Loop or Funnel? Forecasting Whether Swiggy Instamart's Growth Holds
Objective: Read four weeks of real funnel data stage by stage, the way a growth analyst would, and forecast whether the acquisition engine is compounding like a loop or quietly draining like a leaky funnel, before the trend shows up in the topline number anyone in the room is watching.
You're the first growth hire supporting a quick-commerce vertical built on the Swiggy Instamart model during a stretch leadership is calling 'healthy and flat.' Weekly signups have barely moved for a month, which the team reads as stability. You've been handed the raw funnel export and asked for a one-slide verdict: is this actually stable, or is something underneath about to break?
The topline number (signups) is nearly flat across four weeks, that part is true. Your job is to look one layer beneath it, at the stage-to-stage conversion rates, and decide whether the engine underneath a flat topline is holding steady or eroding, using the lesson's own loop-vs-funnel and outputs-vs-outcomes lens.
Before you start
What you'll need
Free path (everything below is enough to finish)
Free, and the entire step-1 and step-2 tables above can be built in it (or any spreadsheet) from the raw export.
No access? Any spreadsheet tool (Google Sheets, Excel) works identically for this pivot.
Free for a single user, enough to produce the one-slide deliverable.
Paid upgrades (optional, faster/deeper)
The free path (a spreadsheet or Looker Studio plus Notion for the write-up) is complete on its own. Amplitude is only worth paying for once the team wants this check running continuously instead of pulled by hand.
An upgrade for a team that wants this table live every week instead of pulled from a CSV, never required to finish this project.
The process
3 steps
Step 01 of 03
The lesson's Step 2: plot your conversion rate at each AARRR stage and find the biggest drop-off, that's your highest-leverage point.
Which stage-to-stage transition is actually moving over the last four weeks, and which one looks frozen only because you haven't broken it into stages yet?
Procedure
- Group the daily rows into four 7-day weeks (Jun 1-7, Jun 8-14, Jun 15-21, Jun 22-28); drop the trailing partial days so every week compares like-for-like
- Sum 'visitors' within each week for each of the five stages: visit, signup, activation, purchase, retained
- Compute four stage-to-stage rates per week: visit-to-signup, signup-to-activation, activation-to-purchase, purchase-to-retained
- Line up the four weeks side by side and scan each rate for a trend, not just a single week's number
Week visit->signup signup->activation activation->purchase purchase->retained Jun 1-7 (W1) 9.9% 56.1% 30.7% 55.4% Jun 8-14 (W2) 10.6% 55.9% 36.5% 57.8% Jun 15-21(W3) 11.0% 54.9% 36.3% 57.6% Jun 22-28(W4) 10.9% 49.3% 35.0% 58.3%
Healthy
Each stage-to-stage rate holds roughly flat or improves week over week while topline visits/signups stay steady.
Unhealthy
One interior stage trends down for two or more consecutive weeks while the stages around it stay flat, a hidden leak a flat topline number will hide from you.
What this means
visit-to-signup is flat-to-improving (9.9% to 10.9%) and activation-to-purchase and purchase-to-retained both hold steady, but signup-to-activation drops for three straight weeks: 56.1% to 55.9% to 54.9% to 49.3%, a 6.8-point fall concentrated almost entirely in week 4. Acquisition and monetization both look fine; activation is quietly breaking.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| Signup-to-activation falling three-plus weeks running while top-of-funnel volume is flat | Pull the activation step itself (what does 'activated' mean here, e.g. first order placed within 48 hours?) and check whether onboarding, catalog load time, or delivery-slot availability changed in week 4 | half day |
| Flat weekly signups being read as 'stable' by leadership without a stage breakdown | Replace the single 'signups' slide with the four-stage table above in the next update, a flat topline is not the same claim as a healthy funnel | 30 min |
Step 02 of 03
The lesson's distinction: funnels drain, loops compound. A loop is healthy when this cycle's output measurably feeds the next cycle's input, e.g. more retained users than the equivalent week prior.
Is week-over-week retained-user volume actually compounding (each week produces more retained users than the last), or is it just riding the same flat signup volume?
Procedure
- Pull the absolute 'retained' visitor count per week from the same weekly aggregation (not the rate, the raw number)
- Compare each week's retained count to the week before it
- Check whether growth in retained users is outpacing, matching, or lagging growth in visits
Week visits retained (count) retained / visits W1 (Jun 1-7) 11,891 112 0.94% W2 (Jun 8-14) 12,654 159 1.26% W3 (Jun 15-21) 11,388 144 1.26% W4 (Jun 22-28) 11,800 130 1.10%
Healthy
Retained/visits ratio holds flat or climbs, meaning each unit of top-of-funnel traffic produces an equal or growing share of durably retained users, the mark of a compounding loop.
Unhealthy
Retained/visits ratio falls even while raw visit volume holds steady, meaning the same traffic now produces a smaller retained base, the activation leak from Step 1 showing up two stages downstream.
What this means
Retained/visits rose from 0.94% to 1.26% across weeks 1-3, then fell back to 1.10% in week 4, the same week signup-to-activation cratered. This is a funnel quietly draining exactly where the deck says it's 'flat,' not a loop compounding.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| Retained/visits ratio falling the same week signup-to-activation falls | Confirm the activation-stage leak from Step 1 is the cause (not a separate week-4-only issue like a delivery outage) before recommending a fix | half day |
| Leadership reading 'flat signups' as evidence the loop is stable | Reframe the forecast: without an activation fix, week 5's retained count is more likely to fall further than to recover on its own | 5 min |
Step 03 of 03
The lesson's Mistake 3: impressions, clicks, and open rates are outputs, revenue, retention, and activation rate are outcomes, every metric on a leadership slide should tie to an outcome, not a vague sense of 'growth is slowing.'
Given everything above, what's the one-line verdict for the leadership slide, and is it a forecast of continued plateau, mild decline, or something worse if nothing changes?
Procedure
- State the verdict in one sentence: healthy plateau, slow leak, or active break
- Name the specific stage responsible (signup-to-activation, week 4) rather than a vague 'growth is slowing'
- Attach the two numbers that prove it: the 6.8-point signup-to-activation drop and the retained/visits reversal
- Give a one-week check-in date to confirm whether the fix reversed the trend
VERDICT SLIDE, Instamart Acquisition Health, Week 4 Read Status: Not a healthy plateau, an activation leak currently masked by flat topline signups. Evidence: signup->activation fell 56.1% -> 49.3% over 4 weeks (-6.8pts); retained/visits reversed 1.26% -> 1.10% the same week. Forecast if unaddressed: retained users, the real compounding input to the loop, keep shrinking even if signups stay flat. Next check-in: pull week 5 numbers after the activation-step fix ships.
Healthy
The verdict slide names the specific stage and week, cites both supporting numbers, and gives a concrete re-check date.
Unhealthy
The verdict slide repeats the topline 'signups are flat' framing without naming the stage-level cause, leadership walks away no better informed than before the CSV was opened.
What this means
A specific, numbered verdict is what turns a spreadsheet exercise into something leadership can act on; a vague 'growth looks okay' slide wastes the four weeks of real data just read.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| Verdict slide says 'growth is flat' with no stage or number attached | Rewrite using the template above: named stage, two numbers, one re-check date | 30 min |
| No re-check date on the slide | Set a specific week-5 date to confirm the fix worked before treating the diagnosis as closed | 5 min |
Final deliverable
A one-slide verdict (see Step 3's template) plus the four-week stage-conversion table built to support it.
See a reference example
Verdict: not a plateau, an activation leak masked by flat signups, signup-to-activation fell 6.8 points over four weeks (56.1% to 49.3%), and retained/visits reversed the same week (1.26% to 1.10%). This is the same distinction Airbnb's growth team drew when it moved off signups and onto nights booked: a flat top-of-funnel number can hide a shrinking value-delivery stage underneath it. Recommended fix: audit the activation step (first order within 48 hours) for anything that changed in week 4 before assuming next week recovers on its own.
Success criteria
You're done when you can:
- Breaks the four weeks into stage-to-stage rates instead of reading only the topline signup number
- Correctly identifies signup-to-activation, not visit-to-signup or activation-to-purchase, as the stage that's degrading
- Connects the retained/visits reversal in week 4 to the same root cause identified in the stage table
- States a specific, numbered forecast/verdict rather than a vague 'growth looks fine' summary