Romb understands how your organization makes sense of a problem, works it through, and decides. Then it reasons with you, from work your teams have already done.
RombYes, with the price phased back in. Last spring's cohort left at the first full-price bill.Growth, Sep 18Pricing, Jun 2Finance, Jul 9
Every point is something this organization worked out.
Example. Asked: should we run the spring promo again? Romb reasons across what Growth, Pricing and Finance have already worked out and answers: yes, with the price phased back in, because last spring's cohort left at the first full-price bill. Each source is cited.
Not long ago, data was a by-product of doing business. Then companies gave it infrastructure, a discipline, and a chief officer.
What teams learn from that data is worth even more, and it never got the same treatment. The findings, the reasons behind each decision, the way each problem was worked through: they still live in threads, decks, and the heads of whoever was in the room.
Romb is the infrastructure for it. An engine that learns how your organization reasons, and puts that reasoning to work for everyone, and everything, that needs it.
growthSep 18
PSPriya Shah
Q3 retention readout is done. Short version: it was the spring promo cohort.
What
The Q3 retention dip came from the spring promo cohort.
Romb
DATA-2140 closed. Drafting the insight from the work
drafted from
#growth threadDATA-2140retention_q3.ipynb
EditPriya Shah
DATA-2140
Q3 retention dip: seasonal or promo?
Priya Shah, Growth
retention_q3.ipynb#growth thread
Why
They reached their first full-price bill in July. No other cohort moved.
retention_q3.ipynb
cohorts.retention()
cohorts by signup month
How
Cohorts by signup month, split by channel, reactivations excluded.
from the #growth threadfrom DATA-2140from retention_q3.ipynb
how it works
Every answer starts as someone's work.
The first answer on this page cited Growth, Sep 18. That day, Priya Shah finished two weeks of work on why retention dipped: the answer posted in a thread, the ticket closed, the notebook saved.
In most organizations, this is where it ends.
Finish the work. Romb does the rest.
No write-up, no form. Romb notices the work is done and drafts the insight from it: the thread, the ticket, the notebook. Priya reads the draft and confirms.
Now it has depth.
Every insight keeps three things together.
what
The Q3 retention dip came from the spring promo cohort.
why
They reached their first full-price bill in July. No other cohort moved.
how
Cohorts by signup month, split by channel, reactivations excluded.
Most tools keep the what, at best. Romb keeps all three, so the next person inherits the reasoning, not only the result.
Then it takes its place.
One point in the record of what your teams have worked out. The same thing is happening wherever work gets finished: a ticket closes, a test reads out, a thread lands on an answer.
Romb also reads back through work that is already done, so the record is full from the first week.
Connected to everything it touches.
Each insight is linked to what it builds on, what it replaced, and where another team saw it differently. What forms is more than a store of findings. It is how your organization reasons: what it checks first, which approaches hold up, where judgment was needed.
what was learned
what was decided
the approach that got there
how one bears on another
Above it, a system of agents.
They work around the clock, so the reasoning is ready before the question is. People confirm what goes in.
Your company's best thinking, in every seat.
Romb answers where the question comes up: in the channel, in the brief, before the meeting, as new work starts. What comes back depends on what the moment needs.
when the question has been settled before
productsix weeks later
DK
Dana Kowalski
Has anyone looked at why retention dipped in Q3? Board prep is Thursday.
Rombapponly visible to you
Yes. Growth traced it to the spring promo cohort reaching its first full-price bill in July. No other cohort moved.
Q3 retention dipPriya Shah, Sep 18
Decided: model promo cohorts separatelyGrowth, Sep 22
Promo cohort analysisapproach, used four times
See the reasoningShare in the channel
Message #product
Ask RombHelen Park, COO
Why is enterprise churn up when the Q2 pricing fix worked?
The reasoningacross three teams
because
The fix cleared up tier confusion for new contracts only. Pricing, May 30
and
Renewals signed before May still carry the old tiers. Sales Ops, Jun 12
but
The confusion comes from two of those tiers, not all five. Support, Jun 30
so
The fix worked where it applied. Extending it to those two tiers at renewal is the open decision.
Which two tiers?Who owns the decision?Show the renewals at risk
InboxMonday, 8:00 AM
Rombto Lena Ortiz, Growth
Monday brief: three things for Growth this week
new this week, not yet worked on
Activation in EMEA fell 9%.
Three hypothesesranked by what caused it before
1
A step added to onboarding on the 14th.The cause in 3 of 7 past drops.
2
The payment provider change in two markets.Flagged by Finance on Sep 30.
3
School holidays.Ruled out twice before.
First check: cohorts by signup date. The approach that settled it last time.
Spring promo: the price step-up decision is due Friday
Pricing and Support read the tier confusion differently
STRAT-118To doTomás Rivera, Strategy
Forecast demand for the Denver launch
Size weeks one to eight for both Denver zones.
opened
just now
due
Oct 30
labels
launch, forecast
Before you startthe approach that held up
Austin launch playbook
1
Start from comparable cities, not the national average.py
comps = similar_cities("Denver", k=3)
base = demand.loc[comps].groupby("week").mean()
2
Count reactivations separately. They get counted twice.sql
select week, is_new, count(*) as customers
from orders
group by week, is_new -- reactivated apart
3
Trust week four over week one.py
# week one runs hot on the launch promo
fit = model.fit(base[base.week >= 4])
Used four times. Last by Marcus Lee, Aug 4.
further in
Answers are just the beginning.
Asking is the smallest thing you can do with a reasoning engine.
It starts before you ask.A pattern emerging across teams. A gap in what is known. A number about to move. It forms the hypotheses before anyone asks the question.
It executes complex work, end to end.It constructs the methodology from the context, approaches and playbooks your teams have already built, carries it out, and verifies the result. Where it is not certain, it stops and asks.
It is where agents come to reason.Build your own on it. Bring the rest. Every agent then reasons from the same ground, and writes back what it learns.