Forecast the Whole Growth System
Task
Formalize forecasting and executive GTM dashboard.
Summary
Bring pipeline, recurring revenue, churn, expansion, acquisition cost, payback, and channel mix into one view.
Prompt Analysis and Research Execution Brief for S5-11
Executive summary
The attachment is not missing. Three files are available: the controlling assignment in prompt.md, a dashboard presentation, and a supporting dashboard workbook. The assignment is to produce one publication-ready, research-backed Markdown article about formalising forecasting and an executive go-to-market dashboard. It is not an instruction to build the dashboard itself; the dashboard is the business evidence that the article must explain.
The article must remain tightly focused on how a growing subscription business should define, operate, and assess an executive dashboard covering pipeline, annual recurring revenue, churn, net revenue retention, customer acquisition cost, CAC payback, and channel mix. Forecast accuracy is the primary measure, but the stated “target variance” is a company-specific working target, not an industry benchmark.
The most important execution issue is a source conflict. The prompt requires inspection of every supplied resource, but it expressly prohibits the final article from mentioning, citing, paraphrasing, or relying on the publisher of the supplied presentation. The correct treatment is to inspect the presentation and workbook for workflow ideas, quarantine them from the article’s evidence base, and independently verify every publishable point through permitted public sources. The presentation’s useful internal concepts include aligning dashboards with decisions, inventorying data sources, documenting definitions, choosing visuals deliberately, adding commentary and actions, and testing dashboards with users.
The workbook offers a practical sequence—audience and goals, metric selection, data readiness, dashboard design, narrative, and action tracking—but should not be treated as a reliable calculation engine without repair. Local inspection found 1,802 formulas, of which 910 contain broken #REF! references, concentrated in hidden dropdown-support sheets. The visible planning tables remain readable, but automated dropdowns and hidden-sheet logic should not be trusted.
A rigorous execution should take approximately 13–18 person-hours over two to three working days, using one researcher-writer plus short reviews from a revenue operations or finance specialist and an editor. No submission deadline is specified.
Attachment status and prompt synthesis
The missing-attachment contingency in the user’s instruction does not apply. The following resources were inspected.
| Resource | Status and contents | Role in the assignment | Treatment |
|---|---|---|---|
prompt.md | Complete assignment, task record, research rules, voice, structure, source hierarchy, prohibited source, and final quality check | Controlling instruction | Follow directly; it governs all other supplied material. |
Build Better Dashboards…pptx | Thirty-four slides covering goal alignment, data sources, metric selection, dashboard design, storytelling, testing, and iteration; no substantive speaker notes were present | Background process material | Inspect, but do not cite, name, paraphrase, or rely on its publisher in the final article because the prompt expressly prohibits that source. |
Dashboard Workbook…xlsx | Eleven sheets covering audience and goals, metric alignment, data readiness, design review, narrative, and action tracking | Internal workflow and template reference | Use only as private process context. Do not publish its sample figures as evidence. Repair or bypass broken hidden-sheet formulas. |
The prompt’s substantive assignment can be reduced to one operating question:
How should a scaling subscription company turn pipeline, recurring revenue, retention, acquisition economics, and channel results into a credible forecast and an executive decision tool?
The article must explain the work rather than merely listing dashboard tiles. It should show how a leadership team defines the measures, establishes ownership, reconciles channel and company totals, compares forecasts with actuals, interprets variance, and decides whether a channel is producing the right customers at an acceptable cost and operating burden.
A second important distinction concerns output versus editorial evidence:
| Item | What it means |
|---|---|
| Assigned output | A complete, publication-ready Markdown article |
| Business evidence discussed in the article | A working executive dashboard with pipeline, ARR, churn, NRR, CAC, payback, and channel mix |
| Primary business measure | Forecast accuracy |
| Working target | A defined variance tolerance, established by the company |
| Separate dashboard file required by this assignment | No |
| Research plan required as final output | No; the prompt expressly prohibits returning a plan instead of the article |
| Deadline | Unspecified |
Extracted tasks and prioritised action items
The priorities below distinguish mandatory work from useful enhancement. All are derived from the controlling assignment.
| Priority | Extracted task | Required action | Evidence of completion |
|---|---|---|---|
| Critical | Resolve source conflicts | Establish a source quarantine so the prohibited presentation cannot enter the article’s citations, examples, wording, or claims | Source log identifying permitted, contextual-only, and prohibited materials |
| Critical | Inspect every supplied resource | Review presentation text and visuals, workbook sheets and formulas, and the complete prompt | Inspection notes; no claim that a file was reviewed solely from its filename |
| Critical | Define the article boundary | Keep the article on forecasting and the executive GTM dashboard; mention adjacent channel-readiness tasks only when necessary | The article does not become a general SaaS metrics, dashboard-design, or channel-strategy guide |
| High | Establish a metric dictionary | Research and clearly define pipeline, ARR, churn, NRR, CAC, CAC payback, channel mix, forecast, actual, variance, and forecast horizon | A consistent definition and calculation basis for every dashboard measure |
| High | Explain forecasting practice | Cover forecast horizons, snapshots, actuals, error, bias, confidence or scenario ranges, judgemental overrides, and review cadence | Readers can understand how a forecast is produced and tested, not merely displayed |
| High | Connect metrics to decisions | Explain what each metric changes about channel investment, hiring, spending, retention work, or operating capacity | Every major metric is tied to a decision, owner, and expected action |
| High | Research measurement limitations | Explain why similarly named metrics may not be comparable and why percentage-error measures can fail in low-volume or zero-actual periods | Definitions, caveats, and alternative accuracy measures are included |
| High | Find public examples | Use one to three public cases, preferably regulatory filings or official investor material, to demonstrate metric-definition differences or dashboard implications | Each example teaches one specific operating lesson and remains subordinate to the principle |
| High | Draft the complete article | Follow the specified editorial arc and plain-language business voice | A coherent 2,500–4,000-word Markdown article |
| High | Cite every material claim | Use approximately 8–15 strong, freely accessible sources, including primary sources and open research | Inline citations throughout and a final ## Sources section containing only sources actually used |
| Critical | Perform final compliance review | Check focus, citations, word count, source accessibility, claims, acronyms, Mermaid syntax, and all prohibited spelling variants | Clean final text with no mention of attachments, prompts, confidential data, or the prohibited source |
| Medium | Add a useful visual | Include no more than two simple Mermaid diagrams when they materially clarify the forecasting process or measurement model | Diagram renders correctly and has an accompanying text explanation |
| Medium | Improve Canadian relevance | Prefer Canadian regulatory, government, and English-language sources when equally authoritative | Canadian securities, privacy, data-management, and visualisation guidance incorporated where relevant |
The article’s practical questions are also mandatory tasks. It must explain the problem being solved, why the work matters at this stage, what good work looks like, what evidence should exist, what to measure, where judgement is required, how superficial completion fails, and what must be true before the organisation relies on the dashboard.
Deliverables, constraints, success criteria, and assumptions
The following deliverables are required unless explicitly described as optional.
| Deliverable | Required format | Required length or quantity | Deadline |
|---|---|---|---|
| Final article | Publication-ready Markdown | Approximately 2,500–4,000 words | Unspecified |
| Article heading | # [Plain-language article title] | One title | Unspecified |
| Task identifier | **Task ID:** S5-11 | One line beneath the title | Unspecified |
| Opening summary | Markdown blockquote | 40–70 words | Unspecified |
| Main body | Practical business nonfiction with explanatory headings | Enough to cover the full editorial arc without padding | Unspecified |
| Public examples | Verified company, institutional, or public cases | At least one; preferably one to three | Unspecified |
| Research base | Strong, freely accessible sources | Approximately 8–15 | Unspecified |
| Academic evidence | Open or openly available academic research | At least one or two when relevant literature exists | Unspecified |
| Primary evidence | Filings, official guidance, official documentation, annual reports, or equivalent | Required for company facts and quantitative claims | Unspecified |
| Visuals | Mermaid diagram only when useful | Zero to two | Unspecified |
| Source section | Final ## Sources section | Only sources actually used | Unspecified |
| Separate executive GTM dashboard | Not required as a file by this article assignment | Not applicable | Unspecified |
| Research plan or outline as final response | Prohibited | Not applicable | Not applicable |
Explicit constraints
The final article must not be an outline, research plan, idea list, partial draft, company profile, promotional piece, biography, service pitch, generic motivational essay, or general explanation of the complete methodology. It must not invent Matt Edwards’s opinions, experiences, customer outcomes, or quotations. It must not present the tracker’s “target variance” as a universal benchmark. It must not expose private business facts or unsupported claims.
All sources must be publicly inspectable. Search-result pages, anonymous blogs, gated analyst reports, sponsored content, unsourced social posts, and Wikipedia are excluded as evidence. Vendor material may be used for product-specific implementation details, but not for broad performance claims when better independent or primary sources exist.
The article must use ordinary business language. Acronyms such as annual recurring revenue (ARR), net revenue retention (NRR), customer acquisition cost (CAC), annual contract value (ACV), and product-led growth (PLG) must be defined when first introduced. The opening must describe a recognisable operating situation rather than begin with a task identifier, abstract definition, or broad claim.
Inferred assumptions
| Inferred assumption | Basis | Recommended handling |
|---|---|---|
| The primary reader is a founder, executive, finance leader, sales leader, or revenue operations leader in a subscription business | The methodology concerns moving beyond founder-dependent growth and evaluating multiple channels | Write for a commercially literate reader, but define technical metrics |
| The company has progressed beyond initial product-market learning | The task sits in the scaled-channel stage | State that the dashboard should not substitute for unresolved product, onboarding, retention, or data problems |
| “Forecast” primarily concerns bookings, new ARR, expansion, contraction, and churn over a defined horizon | The dashboard combines pipeline and recurring-revenue economics | Make the forecast object and horizon explicit rather than using “forecast” generically |
| Multiple channels may touch the same account | Partners, marketplaces, PLG, expansion, referrals, outbound, and enterprise can overlap | Require attribution and ownership rules to prevent double counting |
| ARR, NRR, CAC, and payback are management metrics rather than universally standardised accounting measures | Public issuers disclose materially different definitions | Require a metric dictionary and reconciliation to recognised financial data |
| A target variance must be company-specific | Prompt explicitly warns against treating it as an industry benchmark | Set tolerances by forecast horizon, sales cycle, segment, channel maturity, and data quality |
| The publishing platform supports native citations and Mermaid | Both are explicitly required or encouraged | Validate syntax before delivery; provide explanatory text in case diagrams do not render |
| The article can discuss a generic dashboard without publishing private company data | No company-specific dataset or approved private example is supplied | Use hypothetical structures and public examples only |
| There are operational dependencies despite “none specified” in the tracker | Reliable forecasting requires definitions, ownership, snapshots, source data, and historical actuals | Identify these as prerequisites, not tracker dependencies |
Success criteria
The article is successful when a reader could use it to answer four practical questions:
- What should the dashboard contain and why? The measures form a connected model of demand, recurring revenue, retention, acquisition economics, and channel contribution.
- Can the numbers be trusted? Every metric has a definition, source, owner, refresh cadence, time boundary, segment rule, and reconciliation method.
- Is the forecast improving? Forecasts are saved as dated snapshots and compared with actual results using an appropriate combination of absolute error, relative error, and bias.
- What decision follows? Leadership can decide where to add capacity, reduce spending, investigate churn, change channel rules, or withhold further scaling.
Recommended research sources
The research should start with definitions and disclosure rules, then move to forecasting evidence, public company examples, and visualisation or data-governance guidance.
| Source category | Recommended sources | Rationale and intended use |
|---|---|---|
| Canadian financial-measure disclosure | Canadian Securities Administrators’ National Instrument 52-112 and companion policy | ARR, NRR, CAC, and payback may be supplementary or non-standard measures. Canadian guidance is useful for requiring clear labels, consistent definitions, comparable information, and transparent calculation methods. |
| U.S. KPI disclosure guidance | U.S. Securities and Exchange Commission guidance on key performance indicators and non-GAAP financial measures | Provides a primary-source model for explaining why a metric is useful, how it is calculated, what assumptions it includes, and whether its definition has changed. |
| Revenue-recognition standards | IFRS Foundation material on IFRS 15 and IFRS 18 | Helps distinguish recognised revenue from operational measures such as ARR and highlights the growing importance of transparent management-defined performance measures. IFRS 18 is effective for annual periods beginning on or after January 1, 2027. |
| Public SaaS filings | Latest filings from companies such as Snowflake and Amplitude | These filings provide maintained definitions of ARR or NRR and warn that similarly named metrics may be calculated differently. Comparing definitions is more useful than copying a single company’s benchmark. |
| Forecast-accuracy research | Hyndman and Koehler’s work on forecast-accuracy measures; Forecasting: Principles and Practice | Establishes the limits of common percentage measures, explains MAE, RMSE, MASE, and rolling-origin evaluation, and supports choosing measures appropriate to the data and decision. |
| Judgemental forecasting research | Fildes, Goodwin, and related open research on forecast adjustments | Supports rules for management overrides: document the reason, preserve the original forecast, and test whether overrides improve accuracy rather than assuming executive judgement always adds value. |
| Canadian visualisation guidance | Statistics Canada’s data-visualisation guidance and Canada.ca chart patterns | Provides Canadian, English-language guidance on audience, message, chart choice, labels, sources, accessibility, simplicity, and accompanying tabular data. |
| Canadian data governance | Government of Canada guidance on FAIR data, stewardship, and metadata standards | Supports source inventories, consistent field definitions, controlled access, data quality, interoperability, and lifecycle accountability. |
| Official system documentation | Documentation from the company’s actual CRM, billing, finance, data warehouse, marketplace, and BI tools | Use only for implementation facts such as snapshot features, API behaviour, field lineage, permissions, and refresh limits; do not use vendor marketing claims as general evidence |
| Independent public reporting | Reputable reporting on a documented forecast failure or metric controversy | Use selectively when it adds context or challenges a company’s official account; verify the underlying facts through filings or official records |
The best public examples will likely demonstrate one of three lessons: the same metric can have different definitions, an operational metric is not recognised revenue, or consumption and channel models require different forecasting logic. A single well-documented example for each selected lesson is preferable to a list of company names.
Research and execution plan
The work should proceed from source control to metric definitions, forecasting evidence, application, drafting, and compliance review.
flowchart LR
A[Confirm scope and source rules] --> B[Inspect supplied resources]
B --> C[Build permitted source register]
C --> D[Define dashboard metrics and decisions]
D --> E[Research forecasting methods and accuracy]
E --> F[Select public examples]
F --> G[Draft complete article]
G --> H[Check evidence, citations, and calculations]
H --> I[Run editorial and prohibited-source review]
I --> J[Deliver publication-ready Markdown]
Text description: the process separates attachment inspection from the permitted evidence base, defines the measurement model before drafting, and ends with both technical and editorial quality control.
| Work stage | Main activities | Estimated effort | Resource needs | Output |
|---|---|---|---|---|
| Scope and source control | Parse the prompt; establish the instruction hierarchy; create prohibited-source quarantine; confirm article-versus-dashboard boundary | 1.0–1.5 hours | Researcher-writer | Scope memo and source rules |
| Attachment inspection | Review all slides, visual structure, workbook sheets, formulas, sample data, and hidden-sheet issues | 1.0–1.5 hours | Researcher-writer; spreadsheet and presentation tools | Resource inspection notes |
| Metric-definition research | Research pipeline, ARR, churn, NRR, CAC, payback, channel mix, revenue, and attribution definitions | 2.0–3.0 hours | Researcher-writer; official filings and regulatory sources | Metric dictionary and definition comparison |
| Forecasting research | Research horizons, snapshots, error measures, bias, overrides, ranges, forecast reconciliation, and backtesting | 2.0–3.0 hours | Researcher-writer; open academic literature | Forecasting evidence matrix |
| Example selection | Identify one to three public cases; verify through current filings and, where useful, independent reporting | 1.0–1.5 hours | Researcher-writer | Example briefs with verified facts |
| Article drafting | Write the complete article, including opening situation, principle, application, evidence, failure modes, and conclusion | 3.0–4.0 hours | Researcher-writer | Full Markdown draft |
| Specialist review | Review accounting/metric distinctions, channel attribution, and forecast operations | 0.75–1.25 hours | Finance or revenue operations reviewer | Corrections and practical validation |
| Citation and editorial QA | Verify every material claim, source accessibility, citation placement, dates, acronyms, voice, and Mermaid syntax | 1.5–2.0 hours | Researcher-writer and editor | Publication-ready final article |
| Prohibited-source scan | Case-insensitive scan for all banned spelling variants; check for unattributed derivative wording | 0.25–0.5 hours | Text-search tool and researcher | Compliance confirmation |
Estimated total: approximately 13–18 person-hours. The work can reasonably be completed over two to three working days, although the actual delivery date remains unspecified.
gantt
title Indicative execution timeline
dateFormat YYYY-MM-DD
axisFormat %b %d
section Foundation
Scope and source control :a1, 2026-08-03, 1d
Resource inspection :a2, after a1, 1d
section Research
Metric and filing research :b1, after a2, 1d
Forecasting research :b2, after a2, 1d
Public example verification :b3, after b1, 1d
section Production
Complete article draft :c1, after b2, 1d
Specialist and editorial review:c2, after c1, 1d
Final compliance check :c3, after c2, 1d
Text description: the indicative timeline sequences the work across three concentrated research and production days. Dates show order rather than a promised deadline.
The minimum resource configuration is one experienced researcher-writer with access to public filings, academic literature, spreadsheet inspection, and Markdown validation. A finance or revenue operations reviewer should spend approximately 45–75 minutes checking metric definitions, channel attribution, and forecasting routines. An editor should perform a final plain-language and compliance review.
Risk analysis and templates
| Risk | Probability | Impact | Mitigation |
|---|---|---|---|
| Prohibited-source material enters the article | High | Critical | Quarantine the presentation and derived notes; use it only to generate research questions; independently source every publishable point; run a final case-insensitive text scan |
| The assignment is misread as requiring a dashboard build | Medium | High | State in the working brief that the final deliverable is the article; treat the dashboard as the evidence and operating artefact the article explains |
| Metrics use inconsistent definitions | High | High | Create a metric dictionary before drafting; specify numerator, denominator, population, timing, exclusions, currency, owner, and source |
| ARR is treated as revenue or as a guaranteed forecast | Medium | High | Explain the accounting-versus-operating distinction and use public filings that explicitly warn ARR is not recognised revenue or a revenue forecast. |
| NRR or churn is distorted by cohort changes | High | High | Freeze the starting cohort; define whether acquisitions, reactivations, free plans, currency changes, services, and usage revenue are included |
| CAC and payback are understated | High | High | Define fully loaded acquisition cost; specify whether partner commissions, marketplace fees, implementation support, sales engineering, and allocated labour are included |
| Channel results are double counted | High | High | Establish account ownership, sourced-versus-influenced rules, hierarchy, lookback windows, and conflict resolution before reporting channel mix |
| Forecast accuracy is measured with one misleading percentage | High | High | Report error and bias separately; use absolute and scaled measures; define treatment of zeros and low-volume channels; compare like horizons |
| Executive judgement overwrites forecasts without accountability | Medium | High | Save both system and adjusted forecasts, require a written reason for material overrides, and measure forecast value added |
| “Target variance” is presented as a benchmark | Medium | High | Label it a working tolerance; derive it from historical error, planning needs, horizon, sales cycle, and channel maturity |
| Article scope expands into general GTM strategy | High | Medium | Apply a scope test to every section: it must materially improve forecasting, dashboard credibility, or the supported channel decision |
| Current company facts become stale | Medium | Medium | Use the latest filing available on the execution date; state fiscal periods and measurement dates |
| Private or sample data is published as fact | Low | High | Treat workbook values as placeholders; use public examples or clearly labelled hypothetical figures only |
| Workbook automation produces incorrect outputs | High | Medium | Do not rely on hidden dropdown formulas; use visible sheets as conceptual templates and rebuild calculations independently |
| Mermaid or citations fail to render | Low | Medium | Validate syntax; keep diagrams simple; include a short text explanation; perform a final preview |
Article opening template
# [Plain-language article title]
**Task ID:** S5-11
> [A 40–70 word summary explaining what the article teaches, the decision it supports, and why a credible forecast and dashboard matter before the company scales more sales channels.]
Metric-definition template
| Field | Example specification |
|---|---|
| Metric name | Net revenue retention |
| Business question | Is the existing customer base growing or shrinking before new-customer revenue? |
| Formal definition | Ending recurring revenue from the starting customer cohort divided by starting recurring revenue from that cohort |
| Included changes | Renewal, expansion, contraction, and churn |
| Excluded changes | New customers acquired after the starting date |
| Cohort rule | Customers with qualifying recurring revenue at the beginning of the measurement period |
| Time basis | Trailing twelve months, measured monthly |
| Currency rule | Constant currency or reported currency, explicitly identified |
| Source systems | Billing, contracts, product usage, CRM |
| Data owner | Finance or revenue operations |
| Refresh cadence | Monthly close, with weekly operational estimate if needed |
| Segmentation | Channel, customer size, product, geography, acquisition cohort |
| Reconciliation | Explain how the operating metric connects to recognised revenue and contract data |
| Known limitations | Usage volatility, contract timing, acquisitions, reactivations, credits |
| Decision triggered | Investigate contraction or churn by channel; reassess acquisition spending and expansion capacity |
Forecast-accuracy scorecard template
| Measure | Calculation | Why include it |
|---|---|---|
| Forecast error | Actual minus forecast | Shows direction and size for one period |
| Absolute error | Absolute value of actual minus forecast | Prevents positive and negative errors from cancelling |
| Bias | Sum of forecast minus actual over a defined period | Identifies persistent optimism or conservatism |
| Weighted absolute percentage error | Sum of absolute errors divided by sum of actuals | Gives an interpretable portfolio-level percentage when totals are positive |
| Mean absolute scaled error | Absolute error scaled against a naïve benchmark | Supports comparison across differently sized series and avoids several weaknesses of percentage error |
| Forecast value added | Accuracy after a process step compared with accuracy before it | Tests whether management overrides or added process complexity improve the result |
No single accuracy measure is sufficient in every setting. Percentage errors become unstable or undefined when actual values are zero or close to zero, while scaled measures and rolling-origin tests are more appropriate for comparing methods across series and horizons.
Evidence matrix template
| Claim or lesson | Evidence required | Preferred source | Counterpoint or limitation | Intended article section |
|---|---|---|---|---|
| ARR is not the same as recognised revenue | Accounting standard plus public filing definition | IFRS Foundation and latest issuer filing | ARR definitions differ by company | Metric definition |
| NRR requires a fixed cohort | Maintained public-company definition | Regulatory filing | Consumption businesses may show more volatility | Measurement |
| Judgemental overrides can reduce accuracy | Open academic research | Original paper or institutional repository | Material overrides based on new information may add value | Failure modes |
| Forecast error should be assessed by horizon | Forecasting text or research | Open academic source | Executive planning may require several parallel horizons | Application |
| Channel mix needs attribution rules | Official platform documentation plus operating analysis | CRM or marketplace documentation | “Sourced” and “influenced” may serve different decisions | Dashboard design |
Bibliography
Supplied resources
- Deep Research Article Assignment — S5-11. Controlling assignment and editorial instructions.
- Build Better Dashboards Across Marketing, Sales, and Customer Success. Presentation inspected as supplied context only; prohibited from use as evidence in the final article.
- Dashboard Workbook for Marketing, Sales, and Customer Success. Spreadsheet inspected locally for workflow, sample tables, formulas, and data-readiness fields.
Primary and official sources
- Canadian Securities Administrators. National Instrument 52-112: Non-GAAP and Other Financial Measures Disclosure and companion policy.
- U.S. Securities and Exchange Commission. Commission Guidance on Management’s Discussion and Analysis of Financial Condition and Results of Operations.
- U.S. Securities and Exchange Commission. Non-GAAP Financial Measures: Compliance and Disclosure Interpretations.
- IFRS Foundation. IFRS 15 Revenue from Contracts with Customers supporting and post-implementation material.
- IFRS Foundation. IFRS 18 Presentation and Disclosure in Financial Statements.
- Snowflake Inc. Fiscal 2026 Form 10-K and fiscal 2027 first-quarter filing.
- Amplitude, Inc. 2025 Form 10-K.
- Statistics Canada. Data Visualization: Best Practices.
- Government of Canada. Guidance on Assessing Readiness to Manage Data According to the FAIR Principles.
- Government of Canada. Be Good Data Stewards.
Open and academic research
- Hyndman, Rob J., and Anne B. Koehler. “Another Look at Measures of Forecast Accuracy.” International Journal of Forecasting.
- Hyndman, Rob J., and George Athanasopoulos. Forecasting: Principles and Practice, sections on forecast accuracy and time-series cross-validation.
- Fildes, Robert, Paul Goodwin, and Shari De Baets. “Forecast Value Added in Demand Planning.” International Journal of Forecasting.
- Fildes, Robert, and Paul Goodwin. “Against Your Better Judgment? How Organizations Can Improve Their Use of Management Judgment in Forecasting.” Interfaces.