Venbit

List Websites

List the websites this account can read, with their timezone. Call this first: every other tool needs a `site` from here, and a site not in this list cannot be queried.

Describe Metrics

The vocabulary: every metric and dimension query_metrics accepts, each with a plain-English definition and its caveats. Call this before composing a query rather than guessing at names.

Query Metrics

Traffic and conversion numbers for a window, optionally broken down by one dimension. Exact for the whole window at any size: settled days come from a nightly archive reconciled against the raw records, recent days come from live data, and the `source` block in the response says which is which. Rates are percentages. `visitors_exact: false` means the visitor count is an UPPER BOUND (daily unique counts cannot be summed), so say so if you quote it. `rows_truncated: true` means the breakdown list

Ad Report

EVERY PAID CAMPAIGN, with what it actually produced, and the baseline to judge it against. Use this for any question about advertising rather than assembling one from query_metrics. Each row is one platform and campaign with visits, distinct people, enquiries, conversion rate, engagement and form progress. Campaigns are identified from the CLICK ID where possible (gclid, fbclid, msclkid), so auto-tagged campaigns that carry no utm parameters are included, which is most of them. `summary` carries

Campaign Report

Hand-tagged links broken out by the full (utm_source, utm_medium, utm_campaign) tuple, with visits and enquiries. Use this when you need to tell apart two campaigns that share a name on different platforms, or to audit how a client tags their links; query_metrics with dimension=campaign would merge them. COVERS TAGGED TRAFFIC ONLY: anything arriving without a utm_source is absent, so this is not a picture of all advertising. For that use ad_report.

Find Friction

Where visitors struggled: rage clicks, dead clicks, form errors, JavaScript errors and not-founds. RANKED BY HOW MANY DIFFERENT PEOPLE hit each problem, not by how often it fired, because one furious person clicking forty times is one problem and a dead link fifty people met once is fifty. Weigh a finding by `share_of_visits`. `events_per_person` near 1.0 means many separate people met the same broken thing; a high value is one person repeating themselves.

Top Pages

Every page anybody viewed in the window, with its title, view count, and how many separate visits reached it. This counts ALL pageviews, unlike query_metrics with dimension=path which counts only the page a visit landed on. `views_per_visit` well above 1 means people returned to that page within a single visit.

Search Rankings

Where the site sits in Google for the keywords it tracks: current position, what moved, what is close, and the visibility score. Use this for any search-ranking question rather than guessing or reasoning from traffic. FOUR THINGS YOU MUST NOT GET WRONG. (1) A POSITION IS ONE READING FROM ONE PLACE, taken on a date, from the location and device named on the row. Google personalises results, so what any given visitor sees can differ; never present a position as what everyone sees. (2) A KEYWORD NO

List Goals

The goals the site owner configured themselves, with completions and dollar value for the window. Use these rather than inventing your own definition of success: they are what this business decided counts as a win. THREE THINGS TO READ CAREFULLY. `completions` counts EVENTS, not visits, so one person reloading a thank-you page three times is three completions and dividing completions by visits does NOT give a conversion rate, so use conversion_rate from query_metrics for that. `value` is complet

Visitor Journeys

What INDIVIDUAL people actually did, step by step: landed here, read to 60%, opened the form, gave up at the phone field, left. Aggregates are averages of everybody and the interesting thing is rarely average; a handful of real journeys shows a pattern no chart does. Pick the population with `kind`. Visitors' typed form values are never returned — field LABELS are, so 'gave up at Phone number' is visible, but what they typed into it is not.

Site Structure

What the website actually IS, read from its own HTML: the enquiry forms and which page each sits on, the phone numbers and email addresses it publishes, the pages found, any booking tool or app-store links. NO OTHER ANALYTICS PRODUCT HAS THIS — they only see behaviour. Combine it with the behaviour tools: if /contact carries a form and people abandon it, you can ask whether the site also publishes a phone number they might be calling instead, and then check whether phone taps rose. Use it to gro

Site Speed

Real-user load speed: the 75th percentile Largest Contentful Paint in milliseconds, the same measure Google's Core Web Vitals uses, plus the median for context. Under 2500ms is 'good' by Google's threshold. Readings above one minute are discarded as instrument error (a backgrounded browser tab keeps counting). Returns null when the site has too few measurements.