Funnel diagnosis lagged the business.
When checkout conversion dipped, the analyst cycle to find the cause took the better part of a week. By the time the answer arrived, the team was already debugging the next thing.
+6pp
D30 retention lift on reactivated cohorts (41% → 47%)
View outcome below₹4.2crCAC efficiency reclaimed in Q1
View outcome below<1hrMedian time-to-insight on funnel questions, down from 3–5 days
View outcome below
- Client
- A leading Indian consumer health platform
- Industry
- Consumer healthtech, diagnostics, telehealth
- Registered users
- 6M
- Monthly active users
- 1.4M
- Cities
- 80+
- ARR
- ~₹220cr
- Paid spend
- ~₹28 cr/year across Meta, Google, affiliates
- Agents deployed
- Growth Ops Agents
- Deployment
- 6 weeks from kickoff to all workflows live
The situation
The data was there. The operating tempo was not.
The growth team ran a data-rich operation: Mixpanel for product, Branch and AppsFlyer for attribution, Snowflake as the warehouse, and eight ad accounts across Meta, Google, and three affiliate networks. The team was 14-strong, including three embedded analysts.
Despite that, the gap between a growth question being asked on Monday and a defensible answer landing was three to five days. Decisions that should have been weekly slipped to monthly.
It wasn’t a tooling gap. It was a workflow gap.
Where the workflow broke down
Creative fatigue was always called too late.
ROAS would dip for 7–10 days before anyone noticed. By then, real budget had been burned.
Reactivation was a generic email blast.
Dormant users all got the same offer. Conversion sat at 0.4%, and the team had no reliable way to score who was actually reactivatable.
Analysts were the bottleneck on everything.
Every campaign waited on a custom audience cut. Every funnel question waited on a query.
The deployment
Six weeks. Fourteen hours of client engineering.
Six weeks from kickoff to all three workflows live, read-only. No data left the client warehouse.
Engineering effort on the client side: 14 hours, mostly access provisioning and one workshop on definitions.
Three workflows.
One connected platform.
Trigger
Daily at 7:30 a.m., plus on-demand via Slack
Data sources
Mixpanel events, Branch attribution, internal booking DB, payment processor
How it runs
Pulls the last 24 hours of funnel and compares to 7-day and 30-day baselines. Flags drop-offs >3σ from baseline. Traces flagged steps to source: channel, creative ID, device, city, time-of-day. Writes a 4-sentence root-cause summary.
Human in the loop
Growth lead reviews in the 9 a.m. standup. One click expands the underlying SQL trace.
Output
Slack #growth-daily by 7:45 a.m., plus a saved analysis link.
Before
Three-to-five day analyst cycle, often arriving after the next dip had started.
After
Same-morning answer with a named root cause. Standups go straight to action.
Trigger
Continuous on Meta and Google, with 4-hour Slack-cadence alerts
Data sources
Meta Ads, Google Ads, post-impression conversion data, Mixpanel
How it runs
Calculates rolling ROAS per ad set and creative. Detects when CTR or CVR has dropped >15% over a 5-day rolling window, with a 95% significance test before firing. Separates creative fatigue from audience fatigue. Suggests next-best lookalike cohorts.
Human in the loop
Performance marketer reviews the alert, confirms a pause or creative refresh.
Output
Slack alert with one-click pause. Daily creative-fatigue digest to the head of growth.
Before
Weekly creative review, often a week behind the actual fatigue signal.
After
Fatigue called within 48 hours of statistically significant signal, with the next creative already briefed.
Trigger
Weekly batch every Monday morning
Data sources
User behavioural data, last test booked, propensity model (rebuilt monthly), CRM communication history
How it runs
Identifies dormant users (no booking 90–180 days). Scores reactivation probability. Top decile (~2,400 users/week) is routed to the voice agent; rest go to email and WhatsApp. Voice agent dials with personalised openings, surfaces incentives, books appointments if intent is detected. Conversations are transcribed and tagged.
Human in the loop
Growth ops reviews a 5% sample of voice calls weekly for quality and compliance.
Output
Direct booking, follow-up call tagged in CRM, model retrain signal.
Before
Generic reactivation email at 0.4% conversion.
After
Voice-led top decile at 4.2% conversion. Roughly 180 incremental bookings per week from a population that previously did nothing.
Outcomes
+6pp
D30 Retention on Reactivated Cohorts
41% to 47%, sustained across three monthly cohorts.
Before41%
After47%
- Sustained across
- 3 monthly cohorts
Customer quote
“Our growth team stopped being the bottleneck. The decisions we used to make on Friday based on Monday data, we now make on Tuesday morning. The compounding from that is real.”
VP Growth
From kickoff to production
Week 1
Scoping workshop with growth leadership. Three workflows and the voice agent locked.
Weeks 2–3
Read-only connections to Snowflake, Mixpanel, Branch, ad accounts. Definition layer built with the analyst lead.
Week 4
Funnel Diagnosis and ROAS workflows live.
Week 6
Voice agent live for reactivation. All three workflows in production.
Common questions
How long does a deployment like this take?
Six weeks for this engagement. Most growth-team deployments land in the four-to-eight week range, depending on the number of workflows and the state of the underlying data definitions.
What engineering effort is required on the client side?
Fourteen hours total for the Growth Ops deployment, mostly access provisioning and one workshop. The Actioneer team handles the build.
Does any data leave the client environment?
No. Actioneer connects read-only to the systems already in place. Nothing is ingested, copied, or relocated.
What does the voice workflow actually do?
The voice agent dials the top decile of reactivatable users with personalised openings, surfaces a relevant incentive, and books a return appointment if intent is detected. Conversations are transcribed and tagged; a 5% sample is reviewed weekly for quality and compliance.
How are the numbers audited?
Against a pre-deployment baseline agreed with the client. All headline numbers ship with traceable underlying queries.
Can the workflows be modified after launch?
Yes. Workflows are continuously refined post-launch. Triggers, thresholds, and approval gates can all be edited without re-deployment.
