A Voicify "Competitor A or Competitor B" dental AI battlecard that should have been retired six weeks ago is the single most expensive document in a competitive program. The rep still has it pinned in the sales drive. The opening line still reads true enough at a glance. The buyer on the discovery call has spent more time on the competitor's website than the rep has spent on the card, and they catch the one claim that stopped being true two months ago when the competitor shipped a feature that nobody on the competitive intel team noticed. The card loses the deal — and then it loses the next three deals before anyone marks it stale. Kill criteria are the discipline that turns retirement from a judgment call into a rule, and they are the most under-built layer of most competitive programs.
TL;DR
Kill criteria are the five explicit triggers that retire a Voicify A-or-B battlecard before it loses another deal. Stale cards are more dangerous than missing cards because reps cite them confidently and buyers catch the drift. Five triggers fire a kill: evidence library decay above thirty percent, a competitor's category-redefining feature, a ten-point win-rate drop over two quarters, a regulatory or payer shift, or a customer reference that goes wrong. Any single trigger queues review; any two together fast-track the sunset. The workflow is freeze, notify, debrief open deals, archive. The archive is permanent, searchable, and excluded from the rep-facing index. Build the kill policy as a system that runs, not a meeting that happens.
Why Kill Criteria Are the Hygiene Layer Nobody Builds
Competitive intel teams build battlecards. They refresh battlecards. They occasionally argue about which battlecard to use on which deal. What they almost never build is the policy that retires a battlecard before it goes wrong in the field. The reason is structural. A battlecard is a creation, and creations are easy to author. A kill is a deletion, and deletions are politically harder — somebody wrote that card, somebody won deals with it, and pulling it back feels like an admission. So the card lives on, drifts, and quietly bleeds win rate while the program manager pours new cards on top of it.
The fix is to treat kill criteria as a policy that runs without a meeting. Five triggers, defined in advance, visible on the battlecard analytics dashboard, and reviewed on the quarterly cadence with authority to invoke between cadences when a trigger fires sharply. The card retires on data. The author of the card is not the same person who decides the kill. The rep who used the card last week gets a notification the same week the card goes dark. The discipline is procedural, not personal.
The Five Triggers That Kill a Card
Five triggers carry the policy. Each one is observable on a dashboard so the kill conversation is grounded in data, not opinion. Each one names the owner of the data feed so the trigger does not depend on someone remembering to check.
| Trigger | Threshold | Data owner |
|---|---|---|
| Evidence decay | More than 30% of evidence-library rows expired or weak | Competitive intel research lead |
| Competitor category move | Competitor ships feature that invalidates central claim | Product marketing + field intel capture |
| Win-rate drop | 10-point drop on card-cited deals over 2 consecutive quarters | Sales ops + competitive intel |
| Regulatory or payer shift | New guidance changes compliance posture the card asserts | Compliance + competitive intel |
| Reference customer event | Cited customer churns, goes negative, or is acquired by competitor | Customer success + competitive intel |
Any single trigger queues the card for review at the next quarterly refresh. Any two firing simultaneously fast-track a sunset with no waiting period. A trigger firing during a critical sales quarter does not get a pass — that is the quarter when a stale card is most expensive, and the policy treats it accordingly. The data feeds that drive these triggers are the same feeds that drive the rest of the competitive program; the evidence library carries expiration dates row by row, the field intel capture system surfaces competitor feature moves within forty-eight hours of any rep encountering them, and the win-loss data flows from the win-loss debrief into the analytics dashboard with deal-citation tagging.
The Sunset Workflow: Freeze, Notify, Debrief, Archive
Once a kill criterion fires, the workflow runs the same way every time. The reproducibility is the point — when the steps are the same regardless of who authored the card or who is in the room when the trigger fires, the kill is not personalized and it does not stall.
- Freeze. The card is flagged 'do not use' on the analytics dashboard and demoted in the rep-facing search index below all active cards. Reps searching for the keyword that used to surface the card now see the active replacement first and the killed card behind a warning banner.
- Notify. Every rep who cited the card on a deal in the prior ninety days gets a direct message naming the card, the trigger, the replacement plan, and the timeline. The notification is owned by competitive intel, not by the rep's manager — it goes laterally so the rep gets the information at the same time the manager does.
- Debrief open deals. Open deals that cited the card are reviewed deal by deal with the rep. Any claim that needs to be retracted in writing to the buyer gets retracted in the next touchpoint, with language drafted by competitive intel so the rep does not have to improvise the correction.
- Archive. The card moves to a clearly-marked, read-only archive folder with a kill log entry recording the trigger that fired, the date, the replacement card, and the open-deal cleanup status. The archive is searchable but excluded from the rep-facing index by default.
The workflow is documented in the governance SOP so anyone in competitive intel can run it without inventing a process. The named-step structure also makes the kill auditable — if a regulator or legal team later asks how a claim was retracted, the kill log answers the question with the receipts.
What the Archive Keeps and Why
An archive is not a graveyard. The killed card itself is preserved, alongside the kill log entry, the evidence library snapshot as it stood the day of the kill, and the deal-cleanup summary that records which open deals had to be re-pitched and how. The archive is read-only and is searchable but excluded from the rep-facing search index by default. It is preserved for three reasons.
First, competitive intel may need to compare an old claim to a new one to track how a competitor's narrative has shifted over twelve to eighteen months. Second, if the competitor announces a regression or rollback of the feature that triggered the kill, the archived card is the starting point for the next-generation card rather than starting from a blank document. Third, regulatory and legal reviews periodically ask for a historical record of what claims were in market and when. The archive answers that question definitively, which is a meaningful risk-reduction artifact in a regulated category like dental software.
Who Has Authority to Kill Between Quarterly Reviews
The quarterly refresh is the default cadence for kills, but a card that hits two triggers simultaneously cannot wait for the next quarter. The governance SOP names the head of competitive intel as the authority to invoke an emergency kill, with a sales leadership co-sign required for any card that has been cited on more than five open deals in the prior thirty days. Co-sign is not veto; it is acknowledgment that the kill is happening and that the named open deals will need debrief support from the sales side. The default posture is to invoke the kill quickly and clean up afterward, not to delay the kill while debating it.
The naming rules in the naming rules battlecard ensure killed cards do not get accidentally re-issued under the same identifier. A replacement card carries a new identifier and a clear lineage pointer back to the killed predecessor, so reps using the search index see the lineage and understand that the new card is the version of record.
How Kill Criteria Connect to the Rest of the Cluster
Kill criteria are the hygiene layer that keeps every other artifact in the Voicify A-or-B cluster honest. The evidence library supplies the expiration data the criteria run against. The analytics dashboard surfaces the trigger thresholds and the running count of cards approaching kill so program leadership sees the upstream signal before any card actually fires. The quarterly refresh is where kill candidates get evaluated; the governance SOP names the authority to invoke between cadences; the naming rules protect against accidental re-issue. Together these artifacts make the kill policy a system that runs rather than a meeting that happens — which is the entire reason a serious competitive program in dental AI by 2026 needs to write the policy down.