Churn by active channel count
Observed churn in the following month.
The numbers behind the rates
One check per customer per eligible month.
| Channels | Observed churn | Churns / monthly checks |
|---|
Counts of four and above have small samples.
Which channels are associated with lower churn?
All service models · Zero-channel months excluded · February–July 2026. This ranking is independent of the filters above.
Ranked by observed churn difference. The comparison group publishes through other channels.
| Rank | Activated channel | Churn when active | Without it | Observed difference | Active customers |
|---|---|---|---|---|---|
| 1 | Instagram Reels | 0.0% | 10.5% | 10.5 pp lower | 10 |
| 2 | Google Ads Very low evidence | 0.0% | 7.8% | 7.8 pp lower | 3 |
| 3 | Meta Ads Very low evidence | 0.0% | 7.7% | 7.7 pp lower | 1 |
| 4= | LinkedIn Profile Very low evidence | 0.0% | 7.7% | 7.7 pp lower | 1 |
| 4= | TikTok Very low evidence | 0.0% | 7.7% | 7.7 pp lower | 1 |
| 6 | LinkedIn Page | 2.5% | 9.4% | 6.9 pp lower | 18 |
| 7 | Facebook Posts (Reels not separated) | 6.0% | 12.5% | 6.5 pp lower | 37 |
| 8 | Instagram Posts (excluding Reels) | 6.4% | 10.3% | 4.0 pp lower | 30 |
| 9 | Facebook Stories | 4.8% | 8.1% | 3.3 pp lower | 12 |
| 10 | Instagram Stories | 5.0% | 8.0% | 3.0 pp lower | 12 |
| 11 | Blog | 5.9% | 8.5% | 2.6 pp lower | 23 |
| 12 | 9.4% | 6.7% | 2.7 pp higher | 18 | |
| 13 | Article SEO | 16.7% | 6.5% | 10.2 pp higher | 15 |
| 14 | Product + collection SEO | 22.2% | 5.8% | 16.5 pp higher | 13 |
Small samples are flagged. Zero observed churn does not mean zero risk.
Facebook Reels: no confirmed active sample to calculate a rate. Facebook Posts still contains unclassified videos. Instagram posts and Reels are separated where the format is known; uncertain formats are excluded from the affected comparisons. Product and collection SEO are grouped; article SEO is separate.
Ranking sample sizes and method
The publishing sample includes 49 publishing customers, 157 monthly checks and 12 churn events. Customers can appear in several months and channel groups. The channel-count views use the same publishing sample. Reels and SEO subtypes are split further in this ranking.
Each row compares active and inactive months for one channel among publishing customers. Percentage-point differences are unadjusted associations, not individual risk predictions. Higher observed churn for SEO does not establish that SEO causes churn.
| Channel | Active: churns / checks | Without: churns / checks | Unknown-format checks excluded |
|---|---|---|---|
| Instagram Reels | 0 / 29 | 12 / 114 | 14 |
| Google Ads | 0 / 3 | 12 / 154 | 0 |
| Meta Ads | 0 / 2 | 12 / 155 | 0 |
| LinkedIn Profile | 0 / 1 | 12 / 156 | 0 |
| TikTok | 0 / 1 | 12 / 156 | 0 |
| LinkedIn Page | 1 / 40 | 11 / 117 | 0 |
| Facebook Posts (Reels not separated) | 7 / 117 | 5 / 40 | 0 |
| Instagram Posts (excluding Reels) | 6 / 94 | 6 / 58 | 5 |
| Facebook Stories | 1 / 21 | 11 / 136 | 0 |
| Instagram Stories | 1 / 20 | 11 / 137 | 0 |
| Blog | 3 / 51 | 9 / 106 | 0 |
| 5 / 53 | 7 / 104 | 0 | |
| Article SEO | 3 / 18 | 9 / 139 | 0 |
| Product + collection SEO | 4 / 18 | 8 / 139 | 0 |
What the zero-channel count actually means
Historical monthly checks, not customers inactive today.
| Outcome month | Classified at zero | Customers checked that month |
|---|
Zero means no recorded publication in the preceding 30 days. Monthly and six-month cumulative counts are different measures.
Monthly comparison
Outcome month, 2026. Bars compare the same two channel groups.
The same customers can appear in several months. These are repeated measurements, not six independent studies.
View monthly counts
| Month | 1–2 channels | 3 or more |
|---|
What counts as one channel?
At least one successful publication through Glowtify in the preceding 30 days. Connecting an account alone does not count.
Counted channels: Facebook Posts, Facebook Stories, Instagram Posts, Instagram Stories, LinkedIn Page, LinkedIn Profile, TikTok, Email, Blog, SEO, Meta Ads and Google Ads. Facebook and Instagram posts plus stories can therefore count as four channels. SEO activates when at least one SEO update is published; product, collection and article SEO updates share that channel. A newly published blog article counts as Blog; a separately published SEO update counts as SEO. GEO and analytics connections are excluded. Multiple accounts or workspaces within the same channel do not increase the count. Only publication types confirmed in the historical record count.
Explore the customers behind the analysis
Historical monthly records, not today's customer list or a current risk score. The service-model and zero-channel filters above also apply here.
| Customer | Published channels | Count | Outcome |
|---|
Methodology, evidence & limitations
How the rate is calculated
For each monthly baseline, count channels published to in the previous 30 days. Then check finance's churn ledger for the following month.
Monthly churn = recorded churn events ÷ eligible monthly customer records. One customer can contribute several records. The full sample, before the zero-channel filter, has 66 distinct customers, 244 monthly records and 27 churn events involving 26 customers; one left, returned and left again. By default, only monthly records with at least one active channel are included. A customer can remain included in active months even if their zero-channel months are excluded.
Each count has its own row in the table above. Group rates are observed associations, not individual risk predictions.
Exposure windows are exactly 30 days; outcome windows are calendar months of 28–31 days. The first outcome month is February and the last is July 2026.
What the evidence supports
Three or more active channels is a candidate retention signal, not evidence that adding a channel causes lower churn.
Customer fit, onboarding, tenure and a decision to leave can influence both usage and churn. Bulk SEO updates also differ from social posts in effort and impact.
Eligible customers have positive recognized platform or managed revenue in the preceding month's finance roster. Revenue is not independently verified cash collection. Missing revenue alone does not establish churn.
Unmapped customers, internal accounts and contradictory payment records are excluded. One unresolved refund-related month is excluded. Historical managed status comes from that month's managed-service revenue.
Sample sizes for the selected service model
| Channels | Monthly records | Churn events | Unique customers |
|---|
Customers can appear in several channel groups in different months. Group counts must not be added to obtain the total customer count.
Data note
Provider-confirmed emails count on their actual send date, including historical records still marked scheduled in the app. Some other delivery dates remain unverified, so treat the rates as estimates.
Sources
Publication history comes from BigQuery, deduplicated by workspace and task across the observed history. A publication requires a nonempty publication URL and published status; its date is the later of the first observed published status and the task's calendar timestamp. Confirmed Klaviyo Sent status and provider send time override the app classification when available. Churn comes from Client summary 31-07-2026.xlsx, January–July customer sheets and the CHURN & NEW ledger. Workspace mapping was reviewed against historical records.
Stories are identified by the task automation type; successful Facebook/Instagram destinations come from the publication response. LinkedIn response personId identifies profiles and organizationId identifies pages. Unspecified Meta publications do not add another channel alongside a known destination of the same format. GEO is not a publishing channel. Publishing evidence is re-queried from BigQuery weekly. Verified finance outcomes remain fixed through July 2026 until a newer reconciled finance cohort is supplied; a fresh publishing timestamp does not mean newer churn outcomes.