Shipped: answer-block tagging for win/loss cross-reference
A small change with a downstream payoff. Every answer block now carries tags for buyer, sector, theme, and outcome — wiring the foundation for the win/loss cross-reference work next month.
Shipped this week: answer blocks now carry structured tags. The tags are buyer, sector, win-theme references, and bid outcome. The change is small in surface area and disproportionately useful downstream.
What changed
Before this change, an answer block had:
- A title and body.
- A version history.
- An owner.
- A freshness score.
- A list of citations to source documents.
After this change, every block also carries:
- Buyer tag — the buyer this block was last drafted for, plus a list of all buyers it has been used in.
- Sector tag — the buyer’s industry, normalized.
- Win-theme references — the IDs of the win themes from the bid this block contributed to.
- Outcome tag —
won,lost,pending, orwithdrawnfor each bid the block participated in.
The tags are added at the moment the block is committed to a bid response. They are updated when the bid outcome is recorded.
Why it matters
Three things this enables, two now and one next month.
Now: better retrieval. A query for an answer to a similar question can be filtered by sector — “show me past answers for healthcare buyers” — without retraining or re-embedding. The retriever already knows which blocks have been used where.
Now: theme reuse audit. When a win theme is being constructed for a new bid (see the win-themes field guide on theme construction from capture), the proposal manager can ask “what blocks did we use last time we ran this theme, and how did they score?” The tags make the question answerable.
Next month: win/loss cross-reference. The Win-Loss Intelligence module depends on knowing which blocks contributed to which outcomes. With the tags in place, when a bid is closed as won or lost, the block-level outcome data is available without a backfill. This is the input layer for the section-level outcome correlation the platform page describes, and it unblocks the post-mortem write-back work the eight-stage pipeline post covers.
What this is not
It is not a win-rate report. The tags are the data layer; the analytics layer is separate. We are deliberately shipping the foundation before we ship the surface, so the analytics team has a year of structured outcome data to work with by the time the win/loss UI lands. The build-log discipline is to ship the inputs to a feature six months before the feature itself, so the data is there when the feature ships.
Docs: see the answer-block schema doc on the platform site.