TriHedgehog Book a demo Get a demo
Adaptive retention

Turn customerbehavior into morerepeat revenueand less churn.

TriHedgehog chooses and writes the best message for each customer, then learns from every order, booking or renewal.

15-minute tailored demo · No sales theatre

See one decisionDecision ready

Your customers are individuals. Your flows are not.

MP
Mina P.First order 23 days ago
High intent
✓ Kept her dress 2× Viewed shirt → Left checkout
TRIHEDGEHOG CHOOSES Show the matching shirt in her saved size.
Send at 12:40No discount
NNorth & Loomto Mina

SUBJECTThe shirt that works with your navy dress

Complete the look →

TriHedgehog starts with the business outcome, then chooses and creates the next message for each customer.

How it works

See the person behind the event stream.

  1. 01

    Connect the signals

    Orders, visits, billing and messages from the tools you already use.

  2. 02

    Choose the next move

    Send, wait, offer or stay quiet—customer by customer.

  3. 03

    Learn from outcomes

    Purchases, bookings and renewals improve the next decision.

See it with your journey →
Thinking

How we think about retention.

8 MIN READ

The hidden cost of sending everyone 15% off

Read the article → Browse all thinking →
Integrations

Add a decision layer. Keep everything else.

Connect the stack you already use. No migration required.

Shopify Stripe Mixpanel PostHog Mailchimp HubSpot

Start with one source and one sender. No migration.

You stay in control

Secure decisions you can inspect.

Decision #18,432● Approved
Why this message?

Mina returned to a matching item and chose her usual size. There is no price signal, so no discount was used.

✓ Consent checked✓ Frequency checked✓ Decision logged
Approval firstReview every message before it sends.
Hard guardrailsYou set claims, offers and limits.
Your data stays yoursNever used to train shared models.
Quick answers

Before you ask.

Does this replace Klaviyo or Braze?+

No. TriHedgehog adds the decision layer and can send through the tools you already use.

Do we approve the messages?+

Yes. Start in approval mode, set firm rules and automate only when you are comfortable.

What does a pilot involve?+

One customer moment, one data source, one sender and one measurable goal over 30 days.

Book a demo

Automatically choose, write and improve each customer’s next message.

Share your work email. We will return with a short, specific demo.

30-day pilotNo migrationNo improvement, no charge

Turn customerbehavior into morerepeat revenueand less churn.

TriHedgehog connects to your store and email tools. It chooses the best message for each customer, writes it in your voice and learns from every order, booking or renewal.

A short demo, tailored to your customer journey. No sales theatre.

AS
Aisha S.Next coffee delivery in 8 days
79% intent
✓

Rated first coffee 5 stars

●

Viewed the decaf blend three times

Opened delivery settings

Decision

Send now. Offer a one-click blend swap before the next box is packed.

Aisha likes the product but is actively considering a different blend. A simple switch protects the subscription without a discount or cancellation message.

Aisha, you have been looking at Decaf. Switch your next delivery in one tap, or keep your usual blend. Nothing else changes.

Your customers are individuals.Your flows are not.

Traditional journey builders make your team define every segment, trigger and branch. TriHedgehog starts with the business outcome, then chooses and creates the next message for each customer.

The old way

A forest of rules

Flow: Post-purchase v12Last edited 43 days ago
CONDITIONAL SPLITOrdered before?
TRIGGEROrder completedFilters: 3
CONDITIONAL SPLITHas ordered before?YES　　　　NO
TIME DELAYWait 7 days
SEGMENTHigh intent?Definition stale
EMAILPost-purchase #2Manual content
EMAILOffer 15%?Manual content
DROP NEXT RULE HERE
  • Every edge case becomes another branch to maintain
  • Segments drift as customer behaviour changes
  • A/B tests pick one average winner for everyone
The TriHedgehog loop

One goal.
Continuous decisions.

GOALIncrease memberships
JD
JACOB’S MOMENTThird day pass in 11 days
TRIHEDGEHOG CHOOSESShow his real savingsAfter tonight’s visit · No offer
Joined→The next decision gets better↻
  • One decision for each customer’s real context
  • Message, content, timing or silence chosen automatically
  • Business outcomes improve the next decision

See the person behind the event stream.

A segment says “one-time buyer.” TriHedgehog sees the exact sequence, the likely interruption and the next action worth testing.

32 DAYS AGO
ORDER COMPLETEDBought navy linen dress

£86 · Size 10 · Instagram referral

✓ kept
TODAY
PRODUCT VIEWReturned to matching linen shirt

Second view in three days

CHECKOUTSelected size 10 and started checkout

Same size as her previous order

high intent
SESSION ENDEDLeft before payment

One minute after checkout began

possible interruption
RETURN VISITOpened the product again

Direct visit · Mobile

ObserveThe full sequence, not a segment label
DecideMessage, content, timing or silence
CreateA complete on-brand communication
LearnFrom purchase, booking, renewal or no action

How we think about retention.

Read all thinking ↗
WIN-BACK ECONOMICS

The hidden cost of sending everyone 15% off

How blanket win-back discounts subsidize customers who would have bought anyway.

8 min read → 30-DAY PLAYBOOK

From first order to second: a retention test

A concrete plan for a subscription brand to increase repeat orders without adding more fixed flows.

10 min read → BETTER EXPERIMENTS

Your A/B test found a winner. Who should get it?

Why an average winner cannot choose the right action for the customer in front of you.

7 min read →

Add a decision layer. Keep everything else.

Connect the stack you already use. TriHedgehog reads customer behaviour from your analytics, commerce and billing tools, then sends the chosen communication through your existing marketing platform. No migration required.

Signals
Shopify
Google Analytics
Mixpanel
PostHog
Stripe
Snowflake
TriHedgehogdecision loop
Actions
Mailchimp
HubSpot
SendGrid
Intercom
Your sender

Do not see your tool? We can start with a scheduled export or a small read-only connection.

Secure decisions you can inspect.

See exactly why a message was chosen, control what the system is allowed to do and keep customer data protected at every step.

Decision #18,432
● Approved
LR
Liv R.First-booking intervention
16:10 today
WHY THIS WAS CHOSEN

Liv keeps searching the same dates but has not messaged anyone. She needs a shorter choice.

  • Viewed four local profiles
  • Searched the same dates three times
  • No conversation started in 48 hours
  • Three suitable people are still available
APPROVED ACTIONCurated shortlist email

Send now · Three available matches · No offer

✓ Consent verified✓ Frequency 1 of 3✓ Approved claims only✓ 20% holdout assigned
Inputs, model version, content and outcome loggedView full audit trail →
01

Your data stays yours

We use the minimum fields needed for your goal. Customer data is never used to train shared models or improve another company’s decisions.

02

Hard rules come first

Consent, suppression lists, frequency, quiet hours and offer limits are checked before any model can make a decision.

03

Start with approval

Review the evidence, action and finished message before it sends. Move only proven, low-risk decisions to autopilot.

04

Every action is reconstructable

Inputs, reasoning, content version, delivery and outcome stay connected so your team can review what happened.

The useful details.

How is TriHedgehog different from Klaviyo, Braze or Customer.io?+

Their core model is a canvas: your team defines the segments, triggers, branches, timing, content and tests. TriHedgehog starts with the business outcome. It studies each customer’s behaviour, chooses and creates their next communication, then learns from whether they purchase, book or stay. It can send through those tools, but your team no longer has to manually encode every decision.

Does AI write every message from scratch?+

Not by default. We begin with approved claims, offers and message components. Verified customer facts are inserted deterministically. Generation stays within your brand rules, and high-risk content can always require approval.

How do you know the system caused an improvement?+

We agree one business metric and keep a randomized holdout. We report incremental bookings, orders, activations or retained customers, along with guardrails such as unsubscribes and refunds. Opens and clicks are diagnostic, not the final score.

Will I lose control of what customers receive?+

No. Start in approval mode. You choose eligible customers, available actions, allowed claims, frequency limits, incentives and the metric to optimize. You can pause the system at any time.

Which channels can you use?+

Email is the simplest starting point. The same decision can later reach SMS, push, in-app or a human task through your existing tools. We add channels only when they improve the outcome and the experience.

What does a first pilot look like?+

We choose one valuable moment, connect one event source and one sender, define guardrails, then run approved interventions against a holdout for 30 days. You choose the success metric. If we do not improve it, you pay nothing.

Automatically choose, write and improve each customer’s next message.

Share your work email. We will research your customer journey and return with a short, specific demo.

30-day focused pilotWorks with your stackNo improvement, no charge