CVRGWizard

Sign in

Your email address, then the six-digit code we send it - no password.

Access is by invitation. If you can't get in, ask the person who invited you, or the support team.
CVRGWizard
Dashboard Import coverage Tags Questions Settings People New
Viewing as
Coverage Wizard
Every board sealed from every other
Live data - nothing baked in
CVRGWizard

Coverage Wizard

Coverage highlights

Top 10 by impact, past 30 days
A click opens the item. Each card is rebuilt from the piece itself - the outlet's mark, the section, the headline, the journalist, the opening lines and the lead image the publisher declared for sharing, credited and linked to the source; a print or broadcast piece shows the opening of its transcript instead. Images are withdrawn on a publisher's request.
Latest
Reach lens
Click a start date, then an end dateClear
Weights preview active - scores and bands on screen are recomputed with your draft profile from Settings. Nothing is saved.

By month

A (75+)B (50-74)C (<50)unscored

Quality - where the scores sit

Impact - where the scores sit

Quality by month - average placement score, hollow point: fewer than five items

Impact by month - the impact index, hollow point: fewer than five items

Media mix

Audience mix

B2C v B2B - the title decides

Focus on the brand (share of placements)

Sentiment toward the brand

Top Outlets - click a metric to sort, click it again to flip

Top Journalists - click a metric to sort, click it again to flip

Top Social Accounts - who posts about the brand, click a metric to sort

Social posts - the newest three in view

Coverage

Export this view:
Methodology ▼

Import coverage - this board

Drop a coverage export (Excel, CSV or PDF) or paste article links. The file is read in your browser; nothing uploads until you build the preview. Imported items run through the SAME pipeline as emailed coverage: relevance gate, full-text fetch, AI enrichment, scoring - allow a few minutes after commit.

Drop a file here or click to choose
.xlsx · .csv · .pdf

Column mapping

The engine proposed this mapping from your file's headers. Adjust anything it got wrong, then refresh the preview. Unmapped fields are simply left blank.

Preview - nothing is saved yet

StatusDateOutletHeadlineTypeLink

Duplicates already in the archive are skipped. Items matching an existing story at a different outlet join that story's syndication family. Junk rows are reported, never silently dropped. Every import can be undone in one click - nothing is ever hard-deleted.

Import result

News scans - find coverage the monitor missed

Saved scans query Google News on a schedule from the pipeline - one edition per market selected under Settings > News scans & markets - the same searches the team runs by hand, automated. New finds are held for the primary feed to log first (48h default), then true gaps land in the Review tab for a human. Nothing enters coverage without a click.

-

Recent imports

-

Campaign tags - workspace

A tag is a brief the AI reads, not a keyword. Every new item is judged against the open tags during the read it already gets; creating or editing a tag re-scans existing coverage in the background (progress shows on the card). Matches at 70% or above apply themselves, 40-69% wait below as suggestions, a human decision always wins. A tag on one copy covers its whole syndication family. Filter the dashboard by a tag and Export this view is the round-up.

New tag- the brief is what the AI reads; be specific about what the campaign said and when
A label that sets expectations (a campaign or launch usually has a window, an issue or theme usually runs open). Nothing else changes.
Good briefs name the offer, the product and the hook. "Cruise deals" is too vague to judge; the estimate below turns amber while the brief is under 60 characters.
tooptional
Coverage outside the window is never auto-tagged; a person can still add it by hand in the drawer or with Tag items. Leave blank for an open-ended issue or theme.
A pre-filter, not the rule: the archive scan reads only items carrying a hint (headline, summary or stored text), and the AI treats a hint as strong evidence. Near-perfect precision for a sale with a distinctive name.
Audited (tag.create); the scan starts on the worker's next cycle - about five minutes - and progress shows on the card.

Suggested - needs a look

Keys: A confirm the first, X not this campaign - the review queue rhythm

Matches between 40% and 69%. A suggestion does not filter or count until someone confirms it. Confirm and dismiss are audited; confirming applies the tag to the item and its syndicated copies; a dismissed suggestion is never re-suggested.

Questions from the AI

Loading the questions...

This boardname, brand words, colour, logo SourcesStreem key, look-back, scan Excluded coveragerules, own posts, guidance Known ad linesradio, TV and podcast ad copy AI score tuninga few quick questions each week House defaultsthe scoring every board starts from Modulesextra measurements, board by board Reading and accuracyread in full, the AI against people Scoring modelweights, bands, profile Target audienceswho the brand wants to reach Sentiment & impact6-rung ladder, hero, watchlist Reach & audiencelens, exposure, impact curve AVErate card, CPMs, currency News scans & marketsAU/NZ gate, grace, pause Access & systemwho can see this, health Outlet profileswho each outlet reaches (reference)

Scoring model

How many placement points each component can earn. Seven separate scores; the maximums must add to 100.

Saving writes a new profile version; every existing score keeps the profile it was scored under. Re-scoring all coverage is a separate step that asks for a second click, keeps full history and logs one audit entry.

How the score works, and the profile history
Placement score = the sum of the components above, 0 to 100, banded A / B / C on the thresholds shown. It measures how big and well-placed the coverage is; direction lives on the sentiment axis (next section), so a well-executed negative story scores well here on purpose. Sentiment carried 10 placement points until V4.0 (20 Aug 2026), when the two-axis model moved them onto the other components.

Target audiences

Who the brand wants to reach. Every placement earns audience-fit points for how strongly its outlet's audience indexes against these.

Up to three. "Core" audiences earn full points; "also valuable" ones up to 70%.

Saving is audited and takes effect for newly scored coverage at once; existing scores move when you rescore under the scoring profile (Scoring model section).

How audience fit is scored
For each placement we take its outlet's audience profile and ask how strongly it indexes against each target: the share of the outlet's audience inside the target (say 45% aged 55+) divided by the population share (34%) gives an index of 133 - the same composition index media planners use, 100 = mirrors the population. Age, affluence and gender indices combine by geometric mean. Index 60 or below earns nothing, 140 or above earns everything, linear between (tunable under Scoring model). A placement is scored on its best-matching audience; "also valuable" audiences are worth 70%. A B2B target is categorical on the outlet's audience type - a B2B (trade) title matches, a B2C title does not - and, when the target names sectors (V4.19), on sector overlap: a travel-trade title earns nothing for a technology brand's B2B audience. Profile confidence (grade A currency data 1.0 down to grade E nothing known 0.3) shrinks the points toward a media-type prior, so a guess can never swing the score the way a measured figure can. No targets defined means everyone counts and every outlet, B2C or B2B, earns full points.

Sentiment & impact

The direction axis. Each rung multiplies the placement score into brand impact.

Hero pieces

Band A placement and clearly favourable sentiment (glowing or positive). Counted in the IMPACT tile's hover.

Watchlist

any placement this bad or worse is flagged

Rung factors change only with a new scoring profile, so every historical impact figure stays reproducible.

Where the six rungs came from
Validated against 45 hand-rated items (Aug 2026). "Critical with balance" requires the brand's own side to be carried; guilt-by-association pieces are negative, not damaging. Extreme rungs need full text - a metadata-only judgment is clamped to the mid rung on its side. Editors can re-rate any item from its drawer; the AI never overwrites a human call and every re-rating is audited.

Reach & audience

How estimated views are counted, and how audience size weights impact.

Reach lens - click to switch this session; save to make it the default

Conservative exposure factors - % of the audited audience counted as actual views

Audience impact curve - how much a big audience amplifies impact

Methodology and evidence (V2.9)
Lenses. Conservative = audited potential audience cut by one evidence-based exposure factor per medium (estimated actual views; CoverageBook-style). Standard = the per-item potential audience from the audited currencies (OzTAM / Regional TAM, GfK, EMMA / Roy Morgan, Nielsen / Ipsos) - the basis Streem reports; back-tested against half-year totals reported by a monitoring vendor, on the same items (a 100.2% match). Stretch = the outlet's total audience per placement (Meltwater-style); falls back to Standard where an outlet is not yet curated. Exposure factors. Online 2.5% (industry actual-view constant, validated against measured readership, back-test 12 Aug 2026); print and magazine 40% (issue noting); TV and radio 70% (presence research); podcast 90% (deliberate listening). No prominence ladder in reach - prominence is scored in quality, never double-counted. Impact curve (V4.2). Audience-weighted impact (the product beneath the IMPACT index since V4.25.3) = brand impact x a log-scaled audience weight between the two ends of the market (floor at the smallest per-item figure, cap at the largest), per lens because the three bases live on different scales. Calibrated once per client; a US or China rollout calibrates to that market's ends. Do not chase it - drifting anchors break trend lines.

AVE

Media cost equivalent, served transparently and labelled indicative - never the headline metric.

Rate card - click a card to make it the active calculation; the draft card is for comparison only

Why our AVE differs from Streem's - the three cards by medium on today's coverage

Display currency

calculated in AUD, converted daily for display
Formula, evidence and the AMEC position
AVE = (audited potential audience / 1,000) x the active CPM card, on the Standard basis regardless of the reach lens, with no editorial multiplier (no published vendor formula uses one; controlled studies put the true multiplier at ~1.0). Market card (default): evidence-based Australian transacted rates per 1,000 audience reached, each value sourced (V2.9 methodology), refreshed quarterly. Streem-parity card: the V2.6 calibration retained for continuity - calibration outputs that make totals track the figures a monitoring vendor reported, not market prices (V4.18 restored the x1.5 factor the V2.9 multiplier removal had dropped; the card had been reporting about 69% of Streem). Space-equivalent card (draft, V4.18): market rates x a space factor by focus tier (dedicated 1.0, featured 0.6, included 0.35, passing 0.15) - how advertising space is priced; shown for comparison only - the headline figure stays the audited-audience AVE. Back-test (Aug 2026): on a half-year population the market card lands at 80-82% of the AVE a monitoring vendor reported, with zero calibration - print is most of the gap because size is not priced. AMEC / Barcelona Principles 4.0: AVE-style measures are served for those who ask, labelled indicative, never the headline metric - outcomes are led with.

News scans & markets

Where the platform looks for coverage Streem missed, and which markets count.

Markets that count as coverage - each country adds its Google News edition to the scans and its domain rules to the gate

A find must place itself in one of these markets - by its country domain (.au, .nz, .co.uk ...), a country subdomain or edition path (au.investing.com, msn.com/en-au, theguardian.com/uk-news), or the outlet registry - or it is set aside as outside market. Set-aside finds are restorable, never deleted. Google News publishes country editions, not a free-text country search: an edition biases results toward that country's outlets but does not filter, which is what the gate is for. American outlets rarely use .us, so a United States market admits finds mainly through the registry and your allow list.

one per line or comma-separated - counts as the first market above
junk aggregators, wrong-company sites
Loading set-aside finds...

Timing and safety

hours before a find is shown to you (Streem gets first claim)
pause stops every scan; nothing is lost
Scan queries themselves live on the Import coverage page, next to the finds they produce.

Access & system

Who can see this, and whether everything behind it is healthy.

Access

Everyone signs in with a six-digit email code - no passwords. The API verifies identity (a signed token) and then the membership on every request: which boards this person holds and as what role (Viewer reads, User works the coverage, Editor runs the board including Settings and People; Org Admin and Super Admin are the support team). No membership, no board - even for the support team. Board memberships are managed in People.

System

Brand web domains

The affiliate scan counts a link only when its unwrapped destination is on one of these domains. One per line. Sister brands in the exclusions are never mistaken for the brand. Saved per board, audited.

Data integrity checking...

Every morning about 6:30, and whenever you press Run now, the platform asks 75 standing questions of its own data (58 rules, 17 counts): one gate row and one current score per item, an audience on every reportable item, copies matching their originals, a stronger read never overwritten by a weaker one, human calls still standing, tags inside their thresholds, imports intact, the worker alive, every board sealed from every other, every membership inside its rules, no platform key anywhere but the vault, no Streem connection failing unnoticed and no look-back stopped for good. Green means the data is consistent with itself. It does not mean a judgment is right - what the sweep cannot see is listed at the bottom, on purpose. Every bug found becomes a new rule.

What this does not check

    Outlet profiles

    Who each outlet reaches - one profile per outlet, drafted from published sources, verified by you. Shared across every board. Reference data: folded away unless you need it.

    Show the outlet profiles
    Media type:

    Grades: A currency data (Roy Morgan, Ipsos iris, GfK, OzTAM, Nielsen) - B publisher kit citing a currency - C third-party panel or a quantitative statement - D positioning or estimate - E nothing known (media-type prior). Profiles expire 12 months after verification.

    People on this board -

    PersonRoleScopeSinceAdded by

    Add a person

    Access starts at once for someone who already signs in. Someone new gets a one-time link that lasts seven days - shown here once for you to send (the platform sends no email in Stage 1). Nobody is ever deleted: removing a person revokes their access and keeps the record.

    Invitations

    Loading...

    Set up a new board

    About 7 minutes
    A approveX archiveS skipU undoO open articleT article viewE adjust score←/→ moveEsc close
    Coverage Wizard (working title) - proprietary platform. Every score shows its working; every AI judgment stores its rationale and is human-overridable.