Cover1 / 20
→ next   ← back   S what to say   O map   F full screen
Prototype data is illustrative · sailthru-case-study.pages.dev
Sailthru
One engine. Two moves. Told in three parts.
Senior Product Manager case study  ·  Jyotishman Das
Press  →  to begin
PART 1 PART 2 PART 3 Understand it Trust its reporting Make it earn

Sailthru's job: know each reader, and act on it.

Everything I recommend today compounds that one strength.

ONE
Understand the platform
What each tool is for · where it wins and loses · what the real engine is
signals audience content send learn profile

It's a loop, not a toolbox.

Behaviour updates one profile · the profile drives everything · each send is a little sharper.

Campaigns Templates Lifecycle Audience Reporting interest + intent scored per reader, nightly

Strip the tools away. This is what's left.

My thesis, and I label it as mine. Built for content. Its stated centre for most of its history.

channel breadth · real-time → content depth → Klaviyo Braze Iterable Sailthru

Four platforms, four buyers.

Iterable is the real overlap. No pricing claims either way. Fit, not "we're better".

STRONG · interest personalization, media-shaped prediction WEAK · usability, fragmented reporting, narrower channels UNKNOWN · competitor AI parity, pricing, the owner's roadmap

Strong engine, dated suite, new owner.

The weaknesses are review perception, not measurement, and mostly pre-acquisition.

TWO
Make the reporting answerable
A governed AI answer layer. Not a chatbot, not more dashboards.
Campaignreport Deliverabilityview Segmentreport Client / devicebreakdown GooglePostmasteroutside Sailthru one answer

One question. Five reports. A spreadsheet.

The data is all there. The join between them is a person with a spreadsheet, and that's the gap.

Audience Delivered Opens Clicks Conversions Revenue Apple prefetch

A funnel. One layer that lies.

Apple Mail prefetches images, so reported opens are inflated. Not the same as a bot open.

AI interpret · explain VALIDATOR no number it didn't compute Deterministic every number

AI narrates. Services compute. A validator guards.

The model never emits a figure. When it can't answer, it says so.

−$7.5K revenue week over week, delivery flat click-to-open rate, 94% of the drop audience mix shift, small open rate, not a driver ~2% not attributed, may sit outside email 72% of opens on Apple so read the non-Apple cohort

"Why did revenue drop?" Here.

Live demo now → sailthru-case-study.pages.dev/#reports · ask it, open the evidence, ask a forecast and watch it decline.

PROVES a decomposition that sums every number traceable facts vs interpretation, split it refuses a forecast DOESN'T PROVE a live model writes good prose accuracy on messy real data adoption the 60-second / 90% targets

Honest about the edges.

The targets are proposals. A pilot with a baseline study tests them.

THREE
Turn the engine into revenue
A publisher-first Direct Sponsorship Workspace. A workflow, not a marketplace.
SPONSOR traffickedby hand priced onbroken opens shown toeveryone

Publishers already sell sponsorships. Badly.

beehiiv taught me sequencing and measurement, not what to build.

MARKETPLACE cold start duplicates LiveIntent DSP conflict needs ad-ops WORKFLOW FIRST

The obvious first move is the wrong one.

Workflow first. The marketplace is LATER, and a separate decision.

SPONSOR BLOCK 7 automated checks + editorial sign-off Performance report verified events, not opens

One block. One gate. A report they can defend.

Live demo now → sailthru-case-study.pages.dev/#campaigns

THE TEST pub 1 pub 2 pub 3 one newsletter each, 4 to 8 weeks ANY ONE ENDS IT · WRITTEN DOWN FIRST no demand editors reject deliverability harm nobody pays

Small, cheap, decisive.

Three publishers run it for a few weeks. Any one of these, agreed before we build, stops it.

THREE DECISIONS Engine = the profile-centred loop Reporting = a governed answer layer Monetization = a workflow not a chatbot. not a marketplace. TEST FIRST Is the reporting pain measured? Does the competitive read hold? Do publishers sell sponsorships? a baseline study + 4 to 6 interviews

Narrow and deep, applied three times.

Both prototypes are live. First step in the role: interviews and a reporting baseline.

Questions.
sailthru-case-study.pages.dev
#reports · #campaigns · #logic
What to say

Jump to a scene  ·  press O to close