Replace analyze our competitors with a decision
A local sports club is evaluating coach-rota software: retain spreadsheets, buy an existing service or build a narrow shift-swap entry point? This fictional example uses services Alpha and Beta plus the manual process to connect information collection, feature breakdown, SWOT and industry research. All prices and organization counts are synthetic teaching assumptions rather than procurement or investment recommendations.
Agree on products, geography, observation period, budget scenario, required workflows and the report's audience before research. The owner needs conditions for selecting an option and a next verification step, not the largest collection of screenshots. Model knowledge can suggest questions, but current features, numbers and versions need accessible primary evidence. Missing information remains unknown.
Scenario one: collect comparable pricing and feature facts
List direct competitors, indirect substitutes and the existing process. Direct rivals address coach scheduling; substitutes might include shared calendars; spreadsheets have flexibility and migration costs of their own. Avoid selecting only feature-rich products or changing dimensions after reading sources to favor a preferred answer. Fix three user tasks first: publish a weekly rota, handle a last-minute swap and explain changes to affected people.
For each source record the address, title, observation date, plan, version, region, language, relevant passage and usage constraints. Pricing pages support prices, help documents support rules and authorized trials support interaction observations. Reviews are investigation signals. Preserve screenshot context rather than cropping away limitations. If a page is inaccessible, record the gap instead of inventing a browsing result.
Keep currency, billing unit, monthly versus annual payment, minimum seats, quotas, trials, add-ons and unknown costs in separate fields. Suppose Alpha costs ten units per seat per month with a five-seat minimum and Beta costs sixty per organization per month. Eight coaches allow a base-cost comparison under those fictional assumptions, while real tax, discounts and extra services remain unresolved. An annually billed monthly equivalent is not the same commitment as monthly payment. Do not combine currencies using an unverified exchange rate.
Define support before rating it. A complete swap may require a request, acceptance by another coach, manager approval and a rota update. Calendar editing alone does not establish that workflow. Use complete, partial, unverified and not-applicable states with conditions. A visible button does not demonstrate audit records, mobile support or correct authorization. Where the product offers a workaround, describe its effort and failure points instead of treating it as identical native support.
Ask for a small matrix first. Manually audit the key pricing, swap and notification cells before expanding it. Keep facts, interpretation and recommendations distinct and cross-reference them through identifiers. Channels, positioning and user groups can be included when material supports them; do not turn an imagined typical coach into a research result.
Use a data skill only on the prepared input
Once an authorized matrix exists as a local CSV, Exploratory Data Analysis can profile fields, missingness, duplication and aggregate anomalies. It may expose a missing plan, text in a numeric field or inconsistent-unit clues. It does not crawl pricing pages or interpret a real contract.
Read the pinned SKILL.md for supported formats, scope and dependencies. Core CLIs require Python 3.11 or later, while the full optional environment has additional requirements. Bound the input directory and aggregate outputs, preserve original files and avoid identifying raw records or silent row removal. Stop on unsupported formats. Return suspicious prices to their sources; an expensive option is not automatically noise to delete.
Scenario two: inspect tasks before building SWOT
Conduct an authorized walkthrough using synthetic coaches and shifts. Record each task's starting state, steps, outcome, failure path and recovery. Alpha might reduce clicks while leaving approval ownership unclear; Beta might require more steps but expose a useful change history. Those are observations requiring evidence, not conclusions inferred from appearance. If a trial cannot be run, label the process as documentation-based inference and list what was not experienced.
Keep the environment reproducible: product version, account role, device, fixture, time and relevant settings. A manager's successful action does not prove a volunteer has the same permission. A browser screenshot cannot establish an unseen server-side guarantee. Separate observed behavior from the underlying technical explanation and ask the vendor or engineering owner when it matters.
SWOT strengths and weaknesses concern capabilities under the organization's influence, while opportunities and threats concern the external environment. The club's stable scheduling rules might be a strength, limited staff maintenance time a weakness, a venue willing to pilot an unconfirmed opportunity and a supplier interface change an external risk. Attach evidence, assumption status and impact to every item rather than filling cells with strong brand or big market.
Translate the quadrants into actions. An SO action uses clear rules and a cooperating venue to test swaps. A WO action uses the pilot to establish maintenance ownership. An ST action uses the existing rules to define export and rollback behavior against integration changes. A WT action limits pilot scope and retains a manual route. Give each action an owner, prerequisite, verification and stop condition. These combinations help discussion; they do not require executing all four immediately.
Link feature differences to the decision. Missing exportable change history matters if the club must investigate swap disputes, but its relevance needs a confirmed need. Set weights before scoring, preserve reasons and test sensitivity. If changing one uncertain rating reverses the ranking, report an unstable choice instead of presenting decimal scores as certainty. Do not convert every competitor advantage into a mandatory feature for a custom build.
Scenario three: put competitors in an industry context
Define the market boundary before estimating it. Scheduling software for amateur sports clubs is not the entire human-resources software market. State geography, institution types, size range, observation year and excluded activities. Collect market size, supply, customer groups, channels, substitutes, drivers, institutional constraints and risks within that boundary. A broad industry total is not obtainable revenue.
Use a transparent bottom-up scenario when useful: two hundred eligible organizations paying five hundred units annually would imply one hundred thousand units under those invented assumptions. Real research needs separate evidence for institution counts, willingness to pay and adoption. Distinguish serviceable scope, reach constraints and plausible adopters. Show conservative, central and optimistic assumptions and which variables matter most. A multiplication of scenarios is not a validated forecast.
Avoid false precision when sources disagree. Compare their definitions, dates, currency bases, geography and inclusion rules before combining them. A vendor count may include inactive accounts, a public register may count branches differently, and a survey can overrepresent organizations already interested in software. Describe why one source was preferred or why no single estimate is defensible. Keep a reproducible calculation sheet and label every assumed input.
Map the value chain: club managers, coaches, venues, software suppliers, calendar platforms and messaging providers. Identify who pays, uses, maintains data and bears failure consequences. For direct sales, partners and self-service trials, examine conditions, acquisition effort, support, migration and exit costs. Channel efficiency without evidence remains a hypothesis; a consulting label does not establish analytical quality.
For policy and institutional constraints, record jurisdictional scope, issuing organization, date and questions requiring specialist confirmation. Examples include authority to share coach information, consent for messages and regional differences affecting the workflow. These are research questions rather than legal conclusions. A risk register should distinguish demand uncertainty, data problems, supplier dependency and implementation failure, with observable signals, owners and alternatives.
The original scenario also discusses stock-data comparisons. If company data is genuinely needed, first verify entity, security, exchange, currency, period and data-use rights and assess comparability. Share-price performance is not a proxy for club adoption of scheduling software. A listed parent's financial results cannot automatically establish one product's market share. This example retrieves no securities data and reaches no securities-investment conclusion.
Structure the report so readers can challenge it
A useful report may include executive conditions, purpose and method, competitor facts, pricing and features, task observations, SWOT actions, market boundaries, value chain, drivers, risks and next checks. Combine or split chapters according to the question. Repeating an assertion to reach twelve chapters adds no depth. Important numbers should trace to the source ledger, and conflicting evidence should be explained rather than averaged into a meaningless value.
The executive section should say what is known, what is conditional and what would change the decision. For this fictional club, buying is conditional on verified approval and export behavior, a custom entry point requires a maintenance owner, and retaining spreadsheets remains an option while key needs are unresolved. This is a decision structure for the exercise, not a vendor recommendation or claim that any real service was evaluated.
After factual alignment, Writing Guidelines fits technical-document review of structure, terminology and examples, returning located suggestions. Provide the target text and access to its linked writing handbook, adapting style to the audience. It does not verify market facts, make strategic decisions or automatically produce every file format.
For Word, Excel or PDF, choose suitable export tooling and reopen the result. Check table headers, units, evidence links, pagination and complete text. Conclusions in the summary must agree with the body. Mark trials and retrieval steps not executed. The delivery package contains the report, source ledger, recalculable matrix, assumptions, risks, open questions and review date. A generated PDF is not proof that research or calculations were verified.
Find capabilities without inventing catalog matches
Check whether SkillPI Skill Finder is installed before the target skill; if missing, use its official setup instructions. Search by the current step and inspect version, license, dependencies and permissions. The current SkillPI competitor search returned no matching entries, so this guide associates only the verified data-profile and writing skills at their appropriate steps. The open-source PM Skills project offers competitor-method artifacts as reference without being presented as a verified SkillPI association.
A task may need research access, local processing and document export separately. A skill trained on model knowledge does not become a live collector by receiving a URL, and a user-supplied-data workflow does not imply network capability. Inspect what the selected package actually does. Do not install unrelated tools to fill a six-card layout or assume all packages are free because instructions are public.
Troubleshooting and FAQ
What if old prices appear as current? Require dates and sources for numbers and keep inaccessible current pages unknown. Revisit the source before changing the matrix. Do not silently replace a dated value with a model-generated estimate.
What if the report is long but inconclusive? Return to buy, build or retain the current process. Remove material that does not affect the choice and investigate decisive counterexamples and stop conditions. A shorter report with clear uncertainty can be more useful than a confident but unsupported ranking.
Does a security label prove no data will be uploaded? Inspect actual scripts, external services, permissions and output scope. Static source review is not a runtime guarantee, and installation does not grant additional data access.
Does a framework guarantee depth? No. It organizes questions; evidence quality, hands-on inspection and counterevidence determine whether conclusions stand. When requirements change or catalog skills grow, rerun the same tasks, source fixtures and acceptance rules before refreshing links. Accept the work when facts are traceable, prices comparable, actions executable, market boundaries explicit and limitations visible, rather than because generation was quick or a download count was high.