane maps, RACI accountability matrices, a named and evidence-cited bottleneck, a root-cause diagram, and a complete waste breakdown, all of it grounded in citations back to the customer's own documents. A reader who wants to challenge any single claim can trace it to its source in one click, which is what turns this section from a summary into real backup material.

6:42

Comprehensive Report

Every part of the backup material: the process maps, the root causes, the accountability matrix, and every figure cited back to the document it came from.

This video is shown unedited, so the runtime reflects real processing time. Feel free to drag the progress bar to skip ahead. Total length appears in the lower left when playback starts, and updates to show time remaining as the video plays.

Download the sample PDFComprehensive Report · 116 KB

• A stated process boundary and role list. The comprehensive report opens with the exact start and end point of the analyzed process (for example "from Purchase Need Identified to PO Approved and Sent to Vendor") and every role included in it, so there's no ambiguity about what was and wasn't in scope.
• A worked-through explanation of what each dollar figure rests on. A dedicated "What these figures rest on" section states, pool by pool, which numbers are well-documented in your own records and which ones lean on a labor-rate benchmark because your documents didn't give a precise figure, and closes with a direct, specific ask for what additional numbers would tighten the range (for example "share the full-year PO counts by band and a rough cost per audit exception and we can tighten these ranges into firmer figures").
• A way to unlock pricing on a pool yourself. When a value pool can't be priced because a unit cost or volume wasn't in your documents, the report shows exactly how many times the underlying event happened (pulled straight from your own documents, for example "144 out-of-policy PO approvals a year, from your documents") next to an "Add your cost and volume" control that lets you supply the missing number directly, so you can finish pricing that pool yourself without waiting for a new analysis run.
• An Impact by Category matrix. A grid crosses People, Process, and Technology against Measurable and Qualitative impact, so the report captures real value that a dollar figure alone would miss (for example "VP Finance no longer single-handedly gates every $15,001 to $50,000 PO" is recorded even though it has no dollar figure of its own).
• A consolidated Risks and Mitigations section , gathering every recommendation's stated risk and mitigation into one place with source citations, so a reviewer can scan every risk in the plan without reopening each recommendation individually.
• Recommendation dependency notes. Where one recommendation can't start until another is done (for example "R2 cannot start until R1 is done. Coupa-only enforcement cannot start until infrequent users have a usable interface"), the report says so directly, which keeps the implementation roadmap realistic rather than a list of independent items.
• Automatic detection of contradictions between your own documents. A "Conflicts we found in the documents" table lists every place your source documents disagree with each other (for example a policy spreadsheet and a process diagram naming different dollar thresholds for the same approval tier), stating plainly what each document says, where that leaves the process, and which documents are in conflict.
• Each conflict is tracked as resolved or still open. Every detected conflict is tagged either "still open" (an unresolved contradiction the analysis could not settle on its own) or "Resolved," and a resolved conflict cites exactly which answer you gave in the interview settled it. This means the report doesn't just flag inconsistencies in your own documentation, it tells you which ones your own answers already fixed and which ones still need a decision from your team.
• Every extracted statistic is listed with its own source citation. A running "Key Numbers" list shows every individual fact the analysis pulled out (cycle times, breach counts, approval volumes, compliance rates, staffing figures) each tagged with the exact source document it came from, so nothing in the analysis is presented as fact without telling you where it came from.
• A visual current-state process map you can enlarge. The current process is rendered as a swimlane diagram (organized by role, showing tasks, decisions, waits, and the end point) with a "click to enlarge" control that opens it full-screen for closer reading, and a shared numbering scheme (S for start, T for task, G for decision, Q for wait/queue, E for end) that ties every box on the diagram to the matching entry in the written step list below it.
• A full written walk-through of every process step. Every step in the current process is described in a plain sentence, tagged with the role responsible for it, and cited to the documents that describe it.
• Step-level notes on who does the work versus who owns the outcome. For every step, the report separately states who does the work, who owns the outcome, who has a voice in the decision, and who gets a heads-up afterward, and explicitly flags when a role assignment is "inferred" rather than stated outright in your documents, so you can tell fact from reasonable inference at a glance.
• A standalone Baseline Capability table , restating the "today" numbers for every tracked measure on their own, separate from the improvement table, so the current state can be referenced without scrolling back to the recommendations section.
• A full RACI-style Accountability Matrix for the current process. Every step is crossed against every role in a table marking who does the work (R), who owns the outcome (A), who has a voice (C), and who gets a heads-up (I), giving a complete, structured picture of accountability across the whole process, not just a narrative description.
• Accountability Gaps called out explicitly. A dedicated list names every step where ownership is implied by context but never explicitly stated anywhere in your documents (for example "Step 15 (Wait in CFO Queue): CFO is the implied owner, but the documents do not state it explicitly"), so a genuine documentation gap in your own process is surfaced, not silently assumed away.
• A visual future-state process map, matching the same diagram style. The recommended process is rendered as its own swimlane diagram, also enlargeable, using the same step-code scheme as the current-state map so you can compare the two directly step by step.
• Every future-state step tagged with what kind of change it is. Each redesigned step carries a plain tag such as "Improved," "New step, automated," so you can immediately see which steps are unchanged, which are upgraded, and which are entirely new.
• A Gap Assessment table , restating the same measures with three columns: today, recommended, and the size of the gap closed, giving one clean summary of exactly how much each number is expected to move.
• A full RACI Accountability Matrix for the recommended process , in the same format as the current-state matrix, so you can see exactly how responsibility shifts under the new design, not just that it does.
• An Accountability Changes summary. A "what changes from today" callout lists every new step being added to the process (for example a new automated routing check), giving a quick answer to "what's actually different" without reading the whole future-state walk-through.
• A named, evidence-cited Dominant Constraint. The report identifies the single step that limits the whole process's throughput, states the specific measured cost of that constraint (days added, dollars lost), and backs the claim with cited evidence from your own audit data, cycle-time logs, and even direct quotes from your own leadership documents when available (for example a specific manager's own words about the bottleneck, cited to the transcript it came from).
• A "what it means" plain-language translation of the constraint's evidence. After citing the raw evidence for the bottleneck, the report restates in plain business language what that evidence actually means for how the process runs day to day.
• A visual root cause diagram, also enlargeable. A cause-and-effect (fishbone-style) diagram traces the root causes behind the constraint, organized into categories (for example People, Process, Cross-cutting, Inputs, Measurement), each branch backed by a plain-language explanation and its own source citation.
• A ranked waste chart. A Pareto-style bar chart ranks the categories driving operational friction by how often they occur, with a cumulative percentage line overlaid, and is also enlargeable for closer reading.
• A complete eight-category waste breakdown, every category shown even at zero. A table walks through all eight classic categories of process waste (Defects, Waiting, Transportation, Inventory, Motion, Over-processing, Skills, Overproduction), each with a count of affected steps, a plain-language description of the main driver, and source citations. Every category is shown, including ones with low or zero counts, so you see the complete picture rather than only the worst offenders.
• A one-line "where to concentrate first" closing recommendation , giving a clear, single next step to focus on before the full plan, right at the end of the deep-dive analysis.
• A full Source Documents appendix. Every document used in the analysis is listed with its reference code, its filename, and a one-line plain-English description of exactly what it contributed to the findings, functioning as a complete bibliography for the entire report.
• Inline citation chips throughout the report. Nearly every statistic, quote, and claim in the comprehensive report carries a small clickable reference tag (for example "S18") that links down to the exact source document in the appendix, so any finding can be traced back to its evidence in one click.

Not another AI tool. A harness built around one.

Language models read well and write fluently, and they are unreliable accountants: ask one the same question twice and the answers can drift, and left unsupervised, it can invent a plausible number rather than admit it does not know. ValueAtlas is built so neither failure reaches your report, not by using a better AI, but by building a better harness around one. The AI reads your documents and writes what it finds. The harness makes sure every finding, every decision, and every dollar comes from recorded facts and auditable arithmetic, never from the model's imagination. Your answers and corrections outrank the AI, every time.

Source traceability and evidence

This is not a screen so much as a discipline that runs through every screen: every figure is tagged to the document it came from, every document is accounted for as analyzed, set aside, or used as context, and every gap or contradiction in the customer's own paperwork is stated openly instead of quietly smoothed over. It is what makes the rest of the report something a skeptical reader can actually verify rather than simply trust.

• Every figure is tagged to the document it came from. The report states this as an explicit rule at the top of the Comprehensive Analysis & Findings section: "Every figure is tagged with the document it comes from... and links to Source Documents at the end," and this promise holds throughout, not just in the appendix.
• A stated document intake summary. The report tells you exactly how many documents were read, broken down by format (texts read directly, Word documents with text extracted, spreadsheets read as tabular data, PDFs with text extracted, process diagrams parsed as process models) and an estimated total page count, so you know the real scale of what was actually reviewed.
• A transparent split of your documents into analyzed, set-aside, and cross-cutting. The report states plainly which of your uploaded documents were directly analyzed for this process, which were set aside because they described a different process, and which were used as cross-cutting context (organizational or policy documents relevant to more than one process) without being a primary source, so you can see exactly how your whole document set was used.
• Honest handling of missing or contradictory data throughout, not just in one section. The same "tell you what it can't support" discipline shows up everywhere a number is presented: an unpriced pool, an inferred (not stated) accountability owner, an unresolved document conflict, and an estimate disclaimer are all marked the same honest way rather than smoothed over.

Decisions and reopening figures

Every dollar figure in a ValueAtlas report is the product of specific, visible decisions, not a black box, and the customer can question any of them. Reopening a settled answer reprices the figure in about the time it takes to save the change, with no new analysis run and no charge, and the report's own prose updates to match without breaking any of its citations. A number the customer supplies is never lost or silently overwritten either; corrections are always possible, and the original source material stays on record underneath them.

• Every dollar figure is traceable back to the specific steps and recommendations that produced it. The underlying decisions about what a recommendation actually removes, what it merely improves, and what a root cause actually explains are all recorded and linked together, so a recommendation, a root cause, and a dollar figure can never silently disagree with each other in the report.
• Itemized pricing when your documents give exact timing. When your documents state precisely how long a step takes and how often it happens, ValueAtlas prices the actual time freed by role and rate rather than falling back to a rougher benchmark estimate, and the report tells you plainly whether a given figure is itemized from your own numbers or is an estimate.
• A "Decisions behind this figure" section on the report. A dedicated block shows the actual judgment calls baked into the priced figure (for example, whether a specific step goes away entirely or is simply made faster), stated in plain words, and names who settled each one: you, your documents, or the analysis itself.
• Change your mind on any settled decision, right on the report. Every decision behind the figure carries a "Change this answer" control. Reopening it asks you the same plain question again and reprices the whole figure in about the time it takes to save your answer, with no need to rerun the entire analysis or pay for a new run.
• Your decision carries forward. Once you've answered or corrected a decision, that answer is remembered and automatically applied to every future run of that same process, so you never have to re-answer the same judgment call twice.
• You can fill in a missing step timing yourself, and the figure updates immediately. On any value row that couldn't be priced for lack of a specific timing, you can enter the actual minutes-per-occurrence and how often it happens yourself, and the report reprices right away using your number, clearly labeled as something you supplied.
• Correcting a number you've supplied is always possible, without losing anything. If you realize you entered a figure wrong, you can update it again later.
• Your correction never overwrites what your documents actually said. A number you supply that conflicts with what your documents state is stored alongside the original, not on top of it, so your original source material is always preserved and the analysis prices from your latest, most accurate number.
• Report prose updates automatically to stay consistent with a correction, without breaking citations. When a figure changes because you corrected it, the sentences that quote a dollar amount update to match; every citation and every sentence that isn't quoting the changed number stays exactly as it was, so a correction can never cause the rest of the report's evidence to drift out of sync with itself.