Compare vs. competitor

What this compares

Open Compare vs. competitor from the Compare tab under Brand perception in the sidebar. It compares your brand against one active tracked competitor you choose, over one collection channel and window, using only the answers where both were named (api/services/sentiment_compare.py#sentiment_compare_for_project). Sentiment trends shows each brand's own sentiment across every answer that names it; this screen answers a narrower question, how the two of you look when an answer discusses you side by side.

Everything here is a saved-data comparison within one window, not a before-and-after: it sets your brand's numbers beside a competitor's in the same window, from evidence already saved; it never says why the two differ, and it produces no composite "winner" score.

Choosing the comparison

Pick which active competitor to compare against, one collection channel and the same prompt intent, buyer stage and theme used on Sentiment trends. The filter bar sets a 7, 28 or 90 day window ending yesterday (28 by default) and can narrow by engine, tag, country, persona and language; with an engine chosen, the channel list offers only that engine's channels (api/routers/sentiment_compare.py#ACCEPTS, api/services/sentiment_population.py#segment_targets, api/services/sentiment_compare.py#sentiment_compare_for_project). Your own brand is always one side; you cannot compare two competitors against each other. A sub-brand named alongside your own brand, or alongside the competitor's, counts as its family being named, exactly like Sentiment trends rolls sub-brands into their family (api/services/sentiment_compare.py#_side_case).

Coverage: only you, only them, both, neither

Of every analysed answer in the selected filters, segment, channel and window, the coverage bar splits them into four groups: only your brand named, only the competitor named, both named, or neither. Every other figure on this screen is computed over the both group alone, called shared answers throughout. Fewer than 30 shared answers is marked provisional, the same small-sample threshold used across the product (api/services/brand_metrics.py#MIN_OBSERVATIONS).

Named first, and ties

Within shared answers, "named first" is decided by whichever family's earliest mention appears first in the answer; equal earliest positions are a tie, counted on their own rather than credited to either side. A tie is not a share you can win by controlling the other numbers: it is simply how often the two of you land in the same spot (api/services/sentiment_compare.py#sentiment_compare_for_project).

Sentiment, roles and reasons on shared answers

For each side, the screen shows the positive/neutral/negative/unclassified mix and the recommendation-role mix of that family's own mentions, computed only within shared answers, not across every answer that names it (api/services/sentiment_compare.py#_build_side). With the owner's Brand reasons entitlement, it adds each side's strength and weakness rate per reason and per category, over the shared answers that were analysed for reasons; without that entitlement, this part of the screen shows a locked message instead of zeros (api/schemas/sentiment_compare.py#reasons_locked). Reason-analysed shared answers are a subset of shared answers, not a separate population: an answer collected before reason extraction shipped has no reasons to draw on, so it counts as not-yet-analysed instead, and the two together, reason analysed and not-yet-analysed, always add back up to the full shared-answer count (api/schemas/sentiment_compare.py#reason_analysed_shared, #reason_unanalysed_shared). This part of the screen carries its own provisional flag, fewer than 30 reason-analysed shared answers, separate from the shared-answers provisional flag above (api/schemas/sentiment_compare.py#reason_provisional).

Per-prompt table and paired evidence

The table is keyed on prompt target, not prompt: one row per (prompt, country, audience, language) combination with at least one shared answer, 25 per page, sorted by the most shared answers first, so a prompt tracked in several countries, run as several personas, or tracked in more than one language can appear more than once, each occurrence with its own shared-answer count, each side's positive share, and which side was named first more often; the row itself does not name which language it is (api/services/sentiment_compare.py#PROMPT_PAGE_SIZE, api/schemas/sentiment_compare.py#ComparePromptRow). Selecting one row opens paired evidence for that specific prompt target, up to 25 shared answers per page, newest first, each showing both sides' saved quote (cut to 500 characters), sentiment and role from that same answer (api/services/sentiment_compare.py#EVIDENCE_PAGE_SIZE).

Access

Every current project member can read this screen and its evidence on every plan, the same as Sentiment trends. The reason and category rows need the project owner's Brand reasons entitlement; everything else here does not.

Related

• Sentiment trends: each brand's own sentiment across every answer that names it, not only shared ones
• Brand reasons: the reason categories and per-reason rates this screen locks without that entitlement
• Brands and sub-brands: how a sub-brand's mentions roll up into its family for this comparison
• Personas: the audiences the persona chip filters by, here as on Sentiment trends and Brand reasons
• Competitors: which brands are eligible to compare against