Field Notes
A research debrief, vol. iii

What we learned this quarter.

A short, honest look at what our customers told us between January and March — what's working, what's quietly broken, and what we want to try next.
Foreword
ii

We spent eight weeks listening, and what we heard surprised us in the kindest way.

The team ran twenty-eight long-form interviews, shadowed nine teams during their busiest week of the year, and sat with the support inbox for ten unbroken days.

The themes that emerged were not the ones we set out to find. The brief asked about onboarding; the answers we got were about trust. So we followed the thread.

This deck is the short version. Each insight is a door — open the ones that matter to your team this quarter, and let the others wait.

— The research desk
The Method
iii
i.

Listen

Twenty-eight long-form conversations with customers across four segments and three regions.

ii.

Watch

Nine on-site shadowing sessions during peak workflows. We took notes, not video.

iii.

Read

Ten days inside the support inbox, tagging every message by intent and emotional tone.

iv.

Distill

Three rounds of thematic clustering with the design and policy teams, ending in eight themes.

Insights
iv

Insight #1

Trust is the onboarding

Customers don't churn on day one because the product is hard. They churn because the first three emails feel like a stranger.

Insight #2

Power users dread upgrades

The people we asked to love new features the most quietly resent them. They want fewer surprises, not bigger ones.

Insight #3

Support is product

Half of "feature requests" in the inbox are existing features that customers couldn't find. We have a discovery problem dressed as a roadmap problem.

A closer look · 1 of 3
v
on insight #1

Trust is the onboarding.

The product can be perfect on day one, but if the welcome email reads like a contract, half of new accounts will never log in twice. The fix is small and writerly, not technical.

By the numbers
vi
of new accounts open the third email,
up from 41% last quarter.
68%
28
long-form customer interviews, across four segments.
9
teams we shadowed for their busiest week of the year.
In their words
vii
"
I did not need a better product. I needed it to behave like it remembered me.
Renée, three-year customer Studio of seven, Lisbon
What we'll do next
viii

Three small moves,
before the next debrief.

Each one is small enough to ship inside a sprint, and structured so we can tell, six weeks from now, whether it moved the dial.

i.

Rewrite the first three emails

From templated to written. Owner: lifecycle. By: May 17.

ii.

Quiet upgrades by default

Opt-in for power users; soft rollout for everyone else. Owner: product. By: June 1.

iii.

Make the inbox a search bar

Surface in-product help when a request matches a feature. Owner: support × product. By: June 14.

Findings · Detail
ix
Action title

The trust gap is built in the first 72 hours, not the first 7 days — and the cost compounds for the rest of the lifecycle.

What we found

Three behavioral signals in the first 72 hours predict 18-month retention better than any feature-usage metric we tracked.

  • Email open #2 — opening the second lifecycle email lifts D90 retention by 19 points.
  • Personal salutation — accounts that received a written, non-templated welcome retained 2.4× the cohort.
  • Reply received — a single human reply within 24 hours of signup is the single largest lever we measured.
Source: 14,200 cohorted accounts, Jan–Mar 2026.

Why it matters

$4.1M in projected retained ARR, on the current cohort alone.

The first three days are the only window where customers are both paying attention and willing to write back. Every interaction during this window does the work of roughly four interactions in week three.

The cost of getting this wrong is not refunds — it is the long, quiet churn of an account that never returns to the inbox.

Modelled on FY24 cohort behaviour.

What to do

  • Rewrite emails 1–3 in human voice; ship behind a 50/50 holdout. Owner: lifecycle. By: May 17.
  • Route every signup to a named human for one personal reply within 24h, capped at the top 200 accounts/day. Owner: success. By: May 24.
  • Instrument the 72-hour window as a first-class metric in the weekly review. Owner: analytics. By: June 1.
Pilot scope: top-decile signups by ICP score.
Retention, by cohort
x

The curve bends
around day three.

Cohorts that received a written welcome and a human reply within 24 hours retain at roughly 2× the rate of the templated cohort, and the gap holds steady for the first ninety days.

Templated welcome (control)
Written welcome
Written + human reply
% of cohort active, by day
1007550250
D0D7D14D30D45D60D90
How we'll work
xi

From insight
to shipped change.

A five-step path each pilot follows, end to end, before it is allowed to graduate to the default experience.
i.

Frame

Translate the insight into a single behavioural hypothesis we can falsify.

ii.

Design

Sketch the smallest end-to-end change that lets the hypothesis be tested in one sprint.

iii.

Pilot

Ship to a 50/50 holdout in a single segment. Hold the line for two full cycles.

iv.

Read

Review the cohort against pre-registered metrics. Decide kill, scale, or extend.

v.

Default

Graduate the change to the default surface. Retire the legacy path inside the same release.

Week 1Weeks 2–3Weeks 3–6Week 7Week 8
The three pilots, side by side
xii

Where each pilot
earns its keep.

Scored against the four levers that matter most this cycle. We will only carry forward bets that win on at least two.
Lever
Rewrite welcome
Quiet upgrades
Inbox-as-search
Time-to-impact
≤ 4 weeks
6–8 weeks
≤ 4 weeks
Build cost
Low
Medium
Low
Retention lift (model)
+19 pts D90
+7 pts D90
+5 pts D90
Risk to power users
None
Material
Soft, reversible