feat(leads): intent qualification on by default, at a measured bar

components/leads/campaign-panel.tsx

@@ -9,6 +9,7 @@ import {
   toggleCampaignAction,
 } from "@/app/actions/leads";
 import { LEAD_RUN_CREDITS } from "@/lib/credits";
+import { DEFAULT_MIN_INTENT } from "@/lib/outreach/intent";
 
 export type CampaignRun = {
   ran_at: string;
@@ -84,9 +85,10 @@ export function CampaignPanel({
   // scan of the prospect to contribute. Selecting the audit pitch turns it
   // back on, because that pitch is built from findings.
   const [scanProspects, setScanProspects] = useState(false);
-  // Null means "don't qualify on intent" — the behaviour every campaign had
-  // before this existed, and the default for anything already running.
-  const [minIntent, setMinIntent] = useState<number | null>(null);
+  // On by default for a new campaign. Writing to people who never asked is
+  // the failure mode the whole feature exists to avoid, so it should be what
+  // you turn off deliberately rather than what you remember to turn on.
+  const [minIntent, setMinIntent] = useState<number | null>(DEFAULT_MIN_INTENT);
   const [goal, setGoal] = useState("");
   const [generating, setGenerating] = useState(false);
   const [pitchNote, setPitchNote] = useState<string | null>(null);
@@ -503,7 +505,7 @@ export function CampaignPanel({
               type="checkbox"
               className="mt-1"
               checked={minIntent !== null}
-              onChange={(e) => setMinIntent(e.target.checked ? 40 : null)}
+              onChange={(e) => setMinIntent(e.target.checked ? DEFAULT_MIN_INTENT : null)}
             />
             <span>
               Only work leads who publicly asked to buy

lib/outreach/intent.ts

@@ -18,8 +18,17 @@
 // The adapters live in intentSources.ts; this file knows nothing about where
 // the words came from.
 
-/** Hours after which a signal has decayed to roughly half its strength. */
-const HALF_LIFE_HOURS = 36;
+/**
+ * Hours after which a signal has decayed to roughly half its strength.
+ *
+ * Matches the 72-hour window the Reddit scorer already treats as the edge of
+ * repliable, rather than a number chosen independently. The first value here
+ * was 36, which priced a day-old "can anyone recommend a X" at barely more
+ * than a fresh unattributed grumble — and those threads stay answerable for
+ * days. Replying to a three-day-old request costs little; missing it costs a
+ * lead.
+ */
+const HALF_LIFE_HOURS = 72;
 /** Past this, treat a signal as stale regardless of how strong it was. */
 const MAX_AGE_HOURS = 24 * 14;
 
@@ -340,8 +349,17 @@ export function scoreIntent(input: {
   };
 }
 
-/** Default bar for acting on a signal. Roughly "a fresh solicitation". */
-export const DEFAULT_MIN_INTENT = 40;
+/**
+ * Default bar for acting on a signal.
+ *
+ * Chosen from the measured spread rather than picked round. At 72-hour
+ * half-life a day-old "can anyone recommend a X" scores 56 and a fresh
+ * unattributed grumble scores 42, and 50 sits in the gap between them: real
+ * requests stay in for about three days, complaints from nobody in particular
+ * never get in. A team lead saying "our team is struggling with X" clears it
+ * at 54, because that is a lead and a stranger's grumble is not.
+ */
+export const DEFAULT_MIN_INTENT = 50;
 
 /**
  * Whether a lead clears a campaign's bar.

supabase/migrations/20260729070000_intent_on_by_default.sql

@@ -0,0 +1,18 @@
+-- Intent qualification on by default.
+--
+-- Writing to people who never asked for anything is the failure mode the
+-- feature exists to avoid, so it should be what somebody turns off
+-- deliberately rather than what they remember to turn on.
+--
+-- 50 is taken from the measured spread rather than picked round: at a 72-hour
+-- half-life a day-old "can anyone recommend a X" scores 56 and a fresh
+-- unattributed grumble scores 42, so the bar sits in the gap between a request
+-- and a complaint.
+alter table public.outreach_campaigns
+  alter column min_intent set default 50;
+
+-- Existing campaigns were created before the column existed and would
+-- otherwise keep selecting purely on resemblance forever.
+update public.outreach_campaigns
+set min_intent = 50
+where min_intent is null;

tests/intent.test.ts

@@ -52,8 +52,10 @@ describe("recency is half the signal", () => {
   });
 
   it("halves at the half-life", () => {
-    expect(recencyFactor(36)).toBeCloseTo(0.5, 2);
-    expect(recencyFactor(72)).toBeCloseTo(0.25, 2);
+    // 72h, matching the window the Reddit scorer already treats as the edge
+    // of repliable rather than a second number invented alongside it.
+    expect(recencyFactor(72)).toBeCloseTo(0.5, 2);
+    expect(recencyFactor(144)).toBeCloseTo(0.25, 2);
   });
 
   it("gives up entirely on something ancient", () => {
@@ -124,8 +126,26 @@ describe("qualifying against a campaign's bar", () => {
     expect(qualifies(weak, undefined)).toBe(true);
   });
 
-  it("holds a weak signal below the bar", () => {
-    expect(qualifies(score("Struggling with load testing", 24 * 6), DEFAULT_MIN_INTENT)).toBe(false);
+  it("holds a bare complaint below the bar, however fresh", () => {
+    // The bar sits in the measured gap between "somebody asked for something"
+    // and "somebody grumbled": a fresh unattributed complaint scores 42.
+    expect(qualifies(score("Struggling with load testing", 1), DEFAULT_MIN_INTENT)).toBe(false);
+  });
+
+  it("keeps a day-old request in", () => {
+    // These threads stay answerable for days, and losing them to an
+    // over-eager decay curve is what the 72h half-life fixed.
+    expect(qualifies(score("Can anyone recommend a load testing tool?", 24), DEFAULT_MIN_INTENT)).toBe(true);
+  });
+
+  it("lets a team lead's complaint through where a stranger's does not", () => {
+    // "Our team is struggling with X" is a lead. The same words from nobody
+    // in particular are not.
+    expect(qualifies(score("Our team is struggling with load testing.", 6), DEFAULT_MIN_INTENT)).toBe(true);
+  });
+
+  it("drops a request that has gone stale", () => {
+    expect(qualifies(score("Can anyone recommend a load testing tool?", 24 * 5), DEFAULT_MIN_INTENT)).toBe(false);
   });
 
   it("passes a fresh explicit request", () => {