Playbook: Recency Engagement Score

Introduction

The Recency Engagement Score is most valuable when used as part of an ongoing decision process rather than as a one-time report. The scenarios in this playbook illustrate common situations marketing teams encounter at intake, before sends, after sends, and in ongoing list management, along with questions worth investigating and actions they may consider depending on their organization’s goals and risk tolerance.

It helps to be precise about what the score measures, because it is easy to confuse with the question Inbox Placement answers.

The Recency Engagement Score “Spruce” tells you the recency of truly valid subscriber engagement with marketing email, as observed across AudiencePoint’s second-party data pool. It is a property of the subscriber’s behavior, not of your program.

Spruce identifies patterns that warrant investigation and opportunities to inform your segmentation.

---

A note on the data pool and privacy

Spruce is generated by matching your subscribers against AudiencePoint’s second-party data pool — behavioral signals shared directly between AudiencePoint and its network of participating brands, spanning hundreds of millions of email addresses across hundreds of brands and 85 trillion tracked events. That scale is what makes the score possible: it is how machine-generated opens can be filtered out with confidence, and how engagement your own sends never captured becomes visible.

It is also the part of the product clients ask about first, so it is worth stating plainly how it works:

• Addresses are hashed before matching. Each address is converted to a fixed-length string, so raw email addresses are never exposed in the process.
• No brand’s subscriber records are exposed to any other participating brand. Matching and scoring happen inside a privacy-safe clean room environment.
• No personally identifiable information changes hands. AudiencePoint stores no PII in the global data pool.
• Only the resulting score is returned to you. Not the underlying events, and not their source.
• AudiencePoint acts strictly as a third-party data processor, following controller instructions, supporting data portability, and honoring erasure requests.

The practical consequence for this playbook: the score tells you that a subscriber engaged recently, and how recently. It does not tell you with whom, and it cannot. Several scenarios below depend on understanding that boundary.

---

Score quick reference

Score	Verified activity window	What it asserts
5	Real open and click recorded in the last 14 days	High-confidence active human, right now
4	Real open or click recorded in the last 60 days	Verified activity, recent
3	Real open or click recorded in the last 120 days	Verified activity, moderately recent
2	Real open or click recorded in the last 240 days	Verified activity, fading
1	Real open or click recorded in the last 360 days	Verified activity, distant
0	No real activity in the last 360 days, but real activity within the last 720 days	Known dormant
Unmatched	Insufficient engagement history in the pool to produce a score	Not scored — uncertainty, not confirmation of inactivity
Invalid	Address cannot be reached — syntactically broken, a known spam trap, an abandoned inbox, or otherwise undeliverable	Not scored — unreachable

“Real” means human-initiated opens and clicks. Proxy and automated events are excluded.

The scale reflects recency, not frequency. A subscriber who opened 40 times two years ago scores lower than one who clicked once last week. Recent behavior predicts future behavior more reliably than volume of past activity. It also means a high score is not a measure of how much someone engages — only how recently.

The ladder is not uniform

The windows above look like a single recency ladder, but the scale makes three different kinds of statement, and reading it as uniform leads to predictable mistakes.

The stricter bar at 5 stars is deliberate. Since Apple Mail Privacy Protection and similar features became widespread, an open on its own is weak evidence of human attention. An open paired with a click is strong evidence. A 5-star score is therefore a confidence statement, not merely a recency statement.

Two consequences worth carrying into every scenario below:

• 4 to 5 is not one step up. It is a different assertion with a higher evidentiary bar.
• The 5-star tier is a 14-day rolling window and it churns by design. Subscribers move in and out of it week to week without anything meaningful having changed. Do not build stable operational rules on it.

---

What these scores are not

• Not an inbox placement signal. It says nothing about whether your mail reaches the inbox.
• Not an address verification result. A subscriber can score well and still be risky to send to. A known spam trap, a role address, or a high-complaint address puts your deliverability at risk regardless of its score, and a strong score does not offset that. Verification and engagement scoring answer different questions; run both.
• Not a measure of engagement with your program. It reflects engagement with marketing email generally, across brands.
• Not an affinity, intent, or value score. A 5 means someone opened and clicked something, somewhere, recently. It does not mean they like your brand or will buy from you.

A high score does not guarantee engagement with your next send. It indicates the subscriber is a stronger candidate than one who has not been active.

---

Reading Spruce alongside your other signals

There are specific use cases where the score makes sense on its own, but the most value often comes from resolving ambiguities that your other data structurally cannot.

Coverage comes first

Spruce is available before you send, which is what makes most of Section II possible. That availability depends on AudiencePoint having activity history for those subscribers in the pool.

Every address in an analysis resolves to one of three states: Matched (enough signal to assign a 0–5 score), Unmatched (a real address with insufficient history to score), or Invalid (unreachable). Your Matched share is your coverage. Before drawing any conclusion from a distribution — or building an audience from one — check what yours is. A read on 30% of a list is a read on 30% of a list, and a high Unmatched share is more common in B2B, regional, international, and highly specialized audiences.

Spruce and your own engagement data

	Engaged with your program	Not engaged with your program
Higher Activity Band (3–5)	Healthy core. Your benchmark audience. This is who to lean on during warming, recovery, and business-critical sends.	The opportunity. They engage with marketing email. They are not engaging with yours. That points to relevance, cadence, or placement — not to a dead subscriber. Typically the highest-yield reactivation population on the file.
Lower Activity Band (2 or less)	Worth investigating. If you’re not sharing your engagement data feed with us consider doing so. If you are sharing your feed there are two probable causes for the difference. The most common is that you ESP filters failed to detect an automated engagement when caught. The other is when the engagement was within the past few days and our latest recency engagement scores haven’t been published yet (2x per week).	Genuinely dormant. Two independent signals agree. The most defensible re-engagement and suppression candidates you have.

Spruce and Inbox Placement

These measure different things and should be expected to diverge. Divergence is the informative case, not an error.

	High Spruce (4–5 bands)	Low Spruce (0–1 stars)
Inbox	Your strongest audience. Reachable and attentive.	Reachable but not paying attention. Relevance and cadence work first, re-engagement second, suppression last.
Spam	Highest-priority remediation. There is an actively engaged person behind this address and your mail is being filtered away from them. This is recoverable value with a fixable cause, not a dead contact.	Lowest priority. Both signals negative. Handle per your suppression policy.

The Spam + high-Spruce cell is the clearest illustration of why the two analyses are stronger together. Placement tells you the mail is not landing. Spruce tells you it mattered that it did not.

---

Section I — Intake and Onboarding

Prioritizing a new list before you send to it

Situation: You are about to send to addresses your program has not mailed before, or has not mailed in a long time. This covers three common entry paths: a newly acquired list (purchased, appended, co-registration, or partner-supplied), a list inherited through an acquisition or platform migration, and an owned list that has gone dormant.

What It May Indicate: The causes differ, but the risk is the same. Addresses you have not recently mailed carry unknown reachability and unknown attention. Sending to them at volume before you know which is which is one of the most reliable ways to damage a healthy program. Spruce tells you how much genuine attention exists in the file before you commit any sending reputation to it.

Questions to Investigate

• Whether the addresses have been run through AudiencePoint Address Verification
• Coverage — what share of the file carries a score at all
• The score distribution, particularly the size of the 4- and 5-star population relative to 0- and 1-star
• How the addresses were originally collected, and on what consent basis
• How long since the file was last mailed, and by whom
• Whether Invalid addresses have been identified and removed

Actions to Consider

• Verify before you score, and score before you send. Address Verification removes what cannot be reached; Spruce tells you how much of the remainder is paying attention. Running both first is materially cheaper than recovering from a bad first send.
• Read the distribution rather than the average. A file with a healthy 4- and 5-star population and a long 0- and 1-star tail is a sequencing problem. A file that is almost entirely 0- and 1-star is a different conversation about whether to mail it at all.
• Sequence by score. Do not blast. Begin with the strongest, most clearly consented portion and expand as performance holds.
• Expect a meaningful Unmatched share on genuinely new addresses, and do not treat it as a negative signal. Continue including those subscribers under your standard practices so behavioral data can accumulate.
• For a dormant owned list specifically, weigh whether re-permission is appropriate before resuming normal cadence. Time away changes both reachability and expectation.
• Run Inbox Placement alongside this. A file can be attentive and still unreachable, and you want to know which problem you have before you start.

---

Comparing acquisition sources by score distribution

Situation: You want to know which acquisition channels are producing subscribers who actually engage.

What It May Indicate: Acquisition quality is the most common upstream driver of long-term list health, and it is usually invisible in acquisition reporting, which measures cost and volume rather than attention. Sources that produce disengaged, poorly consented, or stale addresses will show a distribution skewed toward the low end of the scale and a high Invalid rate — often well before the effect appears in campaign metrics or deliverability.

Questions to Investigate

• Score distribution by source, compared at matched cohort ages
• Unmatched and Invalid rates by source
• Consent basis and collection method for each source
• Whether validation happens at the point of collection
• Volume and spend directed to each source

Actions to Consider

• Compare like-for-like by cohort age. This is the discipline that makes the whole comparison valid. Recency scoring always favors younger cohorts, so a source you started using three months ago will look better than one you have run for three years even if the underlying quality is identical. Compare a source’s 90-day-old cohort to another source’s 90-day-old cohort.
• Treat a persistently high Invalid rate as an acquisition hygiene problem, not a list hygiene problem. It points at the form, not the file.
• Improve validation at the point of collection for weaker sources — real-time verification, confirmed opt-in, stricter form handling. Raising cohort quality before addresses enter the program is cheaper than managing them afterward.
• Reassess spend across sources on distribution and downstream engagement, not on cost-per-acquisition alone. A cheaper source producing mostly 0- and 1-star subscribers is not cheaper.
• Consider stricter entry criteria or re-permission for sources that keep producing low distributions despite onboarding adjustments.

---

Setting welcome-series entry policy

Situation: You are deciding how new subscribers should be routed into onboarding, and whether all of them should receive the same treatment.

What It May Indicate: Most programs treat every new subscriber identically. Where new addresses arrive with existing pool history — which is common for anything other than organic first-party signup — Spruce can differentiate at the point of entry, before any sending history exists to differentiate on.

Questions to Investigate

• What share of new subscribers carry a score at entry versus reading Unmatched
• The distribution among those who do carry a score
• Whether score at entry varies meaningfully by acquisition source
• What the transition looks like from welcome cadence to promotional cadence

Actions to Consider

• Use score at entry to set the starting cadence, not to exclude. A new subscriber scoring 0 or 1 has consented and deserves an onboarding experience; they may simply warrant a more measured one.
• Give Unmatched new subscribers your standard onboarding path. They are the population most likely to be genuinely new, and onboarding is how they generate the signal that will score them.
• Route new subscribers who already score 4 or 5 into your normal cadence with confidence. They are demonstrably active in email.
• Invest in the welcome experience regardless of score. Onboarding is where consent and expectation are established, and it does more for long-term engagement than any downstream segmentation will.

---

A new cohort reads mostly Unmatched

Situation: A recently acquired or recently onboarded group of subscribers returns a high proportion of Unmatched values.

What It May Indicate: For a genuinely new cohort this is expected and is not a negative signal. Unmatched means AudiencePoint does not yet have sufficient behavioral evidence in the pool to score the address — not that the subscriber is inactive. Newer subscribers, audiences with limited observable history, and specialized or international audiences all produce higher Unmatched rates. Coverage builds over time as behavioral data accumulates.

Questions to Investigate

• Age of the cohort — how long since these addresses entered your program
• Whether the Unmatched share is concentrated in one acquisition source
• Whether the audience skews B2B, regional, or international, where pool coverage is typically thinner
• Whether you are sending AudiencePoint your open and click data, if the rate is unexpectedly high across the whole file
• If you deliver by FTP or SFTP, whether the file you are sending has been refreshed recently

Actions to Consider

• Continue mailing Unmatched subscribers under your standard practices. This is the central point. Withholding mail from them prevents the accumulation of the very behavioral data that would score them, so the uncertainty becomes self-perpetuating.
• Watch coverage growth over successive refreshes rather than treating a single high reading as a finding. A cohort trending from 70% Unmatched toward 40% is working as intended.
• If Unmatched is very high across your entire file rather than one cohort, treat it as an integration question rather than an audience question, and confirm engagement data is flowing to AudiencePoint. Contact support@audiencepoint.com to set up an automation that keeps lists and profiles current.
• Do not treat Unmatched as a low score. It is not the bottom of the scale; it is the absence of the scale.

---

Section II — Before the Send

Building a high-value campaign audience

Situation: You are preparing an important campaign and want it to reach subscribers who are in a position to notice it.

What It May Indicate: Audience selection influences campaign outcomes more than most teams account for. Spruce provides subscriber-level attention data before deployment, which allows audience decisions to be made deliberately rather than by defaulting to “everyone who has not unsubscribed.”

Questions to Consider

• Coverage across the intended audience
• Score distribution within that audience
• Inbox Placement designations for the same audience
• Your own engagement history with these subscribers
• The campaign objective and what a poor outcome would cost

Actions to Consider

• Check coverage before anything else. If most of the intended audience is Unmatched, the distribution is not telling you much and your existing audience practices should carry the decision.
• Structure the audience by score rather than filtering to a single tier. 4- and 5-star subscribers are your highest-confidence population; 3-star subscribers are a reasonable extension when you need more reach; 0- and 1-star subscribers warrant a deliberate decision rather than automatic inclusion or exclusion.
• Combine with Inbox Placement rather than choosing between them. A high-Spruce subscriber with a Spam designation will not see the campaign regardless of how attentive they are.
• For business-critical sends, some organizations narrow to 4- and 5-star subscribers with a confirmed Inbox designation. This is defensible for a small number of sends per year and unsustainable as a default.
• Resist building the audience on 5-star subscribers alone. See the cadence scenario below for why.

---

Warming a new IP or sending domain

Situation: You are establishing a new sending IP or domain and need to build a positive reputation from a standing start.

What It May Indicate: Warming is one of the few situations where the margin for error is genuinely narrow. Mailbox providers form an early view of a new sender quickly, and that view is difficult to revise. The signal you want to generate during this window is consistent engagement from real, attentive humans.

Questions to Investigate

• Size of your 4- and 5-star population, and whether it is large enough to sustain the warming schedule
• Inbox Placement designations within that population
• Which mailbox providers the warming audience concentrates in
• Your planned volume ramp against the size of the highest-confidence audience

Actions to Consider

• Warming is where the 5-star tier earns its strictness. The open-and-click bar is exactly what you want when the cost of a weak early signal is high and the audience is deliberately small. Everywhere else that volatility is a liability; here it is the point.
• Begin with 4- and 5-star subscribers who also carry an Inbox designation. Attention without reachability generates no signal.
• Expand outward by score as placement holds — 5-star, then 4-star, then 3-star — rather than by raw volume.
• Exclude 0- and 1-star subscribers entirely during the warming window. There is no version of this where they help.
• Hold Unmatched subscribers out of the initial warming audience specifically. This is a narrow exception to the general rule of continuing to mail them: during warming you want only confirmed signal, and you can resume normal treatment once the ramp is complete.
• Monitor placement by provider as you expand, and slow the ramp rather than pushing through a dip.

---

Building engagement-tiered cadence

Situation: You are designing send frequency that varies by how engaged a subscriber is, rather than mailing the whole file on one cadence.

What It May Indicate: Uniform cadence over-mails your least engaged subscribers and under-mails your most engaged ones. Tiering addresses both. It is the most common structural use of Spruce, and also the one most often built in a way that does not hold up.

Questions to Investigate

• Your current cadence and whether it varies by anything today
• Score distribution across the file, and how large each tier actually is
• How much subscribers move between tiers across refreshes
• Complaint and unsubscribe rates by tier

Actions to Consider

• Do not use 5 stars as a cadence boundary. This is the most important guidance in the scenario. The 5-star tier requires an open and a click within a rolling 14-day window, so subscribers enter and leave it constantly without any real change in their behavior, and it covers a small share of most files. Cadence rules built on it will reassign people weekly for no reason.
• Use 4 stars and above — or 3 and above — as your working definition of an engaged tier. These are stable, meaningfully sized, and reflect a real distinction.
• Build three tiers rather than six. Something like: engaged (4–5 stars) receives your full cadence, moderate (2–3 stars) receives a reduced cadence, dormant (0–1 stars) receives a re-engagement track rather than the promotional stream.
• Give Unmatched subscribers your standard cadence rather than a reduced one, unless business considerations suggest otherwise.
• Reassess tier assignment on a defined schedule — monthly or quarterly — rather than on every refresh. Frequent reassignment creates inconsistent subscriber experience and makes results impossible to read.
• Let business value override the tier where it should. A high-value customer who scores 2 may warrant full cadence regardless.

---

Scoping a reactivation campaign

Situation: You are planning a win-back or re-engagement campaign and need to decide who belongs in it.

What It May Indicate: Reactivation is where poor targeting does the most damage. Mailing a large block of unreachable or long-dormant addresses can generate complaints and bounces that harm the program more than the recovered subscribers are worth. Spruce brings precision to the question of who is plausibly recoverable.

Questions to Investigate

• Score distribution among the subscribers you are considering
• Inbox Placement designations for the same group — particularly whether they are reachable at all
• How your own engagement data characterizes them, and how it compares to the score
• Their acquisition source, original consent basis, and historical value
• What reactivation attempts have already been made and how they performed

Actions to Consider

• Start with the opportunity quadrant. 4- and 5-star subscribers who are inactive in your program are engaging with marketing email — just not yours. They are the highest-yield reactivation population you have, and the problem is more likely relevance, cadence, or placement than disinterest. Treat them first and treat them differently.
• For 2-star subscribers, a reactivation attempt is a reasonable investment. They have shown genuine engagement within the last eight months.
• For 1-star subscribers, an attempt may still be warranted, but keep it low volume, closely monitored, and separate from any send carrying deliverability risk you cannot absorb.
• For 0-star subscribers, weigh carefully. Verified activity exists somewhere in the last 720 days but nothing in the last 360. Some organizations make one final attempt; others move directly to suppression review.
• Cross-reference Inbox Placement before you build the audience. A reactivation campaign to subscribers designated Spam will not arrive, and the non-response will be misread as disinterest.
• Use clicks and conversions rather than reported opens to judge whether the campaign worked. Opens are the least reliable measure available for exactly this population.
• Set a decision point in advance. If the campaign does not produce a response within a defined window, move the cohort toward reduced frequency and eventual suppression per policy rather than repeating the attempt indefinitely.

---

Section III — After the Send

Your open rates look healthy but the scores say otherwise

Situation: A segment shows solid open rates in your ESP, but its Spruce distribution is concentrated in the low end of the scale.

What It May Indicate: This is usually a measurement problem rather than an audience problem, and it is one of the most valuable things the score surfaces. Apple Mail Privacy Protection and similar features pre-fetch message content on the subscriber’s behalf, registering an open before — or without — the person seeing the message. Your ESP records that as engagement. Spruce excludes it. When your data says engaged and the pool says dormant, the pool is generally measuring the thing you actually care about.

The broader implication is worth sitting with: if this is happening, your internal definition of “engaged” is inflated, and every downstream decision built on it — segmentation, suppression, cadence, reporting — inherits the error.

Questions to Investigate

• How your organization currently defines “engaged,” and whether that definition rests on opens
• The gap between open-based and click-based engagement rates for the segment
• What share of the segment uses Apple Mail or another privacy-protected client
• Whether click and conversion rates corroborate the open rates or contradict them
• How long the divergence has been present

Actions to Consider

• Reconcile the definition rather than the data. The score is not disagreeing with your ESP about what happened; it is disagreeing about what counts. Decide which definition you want driving decisions.
• Move your internal engagement definition toward clicks and conversions. Opens have been unreliable for years and this quantifies how unreliable they have become for your specific audience.
• Re-examine any suppression or cadence rules built on open-based engagement. If opens are inflated, those rules have been retaining subscribers they should have flagged.
• Treat the affected segment as less engaged than your reporting has been indicating, and adjust cadence accordingly before the discrepancy pressures deliverability.
• Watch Inbox Placement alongside it. Sustained genuine disengagement eventually pressures placement even while open rates look stable.

---

Your data shows more recent activity than the score reflects

Situation: You can see an open or click in your ESP from the last few days, but the subscriber’s Spruce score does not reflect it — sometimes reading 0, 1, or Unmatched despite recent activity in your own records.

What It May Indicate: This is the most common data-trust question the score raises, and there are two ordinary explanations before anything is actually wrong.

Refresh timing. AudiencePoint updates its database several times a week. Activity from the last few days may simply not have been processed into the pool yet. Depending on how your integration is configured, additional sync delay may apply on top of that. A discrepancy measured in days is usually this.

Automated activity filtering. Spruce counts only human-initiated opens and clicks; your ESP generally does not make that distinction. If AudiencePoint classified the event as machine-generated, it will not appear in the score no matter how long you wait. Common sources include Apple Mail Privacy Protection pre-fetching, corporate security scanners and link-protection services following every URL in a message, and spam-filter link checking — all of which look identical to human activity in ESP reporting. This is the more consequential explanation: it means the subscriber may not have engaged at all, and your data is showing you a machine.

Questions to Investigate

• How recent the activity is. Anything inside the last few days may simply be a timing gap.
• Whether the pattern is a single subscriber or a systematic gap across many
• Whether the activity was an open, a click, or both — and whether clicks arrive suspiciously fast after send, or hit every link in the message
• Whether the subscribers use privacy-protected mail clients
• Whether your engagement data is reaching AudiencePoint, if the gap is broad rather than isolated
• When your integration last synced, and whether the file being scored is current

Actions to Consider

• Check the calendar before escalating. Given a twice-weekly refresh, wait for the next update before concluding anything from a discrepancy of a few days.
• Where the gap persists across refreshes, treat it as an automated-activity signal rather than a scoring error. That is the score doing its job.
• Look at the shape of the activity. Clicks landing within seconds of deployment, clicks on every link in a message, and engagement concentrated in a single corporate domain are all consistent with scanning rather than reading.
• If the gap is broad and systematic rather than scattered, verify the integration — confirm open and click data is flowing to AudiencePoint, and that any FTP or SFTP file you supply is being refreshed. Contact support@audiencepoint.com about an automation if the file is maintained manually.
• Use the discrepancy diagnostically rather than treating it as noise. A segment where your data consistently shows more activity than the score reflects is a segment where your engagement metrics are being inflated by machines, and that is worth knowing.
• Escalate to your AudiencePoint representative if a material share of the file shows recent verified human activity in your data that never appears in the score across multiple refreshes.

---

A campaign underperformed — audience or content?

Situation: A campaign returned weaker results than expected and you need to know where to look.

What It May Indicate: Underperformance has three broad causes: the mail did not arrive, the audience was not in a position to notice, or the message did not land. Inbox Placement addresses the first. Spruce addresses the second. Isolating both narrows the third considerably.

Questions to Investigate

• The score distribution of the audience that received the campaign, compared to audiences for campaigns that performed normally
• Inbox Placement for the same audience
• Whether the audience composition differed from your usual send — a broader list, a new cohort, a reactivation block folded in
• Which metric underperformed, and where in the funnel
• Creative, subject line, offer, and timing relative to campaigns that performed

Actions to Consider

• Compare the audience’s distribution against a campaign that performed as expected. If this send reached a materially less engaged population, that is likely most of your answer.
• Rule out reachability first. A healthy score distribution with poor placement means the audience was attentive and never saw the message.
• If both the distribution and placement were healthy, the constraint is the content. That is a genuinely useful conclusion, and it is difficult to reach with confidence any other way.
• Check whether audience composition drifted. Campaigns often underperform because the audience quietly broadened, not because the creative got worse.
• Avoid drawing conclusions from a single send. Look at whether the pattern holds across comparable campaigns.

---

Judging whether reactivation actually worked

Situation: You ran a re-engagement campaign and need to determine whether it produced durable results.

What It May Indicate: Reactivation is unusually easy to misjudge. Opens are inflated for precisely this population, and a single response does not establish that a subscriber has genuinely returned. Score movement across refreshes is slower but far more reliable, because it reflects verified human activity and cannot be inflated by proxy opens.

Questions to Investigate

• The cohort’s score distribution before the campaign, captured as a baseline
• Distribution across subsequent refreshes — has any share moved up a tier?
• Clicks and conversions, rather than opens, during the campaign itself
• Whether movement concentrates in subscribers who were 2-star versus 0- and 1-star
• Inbox Placement for the cohort, in case non-response reflects non-arrival

Actions to Consider

• Capture a baseline distribution before you send. Without it there is nothing to measure against, and this is the most common reason reactivation results cannot be evaluated afterward.
• Give it time. Tiers are defined in windows of 14 to 360 days and the database refreshes several times a week, so meaningful movement takes weeks. A verdict one week after the campaign is premature.
• Judge in-campaign performance on clicks and conversions. Opens will overstate success for this cohort specifically.
• Segment the result by starting score. If 2-star subscribers moved and 0- and 1-star did not, that is a useful finding about where to spend reactivation effort next time.
• Act on the outcome. Subscribers who did not move after a well-constructed attempt have now been tested, and that strengthens the case for whatever your suppression policy calls for next.

---

Section IV — Ongoing List Management

⚠️ Draft note for AudiencePoint — please review before publication. The four scenarios in this section depend on two behaviors that are not yet documented and were not confirmed at the time of drafting:

1. What a subscriber becomes when verified activity passes 720 days. These scenarios are written to the assumption that they fall back to Unmatched, which is the more cautious of the possible answers. If instead they remain at 0 indefinitely, the aged-out/never-scored distinction below is unnecessary and those passages can be cut cleanly.
2. How Unmatched and Invalid are represented in the data written back to your ESP — distinct field values, nulls, or separate flags. The Invalid scenario is written generically as a result.

Passages contingent on these are marked inline.

Your score distribution is declining across the program

Situation: Tracked over months, your program’s average Spruce score is trending down.

What It May Indicate: In most cases, nothing is wrong. A score can only fall as time passes unless new verified activity is recorded. A subscriber who engaged 100 days ago scores 3; the same subscriber, having done nothing further, scores 2 at day 240 and 1 at day 360. This means a fixed set of subscribers watched over time will show a declining average by construction. It is arithmetic, not a finding.

The average becomes meaningful only when read against what is happening to list composition. A program acquiring healthy new subscribers offsets the decay. A program that has slowed acquisition, or is acquiring lower-quality subscribers, will show the decline plainly. The number is telling you about the shape of your list, not the quality of your email.

Questions to Investigate

• Acquisition volume over the same period — has it slowed or stopped?
• The distribution of new cohorts entering the program, compared to earlier cohorts
• Whether the decline appears within individual cohorts, or only in the blended average
• Suppression activity over the period, which mechanically raises the average when it removes low-scoring subscribers
• Whether your own engagement data shows a comparable trend

Actions to Consider

• Read cohorts, not the blended average. Compare subscribers acquired in a given month against subscribers acquired in the same month a year earlier, at the same age. This is the only comparison that isolates quality from aging.
• Check acquisition volume first. A declining average alongside slowed acquisition is expected and needs no remediation beyond acquisition.
• Compare new-cohort distributions over time. If newly acquired subscribers are entering at lower tiers than they used to, that is a real finding and it points upstream to acquisition.
• Do not respond by suppressing the low tiers to raise the number. That improves the metric without improving the program, and it removes subscribers who may be recoverable.
• Set a review cadence and hold to it. Monthly or quarterly cohort review will show real movement; watching the blended average weekly will show noise.

---

The engaged-elsewhere segment

Situation: A meaningful share of your file scores 4 or 5 while showing little or no engagement in your own program.

What It May Indicate: These subscribers are actively engaging with marketing email. They are not engaging with yours. That is a distinction your own data cannot make, and it changes the conclusion entirely — this is not a dormancy problem, it is a competitive one. The likely causes are relevance, cadence, or reachability, in roughly that order of frequency.

This is typically the highest-yield population on the file, and it is routinely mishandled because standard engagement rules classify them as inactive and route them toward suppression.

Questions to Investigate

• How large the segment is, and whether it concentrates in a source, cohort, or mailbox provider
• Inbox Placement for the segment — the first thing to rule out, since a Spam designation explains the non-engagement completely
• What content and cadence they currently receive, and whether it matches what they signed up for
• How long they have been inactive with you, and whether there was an identifiable point when it started
• Whether their acquisition source suggests a mismatch between what was promised and what is being sent

Actions to Consider

• Rule out placement before assuming relevance. A high-Spruce subscriber designated Spam is an attentive person your mail is not reaching. That is a deliverability fix, not a content fix, and it is the single most recoverable situation in this playbook.
• Where placement is fine, treat it as a relevance problem. These subscribers open other brands’ mail. Something about yours is not earning the open.
• Test cadence in both directions. Over-mailing and under-mailing both produce this pattern, and the answer differs by program.
• Give them a distinct reactivation track rather than folding them into a general win-back. Their profile is completely different from a genuinely dormant subscriber and the same message will not serve both.
• Exempt them from routine suppression rules. Any policy that suppresses on your own engagement data alone will remove this segment, and it is the last population you should be removing.

---

One segment’s distribution diverges from the rest

Situation: A particular audience segment consistently shows a materially different score distribution than your other segments.

What It May Indicate: Divergence usually reflects how the segment is defined rather than anything about how its members behave. A segment built on lifecycle stage, acquisition source, product interest, or an existing engagement rule will inherit the score profile of whatever it selected for. A segment defined as “has not purchased in a year” will naturally skew low, and that is a description of the definition, not a discovery.

Questions to Investigate

• How the segment is defined, and what its members have in common
• Average age of subscribers in the segment relative to others
• Acquisition source concentration within the segment
• Content and cadence the segment receives
• Whether the divergence is stable or widening

Actions to Consider

• Check the definition before investigating behavior. In most cases the segmentation logic explains the distribution, and there is nothing further to find.
• Where the definition does not explain it, look at cohort age and acquisition source next.
• Adjust content and cadence for the segment where the divergence suggests a genuine mismatch of interest.
• Be cautious with mailbox-provider divergence specifically. Because Spruce is cross-brand, a lower distribution at one provider generally reflects the demographics of who uses that provider, not a deliverability problem with that provider. Use Inbox Placement for provider-level deliverability questions; Spruce is the wrong instrument for it.
• Watch the trend rather than the snapshot. A stable difference is a characteristic; a widening one is a change worth explaining.

---

A large share of the list reads Unmatched

Situation: A substantial proportion of your file returns Unmatched rather than a score.

What It May Indicate: Unmatched means insufficient behavioral evidence exists in the pool to produce a score. Some Unmatched is normal in every program. Higher rates are expected with a newer integration that supplied no historical interaction data, with newer subscriber cohorts, and with B2B, regional, international, or highly specialized audiences where pool coverage is naturally thinner.

A distinction to confirm. If subscribers whose last verified activity passes 720 days fall back to Unmatched, then this bucket holds two populations with opposite implications: subscribers who have never been scored (genuinely unknown, often your newest, and not suppression candidates), and subscribers who were scored and have aged out (your deepest-dormant population, and among your most defensible suppression candidates).

Score history separates them. A subscriber who previously read 0 or 1 and now reads Unmatched has aged out. A subscriber who has only ever read Unmatched has never been scored. Retaining prior scores is what makes this distinction available — a Score Change Tracking automation preserves that history automatically. Contact support@audiencepoint.com to set one up.

Questions to Investigate

• Age profile of the affected subscribers
• Whether Unmatched concentrates in a specific source, segment, or provider
• Whether coverage is improving across successive refreshes
• Whether engagement data is reaching AudiencePoint, if the rate is very high across the whole file
• If you deliver by FTP or SFTP, whether the file has been refreshed recently

Actions to Consider

• Continue applying your standard audience management practices to Unmatched subscribers. Withholding mail prevents the accumulation of behavioral data and makes the condition permanent.
• Monitor coverage growth over time rather than treating a single reading as a finding.
• Use engagement, consent status, and organizational policy to make cadence decisions for this population in the interim. Some organizations apply a reduced cadence to very large Unmatched populations; others continue standard sending until data accumulates. Both are defensible.
• Where score history is available, separate never-scored from aged-out subscribers before making any suppression decision that touches this bucket.
• If the rate is very high across the entire file rather than a segment, treat it as an integration question and confirm data flow to AudiencePoint.
• Do not treat Unmatched as equivalent to 0. It is the absence of a score, not the bottom of the scale.

---

Designing or revising suppression policy

Situation: You are establishing or revisiting the rules that determine when a subscriber stops receiving mail.

What It May Indicate: Suppression is where reachability, attention, business value, and compliance intersect, and it is the decision most likely to be made badly with a single metric. Spruce contributes a dimension your own data cannot supply — whether a subscriber is engaging with email at all — but it is one input among several and never a trigger on its own.

Questions to Investigate

• Current suppression criteria, and what they are actually built on
• How “engaged” is defined internally, and whether that definition survives the MPP question above
• How Spruce, Inbox Placement, and first-party engagement interact in the current rules
• How Unmatched is treated today
• Business value, customer status, and lifetime value of the populations the rules would affect
• Consent basis, jurisdiction, and any regulatory constraints on retention or removal

Actions to Consider

• Never suppress on Spruce alone. A low score means low recency of engagement with email generally. It does not mean low value. A customer who transacts by phone, a B2B contact who reads without clicking, or a high-value account with a quiet inbox can all score low and still warrant mail.
• Require agreement across signals before suppressing. The defensible case is a subscriber who is unengaged in your program, low-scoring in the pool, and either unreachable or persistently Spam-designated. Any one alone is insufficient.
• Exclude Unmatched from suppression logic. It reflects missing data, not evidence of dormancy. (If aged-out subscribers fall back to Unmatched, this needs a carve-out: never-scored subscribers stay excluded, while subscribers with score history showing they aged out may be treated as the deepest-dormant tier. See the Unmatched scenario above.)
• Attempt re-engagement before suppression wherever the subscriber is still reachable. A subscriber with an Inbox designation and a low score is a win-back candidate; one designated Spam with a low score is not.
• Build tiers rather than a binary. Full cadence for the engaged, reduced cadence for the slipping, a re-engagement track for the dormant, and suppression reserved for those who fail the re-engagement track.
• Separate suppression from loss when you socialize the policy internally. Holding a 1-star subscriber out of a major campaign is not the same as losing them — it protects the deliverability of every other message you send while you decide what to do with them.
• Let business value and consent requirements sit above the score, and handle those cohorts deliberately rather than by rule.
• Review the policy on a set cadence. Audience composition, provider behavior, and business objectives all shift, and thresholds set once quietly become wrong.

---

Handling Invalid

Situation: Your analysis returns addresses designated Invalid.

What It May Indicate: The address cannot be reached. It may be syntactically broken, a known spam trap, an abandoned inbox, or otherwise undeliverable. Unlike every other value in this playbook, this one carries no ambiguity about attention or intent — there is no one there.

Questions to Investigate

• How many, and what share of the file
• Whether they concentrate in one acquisition source, which points at a collection problem
• Whether validation happens at the point of collection today
• Whether any have been receiving mail, and for how long

Actions to Consider

• Remove them. This is the one place in this playbook where a near-automatic action is appropriate. Continuing to mail undeliverable addresses generates bounces and, in the case of spam traps, direct reputation damage.
• Trace concentrations back to their source. A single form or partner producing most of your Invalid addresses is an acquisition problem that will keep producing them.
• Add validation at the point of collection so these addresses stop entering the program.
• Do not confuse Invalid with a low score. A 0 is a real person who has not engaged recently. An Invalid address is not a person at all, and the two warrant completely different handling.

---

Operational Principles

• Read distributions, not individual scores.
• Compare cohorts of similar age. Recency always favors the younger cohort.
• A declining average on a list you are not growing is arithmetic, not a signal.
• Check coverage before drawing a conclusion from any distribution.
• Expect Spruce and Inbox Placement to disagree. Divergence is the informative case, not an error.
• Spruce describes the subscriber. Your engagement data describes the relationship. Decisions need both.
• Unmatched is the absence of a score, not the bottom of the scale.
• Never suppress on the score alone.
• Investigate before making major operational changes, and look for trends rather than reacting to a single refresh.
• Use the score to shape future campaigns, not only to explain past ones. The greatest value comes from building subscriber-level attention data into ongoing audience planning, cadence, and list management.
• Every organization’s threshold is different. Business value, consent basis, and risk tolerance sit above the score.