Opening proof path
SIRJ is loading the workspace, connector, or public page with the same proof path intact: ad spend, lead, booked job, invoice, paid.
See how SIRJ Ask Data answers revenue questions with scoped query specs, cited metrics, chart rows, cache rules, and safe unsupported responses.
Buyer answers
Buyers should understand which questions Ask Data can answer, how scope is protected, what citations look like, and where SIRJ refuses to answer without enough support.
Ask Data can answer spend, revenue, ROAS, top campaign revenue, campaign budget recommendations, cost per lead, source performance, spend versus revenue, and uncontacted-lead questions when the dashboard data supports them.
The buyer sees a focused revenue assistant, not a broad chatbot. The supported question set is deliberately tied to the dashboard and source records.
Follow the evidenceEach query spec carries tenant and client scope, then dashboard config, events, connections, cache records, and execution results are checked against that scope before an answer can be returned.
Agency teams can ask client-specific questions without blending records across workspaces or turning one client's revenue into another client's answer.
Follow the evidenceA supported answer includes status, intent, answer text, cited metrics, query context, and when useful, rows or a chart contract with fields such as campaign, attributed revenue, spend, leads, ROAS, or uncontacted lead rows.
The answer gives the team a claim and the proof path at the same time, so follow-up questions can move into source context instead of opinion.
Follow the evidenceUnsupported questions return a safe response that lists the supported Ask Data areas and avoids reading dashboard data, caching a made-up result, or presenting a confident answer without evidence.
The refusal is part of the product value. Buyers can trust the answers more because the product is willing to say what it cannot answer yet.
Follow the evidenceSupported questions
SIRJ answers the questions agencies ask when a client wants the source path behind spend, revenue, campaigns, leads, and follow-up.
Question map
The examples are intentionally practical. Each answer path points to the dashboard fields, normalized events, citations, or chart rows that make the response defensible.
| Question type | Examples | Answer path | Visible proof |
|---|---|---|---|
| Performance totals | What is ROAS? How much ad spend did we have? What revenue was attributed? | Dashboard metric query specs read allowed dashboard cards and cite fields such as totalSpendCents, attributedRevenueCents, and roas. | The user receives a cited metric rather than a paragraph that hides which dashboard card supplied the number. |
| Campaign decisions | Which campaign drove the most revenue? Which campaign should we increase budget on? | Top-row and computed query specs sort campaign performance rows or evaluate campaigns with spend, leads, revenue, and ROAS. | The response can include a campaign chart, ranked rows, cited campaign metrics, and the calculation basis for budget recommendations. |
| Source-specific questions | How did Google Ads perform last month? What was Facebook Ads cost per lead? | Source filters identify known ad sources, date parsing resolves month ranges, and normalized events supply spend, lead, click, impression, and revenue context. | The answer names the source filter, date range, cited metrics, and chart rows used to explain source performance. |
| Operational follow-up | Which leads have not been contacted? How many leads need follow-up? | Uncontacted-lead queries require lead data and a communication source, compare contact identity after lead time, and return bounded actionable rows. | The product can show a count plus returned lead rows while refusing the question when communication evidence is unavailable. |
Answer workflow
Ask Data answers quickly in the interface while preserving the trust path underneath: parse, scope, dashboard data, allowed execution, and safe cache behavior.
| Step | What happens | User-visible proof |
|---|---|---|
| Parse the question into intent | Ask Data starts by identifying the supported intent, optional date range, and optional source filter from the question or from an injected parser client constrained to allowed intents and dashboard cards. | Unsupported questions stop early with a safe answer, while supported questions move forward as explicit query specs. |
| Build a scoped query spec | The query spec records account id, client id, query type, intent, allowed card or computed metric, sort direction, limit, date range, and source filter. | The user-visible answer can be traced to a bounded query contract instead of an open-ended prompt. |
| Read the current dashboard dataset | Runtime Ask Data builds the dashboard dataset for the active request, including dashboard config, normalized events, and connection state within the selected date range. | Answers reflect the same dashboard context the user is reviewing, so the assistant and the report do not tell different stories. |
| Execute against allowed fields | Metric and top-row queries read only allowed dashboard fields, while computed queries run explicit functions for CPL, source performance, spend versus revenue, budget recommendation, and uncontacted leads. | The response includes cited fields, rows, or charts that match the query type rather than free-form generated claims. |
| Cache only safe answers | Cache keys include scope, question, query spec, date range, and dashboard snapshot time, while row-level uncontacted-lead answers are excluded from cache. | Repeated metric questions can stay fast without replaying sensitive row-level follow-up lists from cache. |
Guardrails
The product earns trust by narrowing the answer surface. A question must pass field, scope, evidence, and unsupported-answer gates before it becomes a client-facing claim.
Ask Data validates dashboard card fields and top-row sort fields against an allowlist before a query spec can be created.
Query specs, dashboard configs, events, connections, and cache records must match the active account and client scope.
Computed queries return insufficient-data answers when required spend, lead, communication, campaign, or revenue evidence is missing.
Unsupported questions return a constrained answer listing what Ask Data can answer, without reading the dashboard or saving a fake cached result.
FAQ
Ask Data has to make restraint feel like a feature. These answers explain where it helps and where connected evidence still decides.
No. Ask Data is constrained to supported revenue questions, dashboard fields, computed metrics, source filters, date ranges, citations, and safe unsupported responses.
Yes, when the active account and client scope has the supporting dashboard data. Scope checks prevent one client's records from being used in another client's answer.
Yes. Top campaign, budget recommendation, source performance, and spend-versus-revenue answers can return a bar-chart contract with encoded fields and supporting rows.
The answer should return insufficient data or unsupported status rather than pretending the dashboard can prove something the connected sources do not support.
Ready when the numbers need to hold up
Open the demo to see Ask Data beside the dashboard, or review data routing to understand how the answer stays tied to source evidence.