SpendSignal inside your AI assistant

Get data in and check it is good enough to model.

• get_workspace_status

  Start here. Lists your SpendSignal workspaces, the data in the current one (datasets, connected sources, freshness), recent analysis runs, and which analyses can run now and what each still needs.

• get_upload_templates

  Column specs and sample rows for each file SpendSignal accepts (revenue, channels, ads, keywords, search_terms, hourly), and which analyses need which files. Use it to reshape the user's exports before ingest_rows.

• ingest_rows

  Save rows the user shared (pasted table, small CSV) as one file of a SpendSignal dataset. Call once per file; pass the returned dataset_id to add more files to the same dataset. Max 5000 rows per call; for bigger files use create_upload_link.

• create_upload_link

  Create a one-time link (valid 30 minutes) where the user can drop large CSV/XLSX exports into SpendSignal. Give the link to the user; afterwards call get_workspace_status to find the new dataset.

• connect_source

  Get the link where the user connects Meta Ads, Google Ads, Shopify, BigQuery or Postgres to SpendSignal (credentials are entered on SpendSignal, never in chat). Also shows already-connected sources.

• import_connected_data

  Build a dataset from data synced by connected sources (Meta Ads and Google Ads spend, ads, keywords, search terms and hourly; Shopify revenue) for a date range, ready for both engines. Use a range of 180+ days.

• validate_dataset

  Check whether a dataset is good enough to model: days covered, gaps, spend variation per channel, channels that move together, and which analyses (engine A modules, engine B) can run.

Weekly decisions: verdicts, budget plan, creatives, search, forecast.

• run_cockpit

  Fit engine A on a dataset with the business's unit economics and return the overview: KPIs, per-channel platform ROAS vs iROAS (with band), true ROI, Scale/Hold/Trim/Cut verdicts and top recommendations. Other engine A tools reuse this run.

• get_budget_plan

  Recommend a daily budget split across channels from engine A response curves, optionally changing the total budget. Increases stop where the next unit of spend would lose money (reported as unspent). Shows interactive response curves.

• get_incrementality

  Engine A incrementality: platform-reported ROAS vs modelled iROAS (with band) per channel, platform over-credit, incremental revenue, and baseline vs paid revenue over time. Shows an interactive chart.

• get_forecast

  Engine A revenue forecast (P10/P50/P90) for the next 7–90 days at a planned spend level, split into baseline and paid. Shows an interactive chart.

• review_creatives

  Engine A creative tests: which ads to scale or pause (tested against target CPA, their own past and peers) and which creative tags (hook, format, creator, offer) win. Needs the ads file. Shows an interactive chart.

• review_search

  Engine A Google Search review: keywords to cut or scale, search terms to add as negatives or promote, brand share of spend, and hour-of-day budget shifts. Needs the keywords file (search_terms and hourly optional).

• get_recommendations

  Ranked actions from engine A across channels, campaigns, ads and keywords: what to do, why, which model said so, estimated monthly impact and confidence.

Deep runs on longer or warehouse data, budget-neutral what-ifs.

• run_deep_analysis

  Fit engine B (ridge MMM with counterfactual attribution) on a dataset: model fit, true ROAS per channel, contribution over time, a budget-neutral reallocation and a 30-day uplift forecast. Slower than engine A; good for longer or warehouse data. Shows an interactive chart.

• simulate_reallocation

  Score a budget-neutral what-if with engine B: set spend for some channels (period totals, same units as the data) and see predicted revenue, true ROAS and the re-optimised split. Use get_budget_plan for budget increases or cuts.

Cross-check before big moves; explain any metric.

• compare_engines

  Run both engines on the same dataset and compare per-channel returns: engine A iROAS with band vs engine B true ROAS, rank correlation, and where they agree or disagree. Use before big budget moves. Shows an interactive chart.

• explain

  Plain-language explanation of a SpendSignal metric or method. Topics: iroas, troas, platform_roas, marginal_iroas, breakeven_iroas, true_roi, verdict, saturation, adstock, baseline, road, unspent, test_results, engine_a, engine_b.