# The Signal-Stack Priority Score: Same-Day, This-Week, or Watch

*Take the signals firing on one account today and assign a defensible action tier that another rep would grade the same way.*

- Canonical URL: https://www.refolk.ai/guides/signal-stack-priority-score
- Pillar: Sales and go-to-market
- Format: Framework
- Published: 2026-09-16
- Last reviewed: 2026-09-16
- Reading time: 15 min

Several buying signals are firing across your accounts this week and you have to decide which one to call today. This guide is for founders selling their own product, account executives, SDR leads, and partnerships teams who need to turn a live stack of co-occurring signals on a single account into a defensible action tier: same-day call, this-week touch, or watch. It gives you a fit gate, a weighting rule, and a decay window so that another rep grades the same account the same way.

The library already catalogs triggers, grades whether a signal type is worth tracking, and routes a single inbound hand-raiser. None of them scores the live stack of signals on one account to set an action tier. That is the gap this fills.

## Why the "three signals equals a call" rule fails

The dominant public rule is that three stacked signals means a same-day call. That rule is broken because it counts instead of weighting. SalesIntel treats a single trigger as a watch item and three stacked triggers as a same-day call; Salesmotion says a website visit plus a new CRO hire plus a hiring spike indicates an active buying window. Both assert a count threshold and stop there.

The problem is that identical counts carry opposite value. "Funding plus a generic blog visit" is a weak two-signal stack. "New VP plus SDR hiring plus tech-stack change" is a strong three-signal stack. A pure count cannot tell them apart, so two reps looking at the same account reach the same number and different conclusions. That is why the score sits in the CRM and nobody sorts on it.

A defensible score fixes three things the count rule ignores: it gates on fit before it looks at intent, it weights each signal by strength and family diversity, and it decays each signal by age. Get those three right and the grade becomes reproducible.

> The escalation math is a noise-cancellation argument, not a vibe: independent signal families fail independently.

Signal stacking works statistically because independent families fail independently. Hiring can be a backfill. Funding can sit in the bank for two quarters. A tech detection can come from a contractor's job posting. But when two independent families point the same direction in the same month, the probability that all of them are noise drops sharply. That is the real reason a stack beats a single trigger, and it is why family diversity, not count, is the thing to reward.

## The four inputs that set the tier

Four inputs decide the tier: fit, family diversity, signal strength, and recency. Fit gates. The other three set the score for accounts that clear the gate.

Practitioners converge on four to six signal families. A workable grouping for account prioritization:

- **Account-level events** - funding, M&A, leadership changes, hiring surges.
- **Technographic shifts** - a competitor tool installed or removed, a stack change.
- **First-party engagement** - pricing or integration page visits, captured on your own channels and tied to a real buyer at a specific moment.
- **Third-party topic intent** - inferred in-market behavior from outside your channels.

First-party signals are the most actionable because they are observed, not inferred. Third-party intent is the noisiest and decays the fastest. The distinction matters when you weight a stack: two first-party families beat one first-party family plus one inferred topic.

> **Rule:** Count families, not signals
>
> Escalate on two or more distinct signal families inside the same decay window. Five instances of one family is one story, not a stack.

Signal strength is the second input. New executive hires produce the highest reply rates, 14 to 25%, and new leaders spend 70% of budget in the first 100 days. Funding rounds follow at 12 to 20% reply, with 71% of funded companies choosing vendors within 90 days. Department hiring surges run 10 to 18%, third-party intent surges 8 to 12%, and generic no-signal outreach only 1 to 3%. A leadership change is simply worth more points than a topic-intent blip, and your weighting should say so.

## Why fit gates everything downstream

Fit-gating is the highest-leverage step in the whole model because the noise floor is enormous. Less than 1% of website visitors match ICP, so a fit check removes 99% of noise before a human ever looks at the account. No amount of downstream weighting recovers that much signal.

The rule practitioners repeat: fit builds the list, intent orders it. Intent without fit floods reps with in-market companies that will never close. Fit without intent leaves you guessing on timing. You need both, applied in that sequence.

There is a genuine order disagreement worth naming. Most sources gate fit first, then score intent on the survivors. Apollo scores a composite first, then runs fit as a downstream disqualifier, discarding low-fit accounts even when intent is high. Both are fine. What matters is that the fit floor is explicit and written down, for example a 70 of 100 minimum, so no high-intent account routes to a rep until it clears the bar.

> **Watch out:** Intent with no fit gate flattens fast
>
> A new intent layer with no fit floor produces a small lift, then flattens. Reps chase false positives and conclude the list is only marginally better than a good ICP list. Disqualify below the floor before routing.

Fit-gating also protects the score's negative side. Without negative scoring, a poor-fit prospect can accumulate enough behavioral points from casual content consumption to cross the threshold anyway. Subtract points for a size floor breach, a competitor domain, or a personal email domain so a genuinely bad-fit account cannot buy its way into the queue.

## Reading decay windows correctly

Decay determines the tier as much as the stack does. The same three signals score differently depending on age, because each signal type has a shelf life measured in hours to weeks. A same-day account at week one is a watch account at week six.

The math is concrete. Funding decays 50% every 60 days. Third-party topic intent decays 50% every 30 days. A leadership change is freshest in the first 30 to 60 days in role. Multiply each raw signal by a recency factor before you sum, and let stale signals expire on their own curve.

| Signal | Freshness window | Source |
|---|---|---|
| Pricing-page visit | hours to under 1 week | Clay |
| Third-party topic intent | 7 to 14 days | industry |
| Funding round | 2 to 4 weeks | Clay |
| Leadership change | 30 to 60 days | Clay |
| New VP hire | 30 to 90 days | industry |

Sources disagree on funding: Clay holds it useful for two to four weeks, while another view says it loses urgency after 48 hours. Where sources conflict, pick the more conservative window for your own SLA and re-check it against your closed deals, which is the only ground truth that matters for your product.

#### How one signal becomes an action tier

1. **Fit gate** - Discard accounts below the fit floor, removing about 99% of noise
2. **Stack check** - Require two or more distinct signal families in the same window
3. **Weight** - Score each signal by strength, favoring executive and funding events
4. **Decay** - Multiply each signal by its recency factor before summing
5. **Tier** - Route to same-day, this-week, or watch on the summed score

*Fit gates first, then weighting and decay set the tier for the survivors.*

Leadership-change signals justify the highest tier precisely because two clocks overlap. The signal is freshest in the first 30 to 60 days, and the new leader spends 70% of budget in the first 100 days. The decay window and the spending window line up, which is what makes a fresh leadership signal worth a same-day call on its own strength when paired with any second family.

## The scoring procedure

Run these steps in order. The first three are RevOps setup you do once; the last five are the repeated judgement call you make per account.

#### From firing signals to an action tier

1. **Write down the ICP and fit rubric** - Build the fit rubric from closed-won and closed-lost accounts, scoring industry, size, technographic, persona, and stage as separate fields. Done means a documented field dictionary so two people score the same account the same way.
2. **Assign point values and set a fit floor** - Score fit and signal on separate scales and set a fit-gate minimum below which no intent score routes to a rep. Done means the floor is written down, for example a 70 of 100 fit minimum.
3. **Centralize signals into one feed** - Map which signals precede your own closed deals and resolve all signal families against one account record. Done means every firing signal lands on the same account object rather than in scattered tools.
4. **Apply decay weighting** - Multiply each raw signal by a recency factor before summing, using the published freshness windows. Done means stale signals auto-expire instead of scoring forever.
5. **Require a stack, not a single trigger** - Escalate only accounts where two or more distinct signal families converge in the same window, and suppress single-contact spikes. Done means no account reaches the top tier on one family or one contact alone.
6. **Assign the action tier and route with context** - Contact 80 to 100 same-day, work 60 to 79 this week, hold 40 to 59 in nurture, and pass below 40. Done means each account arrives in the rep queue with its score and the signals behind it.
7. **Act inside the shelf life** - Match response speed to each signal's decay curve rather than one standard SLA, calling pricing-page hits within hours and funding or leadership stacks within days. Done means the SLA is tied to the signal, not the clock.
8. **Review and recalibrate** - Review drivers monthly (false positives, rejected leads, conversion gaps) and recalibrate quarterly against multi-quarter cohorts. Done means the model is adjusted from what actually closed, not left frozen.

A published 50/50 split is a good starting rubric if you have none: a 50-point fit bar (industry, size, technographic, persona, stage) and a 50-point signal bar (executive hire, hiring surge, funding, intent, tech change), with Tier A above 80. Published per-signal points give you defaults to argue from: a leadership hire around +30, funding +25, expansion +20, and technology adoption +15. Treat these as a floor to calibrate, not gospel; no single cross-vendor weighting exists publicly, so yours has to earn its numbers from your own closed deals.

**100x - Drop in contact odds at 30 minutes versus 5 minutes**

The MIT/InsideSales study analyzed 15,000-plus leads and 100,000-plus call attempts; qualification odds fell 21x on the same delay.

Speed is why the tier exists at all. Teams that act on intent within 24 hours see a 29% lift in opportunity creation over slower responders. The tier is not a label; it is an SLA. Same-day means today, not this week.

## Where this goes wrong

The failure modes below are the most valuable part of this standard, because each one is a specific way a stack lies. Every check tells you what the signal looks like when it is fooling you.

| Failure mode | What it looks like | Check |
|---|---|---|
| Counting instead of stacking | Five hits from one family read as hot | Require two distinct families |
| Intent with no fit gate | New layer lifts, then flattens | Disqualify below fit floor first |
| No decay | A Q1 pricing visit still scoring in Q3 | Expire signals on recency |
| No negative scoring | Poor fit crosses threshold on casual content | Subtract for size, competitor, personal domain |
| Weak stack looks strong | Funding plus a generic blog visit | Weight strength, not raw count |
| Single-contact spike | One person, one page, one session | Require 2 contacts, 2 topics, 2 sessions |

Two of these deserve extra weight. The first is chasing corpses. Intent has a shelf life measured in weeks; research from six months ago is historical interest, not current buying intent. A Q1 pricing visit that still scores in Q3 is a false positive that pulls a rep away from a genuinely fresh account. Recency expiry is not optional.

The second is the single-contact spike mistaken for account intent. Apollo's suppression rule is the cleanest guard: require activity from at least two contacts, across at least two intent topics, in at least two separate sessions, and suppress single-contact spikes. One person hammering your pricing page is curiosity, not a buying committee.

I ran this search: `SaaS companies that raised a Series B in the last 30 days and are hiring SDRs` - [see the full result list](https://www.refolk.ai/s/fkpr9ek88m).

*Returns accounts where a funding event and a hiring surge stack inside overlapping decay windows, ready to fit-gate and tier.*

Building that centralized feed by hand is the friction this whole model runs into. Resolving funding, hiring, technographic, and engagement signals against one account record, across GitHub, LinkedIn, and the open web, is where most stacks fall apart. [Refolk](/) lets you ask for the stacked condition in plain English and get the accounts back already joined, so the judgement you are left making is the tiering, not the data plumbing.

## Keeping the score credible

False positives do not just waste time; they permanently kill the score's credibility. Once a rep works a dozen "hot" accounts that turn out cold, the score stops driving behavior. It stays in the CRM, and nobody sorts on it. This is why the maintenance thresholds are load-bearing, not housekeeping.

The recommended targets: keep the false-positive rate below 30%, the false-negative rate under 10%, and review conversion by score range every 90 days. Review drivers monthly and recalibrate quarterly against multi-quarter cohorts. The model has to learn from what actually closed, and it cannot learn if reps reject leads without saying why.

> **Tip:** Make rejection reasons mandatory
>
> When sales rejects a routed account without logging a reason, RevOps has no signal to recalibrate. Require a rejection reason code in the CRM on every passed lead so a false positive becomes data instead of a shrug.

Calibration needs an owner, and the span is wider than most teams assume. In Refolk's index, US SDR leadership numbers 334 people and US Revenue Operations 402, against 22,714 frontline SDRs. That is one calibration owner per 68 reps on the leadership side and per 56 on the RevOps side, so the recalibration process has to be documented and cheap to run, not dependent on one person's memory.

| Role band (US) | Count | Reps per owner |
|---|---|---|
| Frontline SDR | 22,714 | - |
| SDR leadership | 334 | 68.0 |
| Revenue Operations | 402 | 56.5 |

Counts from Refolk's index; ratios derived as a proxy for how many reps one calibration owner supports.

The tiering rule itself is worth pinning where the whole team can see it. This is the copy-paste version of the four inputs.

**Signal-Stack Priority tier rule**

```
STEP 1 - Fit gate: if fit score < floor (e.g. 70/100), STOP. Route to nurture or pass.
STEP 2 - Stack gate: require 2+ distinct signal families in the same window, and 2 contacts / 2 topics / 2 sessions for first-party intent. Else demote one tier.
STEP 3 - Score: sum (signal strength weight x recency factor) for each live signal.
  Suggested weights: leadership hire 30, funding 25, expansion 20, tech change 15, hiring surge 15, third-party intent 10.
  Recency factor: 1.0 fresh, 0.5 at the signal's half-life, 0 past the freshness window.
STEP 4 - Tier: 80-100 SAME-DAY call | 60-79 THIS-WEEK touch | 40-59 WATCH in nurture | below 40 PASS.
STEP 5 - SLA: same-day = today within the signal shelf life; this-week = within 5 business days.
```

*Adjust the fit floor and point weights to your closed-won data; keep the tier bands fixed so grading stays consistent.*

Before you call an account graded, run this checklist. It is the difference between a score another rep trusts and one they quietly ignore.

#### Before you call the tier final

- [ ] The account cleared the documented fit floor before any intent scored
- [ ] At least two distinct signal families are firing in the same window
- [ ] First-party intent meets the 2 contacts / 2 topics / 2 sessions bar
- [ ] Each signal was multiplied by a recency factor, and stale signals expired
- [ ] Negative scoring was applied for size floor, competitor, or personal domain
- [ ] The account arrives in the queue with its score and the signals behind it
- [ ] A rejection reason code is required if a rep declines to work it

## What to do next

Start with the fit gate, because it does the most filtering for the least effort, and less than 1% of your inbound will clear it. Write the floor down first, then add the stack rule, then layer decay. Build the model in that order and you can ship a defensible tier this week rather than waiting for a perfect weighting.

Then close the loop. Pull a cohort of accounts you tiered same-day 90 days ago and check what closed. If your false-positive rate is above 30%, tighten the fit floor or raise the stack bar before you touch the weights, because a leaky gate corrupts every number downstream. Re-check your decay windows against your own funnel too; the published half-lives are a starting point, but your product's real shelf life is whatever your closed-won data says it is. Keep that recalibration on a 90-day clock and the score stays something the team actually sorts on.

## Frequently asked questions

### How many stacked signals mean I should call today?

Not a raw count. Two distinct signal families converging inside the same decay window, weighted by strength, beat five instances of one family. A new VP hire plus an SDR hiring surge plus a tech-stack change is a same-day call; funding plus a single generic blog visit is not. Require two different families, weight each by strength and recency, and only escalate accounts that clear your fit floor first.

### Should I gate on fit before or after scoring intent?

Sources disagree. Most gate fit first, then order the survivors by intent, which is cleaner because less than 1% of visitors match ICP and the gate removes 99% of noise before a human looks. Apollo scores a composite first, then runs fit as a downstream disqualifier. Either works if the fit floor is explicit; pick one and document it so reps do not diverge.

### How fast do buying signals decay?

By signal type. A pricing-page visit holds full value for hours and is cold within a week. Funding stays useful two to four weeks, decaying 50% every 60 days. A leadership change is freshest in the first 30 to 60 days in role, and third-party topic intent decays 50% every 30 days. Build SLAs around each signal's curve, not one standard response time.

### Why do two reps grade the same account differently?

Because the common public rule, three stacked signals equals a call, has no weighting, no decay, and no fit gate. Two reps counting the same three signals reach the same count but different value, since one stack may be weak and stale and the other strong and fresh. Writing down the fit floor, the per-family weights, and the decay windows is what makes the grade reproducible.

### How do I keep the score from losing credibility?

Watch the false-positive rate. Once a rep works a dozen hot accounts that turned out cold, they stop sorting on the score and it dies in the CRM. Hold false positives below 30% and false negatives under 10%, review every 90 days, and require a rejection reason code on every passed lead so RevOps has something to recalibrate against.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/signal-stack-priority-score*