01 · The score
Share of Voice
The big number at the top is HP's slice of every tracked citation, HP's plus every tracked competitor's, blended across five AI engines. The five smaller tiles are the same math, one engine at a time.
Overall SOV
34%
▲ vs last week
Share of VoiceHP citations divided by all tracked citations for that query set. A competitive ratio, not a raw count.
PresenceThe smaller number under each tile: how often HP was cited or named at all, win or lose against competitors.
The delta arrowChange versus a run at least a week old, never a same-day rerun, so it reflects real movement.
Each tile also carries a small gray trend line, the last seven weekly runs. Click any tile and that number's full history opens directly beneath the row. Section 06 covers what you'll see there.
06 · Change over time
Reading the dashboard across runs
The dashboard re-measures every Monday. Everything above describes one run at a time; this section covers the views that line runs up side by side, and how to get back.
Way in, one number
Perplexity 21% ▼
Click any of the six score tiles. That number's history opens right under the tile row, with the same chart the full view uses. Click the tile again, or the Close button, and it folds away. Click a different tile and it swaps. Nothing else on the page changes.
Way in, everything
NowOver time
Use the switch in the header, or the “Everything over time” link at the right of the Scope row (also inside any expanded tile). The whole page changes to the trend view.
To get back, click Now. Your Scope choice carries across; nothing is lost either way.
Which runs count. Trend lines use cadence runs only: the weekly Monday runs, the monthly and quarterly sweeps, and the first full baseline. Manual reruns (a Run Now click, a verification run) stay in the archive but are left out, so a same-day rerun never reads as a trend. The page says how many runs each line is drawn from.
What the Over time view shows, top to bottom
Share of voice over timeOne line, in whichever Scope is selected. A hollow dot marks the Aug 12 baseline. The pill beside the controls states the typical week-to-week move for this line, measured from the runs themselves.
Numbered markersDashed vertical lines with a number: entries from the actions log. A circle is something the team did; a square is a change to how the dashboard measures, and numbers on either side of a square are not comparable. A dashed marker tagged detected was found automatically by the dashboard run from the client's own feeds and filings (press releases, investor relations, SEC filings) and has not been reviewed yet; treat it as provisional until someone confirms or removes it.
By engine, by categoryThe same line, one small chart per engine and per category, on a shared scale so shapes compare honestly. Each carries its current value and its change across the selected range.
What moved, query by queryQueries that changed in the range: how many engines HP is present on now, the change in engines since the start of the range, and a run-by-run strip. Click a row to open that query in the Now view.
Actions logEvery logged action with a before-and-after readout. Detail below.
Range4 weeks, 8 weeks, or all runs. Narrows every chart and the movers table at once.
The strips
A strip is one cell per measured run, oldest on the left. Two places use it, with two different colorings.
In the movers table darker means HP was present on more of the five engines that run; gray is absent everywhere; a dashed cell means the query wasn't measured that run.
In a query's History block one row per engine: dark for named and cited, mid for cited only, light for named only, gray for absent. The text beside it, “present on 3 of 7 runs,” is the number to carry, not any single cell.
Extended and Probe queries only re-run on the monthly and quarterly sweeps, so their strips are mostly dashed. That's expected, not missing data.
The actions log
An action is anything the team did that might move a query: a placement that ran, a page that went live, a site change. Each entry names its date, what it was aimed at (queries, a category, or an engine), and gets a readout.
2
Work Relationship Index feature story · Sept 2 · earned
1.0 of 5 engines before (Sept 1)→2.0 after (Sept 7), same queries
The targeted queries gained about 1 engine on average after this date. For scale, roughly 11% of engine results flip in a normal week with no action, so a gain that holds across the next two runs is worth taking seriously; a gain on one run is not.
The readout compares the last weekly run before the date with the first run at least three days after it, counting only queries measured in both, and puts the normal week-to-week noise next to the change. It deliberately never says “caused.” Read it as “here is what the numbers did around this date, and here is how big a move would have to be to mean anything.”
Logging an action. Click “Log an action” at the top of the actions log. It gives you a short template: date, type, title, outlet, a one-line summary, and the queries or category it was meant to move. Fill it in and send it to Pete; it shows up on the dashboard with the next update, with its before-and-after computed automatically. Naming the target queries is the part that matters most, since that is what the readout is measured against.
Reading it honestly. These answers are stochastic. About one in nine engine results flips from one weekly run to the next with nobody doing anything, and the overall share of voice line moves about half a point on its own. A single query flipping on a single engine for a single run is noise. A change that holds for two or more runs, across more than one engine, is a signal. The page shows both yardsticks so you don't have to remember them.