Features
The capabilities, not the marketing version.
Five surfaces, one data lake, one reasoning engine. Below is what Combot does today — not what's on a roadmap slide.
Surface 01
Continuous monitoring.
Pipelines run on 15-min, hourly, and daily cadences. Source data lands in BigQuery, partitioned and clustered, with a per-pipeline observability record.
GA4 · Rybbit · GTM
Live API for the freshest 24h, BigQuery export for everything else. GTM container audits flag broken or duplicated tags.
Search Console · CrUX · PSI
GSC live + bulk export. URL-level Core Web Vitals from CrUX. PageSpeed Insights for lab Lighthouse. SERP via HasData for live position tracking.
Google Ads · Merchant Center
DTS for historical, API for live. Feed status, price freshness, Shopping Pack appearances tracked nightly.
WooCommerce · HubSpot
Orders, inventory, contacts, deals — joined to attribution signals from the web analytics stack.
Cloudflare · nginx
Edge analytics, DNS state, bot identification from nginx logs. Distinguishes human, search, and AI bot traffic.
Airtable · Sheets · Linear
Editorial calendars, action queues, findings tracking. Combot reads from and writes to all three.
Surface 02
Anomaly + correlation.
Each metric is checked against multiple baselines: rolling week-over-week, year-over-year, learned seasonality. The interesting signals are compound — a spend spike that doesn't match the click curve and lines up with a regressed Core Web Vitals release.
Time-series detection
Z-score against rolling baselines plus a learned seasonality model. Threshold tuned per-metric, not one-size-fits-all.
Compound flagging
When ranks drop and CWV regressed and Ads spend held flat — that's one story, not three. Combot stitches it together.
Trend extrapolation
7-day, 30-day, and 90-day forecasts for revenue-adjacent metrics with confidence bands. Surface what's drifting before it crashes.
Surface 03
AI visibility module.
The newest pillar. A dedicated 7-layer model of how brands appear in language-model answers, with a north-star AI Share of Recommendation metric, nightly prompt runs across Claude / GPT / Gemini / Perplexity, citation extraction, fact accuracy scoring, and competitor benchmarking.
Surface 04
Chat-native command.
Every Combot capability is invokable from chat. The bot lives in the channels your team already uses — Slack, Teams, Mattermost, or any platform connected via API integration. The answers come back where the conversation is. No new dashboard URL to bookmark, no new login to remember.
Conversational queries
@combot weekly anomaly report, @combot why did organic drop on /pricing, @combot SOV for elisa vs dna on 5g queries. Plain English, real answers.
Digests + alerts
Weekly digest to a channel. Real-time alerts when a metric crosses a threshold or an anomaly fires. Routing per team, per client, per severity.
Channel-scoped access
Tenant isolation extends into chat: which channel sees which client's data is configured per workspace. Cross-client leakage prevented at the bot layer.
Surface 05
Audits as code.
Combot's compound audits run on demand or on schedule. Each is a tool the Claude agent can call; results are streamed back as structured findings, ranked by severity and confidence.
SEO audit
Indexing state, ranking trends, CWV per template, schema coverage, content gaps, sitemap health. Compound — not 10 silos.
Ads audit
Budget pacing, Quality Score by ad group, waste segments (zero-conv impressions), ROAS trend per campaign.
AI bot audit
Bot identification from nginx logs. Which AI crawlers reach your pages. Which never see your content because they can't render JS.
Traffic audit
GA4 + Rybbit + nginx joined to surface session-level anomalies that single-source dashboards miss.