Internal · Reasoning Diff · Not part of the product experience

Strong acquisition, failing retention

Why Aegis reached a different conclusion than the benchmark. This page collects evidence from one failed scenario. It does not create reasoning rules — those come only after the same failure appears across many scenarios.

Back to labDisagreementNew customers arrive efficiently but do not stay.
1

Belief comparison

Expected belief

Acquisition is working and retention is cancelling it out: signups +75% at falling CAC while the customer base is flat and churn doubles.

Confidence · high

Aegis belief

The business is defined by the relationship between Churn Rate · CAC, not by any one line: Churn Rate and CAC are pulling apart, so the period's result is a net of two opposing forces rather than a single trend.

Confidence · high

Aegis did not reach the expected dynamic — the conclusion rests on a different relationship.

2

Data sensing

Did Aegis look at the wrong evidence?

Benchmark prioritized

  • Churn Rate

    deteriorating · importance 100 · seen in 4 periods · established-pattern

  • Customers

    volatile · importance 53 · seen in 4 periods · one-off

  • Trial Signups

    improving · importance 74 · seen in 4 periods · established-pattern

Aegis prioritized

  • Churn Rate

    deteriorating · importance 100 · seen in 4 periods · established-pattern

  • CAC

    improving · importance 67 · seen in 4 periods · established-pattern

  • Customers

    volatile · importance 53 · seen in 4 periods · one-off

  • Trial Signups

    improving · importance 74 · seen in 4 periods · established-pattern

  • Monthly Revenue

    flat · importance 60 · seen in 4 periods · emerging-pattern

Ignored or underweighted

Nothing expected was missed.

Carried in addition: CAC, Monthly Revenue.

No — every expected metric was read, but the conclusion also carried metrics the benchmark treats as secondary.

3

Mechanism misalignment

Why did Aegis reach a different conclusion?

Expected business explanation

Acquisition is working and retention is cancelling it out: signups +75% at falling CAC while the customer base is flat and churn doubles.

Aegis business explanation

Churn Rate moved +22.7% (6.6% → 8.1%) while cac moved −4.8% ($126 → $120). Churn Rate and CAC are pulling apart, so the period's result is a net of two opposing forces rather than a single trend.

Where it diverged

Aegis had the right evidence in hand but did not connect it: churn-vs-revenue, funnel-quality was not surfaced, so the conclusion stopped at description rather than dynamic.

4

Decision impact

Expected recommendation

act-now

Aegis recommendation

act-now

The reasoning difference did not change the business decision — both readings resolve to "act-now". Only the explanation behind it differs, which affects trust in the review rather than the action taken.

5

Engine lesson

Failure type

Unformed relationship

Observed pattern

CAC vs Monthly Revenue

Category

The right metrics were read, but the relationship between them was not formed.

Root cause

Aegis held every metric the conclusion needed but stopped at description. churn-vs-revenue, funnel-quality required the metrics to be read against each other, and each was explained on its own instead.

Supporting evidence

  • CAC

    improving · importance 67 · seen in 4 periods · established-pattern

  • Monthly Revenue

    flat · importance 60 · seen in 4 periods · emerging-pattern

  • Churn Rate

    deteriorating · importance 100 · seen in 4 periods · established-pattern

  • Customers

    volatile · importance 53 · seen in 4 periods · one-off

Business impact

The recommendation survived — both readings resolve to "act-now" — but it was reached for the wrong reason, which costs trust in the review rather than the action.

Candidate principle

None. A single scenario is not evidence. This observation is filed to Judgment Memory and a principle is proposed only once the same behaviour recurs across scenarios.

Status · Evidence CollectionSee accumulated evidence for this pattern

Suggested improvement

Capture the exact readings for churn_rate, customers, trial_signups and the threshold each pattern rule needed. If the same near-miss recurs, the gap is in the rule's tolerance, not in the data.

Notes

  • Scenario family: acquisition-quality. Posture read as mixed; benchmark expected mixed.
  • Evidence overlap 100% · belief 0% · decision 100% · confidence 100%.
  • Evidence only. No reasoning rule should be changed from this single scenario.