Growth Marketing: 7 Levers and How Long Each Takes

The difference that actually matters: a funnel spends, a loop compounds.

Same job, two names, fifteen years apart. Our growth hacking guide owns the term, its tactics and the companies that made it famous. This page covers the other half: what growth marketing looks like as a standing practice, week after week, once the novelty of the word wears off.

Growth marketing vs growth hacking: the honest difference

Growth hacking is a job title from 2010. Growth marketing is what that job turned into when companies had to keep doing it for ten years.

Sean Ellis coined the term in "Find a Growth Hacker for Your Startup" on 26 July 2010, describing "a person whose true north is growth" and asking for three things: creativity, discipline in a systematic testing process, and the analytical judgement to decide what to keep and what to drop.

Two of those three are process, not tricks. The word "hack" is what aged badly: it got attached to one-off exploits, invite spam and scraped contact lists, until the label became a warning sign to investors. What survived is the part Ellis described: one person accountable for a growth number, working a testing process.

• Growth hacking, as the term is used: a tactic, a clever exploit, usually told as a story after it worked. Our growth hacking guide owns that territory on this site, including the tactic list.
• Growth marketing, as a practice: a loop you run whether or not this month produced a clever idea. Instrumentation, a queue of experiments, a decision rule, a written log.

If someone sells you growth hacking in 2026, ask which of their last five experiments failed.

Funnels versus loops, with a worked example

A funnel counts people dropping out at each stage. A loop is a system where the output of one cycle becomes the input of the next. Most Indian D2C and SaaS brands run on the funnel and should stop pretending otherwise.

The funnel most growth teams learn is AARRR, or "pirate metrics": Acquisition, Activation, Retention, Referral, Revenue. It comes from one person, Dave McClure, in a 2007 deck called Startup Metrics for Pirates (Mind the Product dates it and tracks Gabor Papp's 2017 RARRA variant, which puts retention first). A useful list to measure, never a standard, and treating it as one is how teams end up with five metrics and no decision.

The loop model is also one firm's argument, not a law. Brian Balfour, with Casey Winters, Kevin Kwok and Andrew Chen, published "Growth Loops are the New Funnels" on Reforge on 31 July 2018, defining a loop as a system where "the inputs through some process generates more of an output that can be reinvested in the input." Their objection is that funnels are linear and create silos. Fair, and still one point of view, from people working mostly on software-native distribution.

A worked example (hypothetical brand, illustrative numbers, not client data). Kaveri Roasters is a Pune D2C coffee brand. On the funnel view, spend goes in, some people buy, and next month you spend again for the same shape.

The loop version: every parcel ships with a card giving the buyer's friend ₹150 off a first order and the buyer ₹150 back when it is used. Suppose 8 of every 100 buyers get one friend to order. Each customer then produces 0.08 new customers, and those produce 0.08 more. That does not compound: to sustain itself, each customer has to produce roughly one more. What it does is take about 8% off blended acquisition cost. A CAC discount, not an engine, and most Indian D2C referral programmes we see land around here.

Where growth sits against brand and performance

Brand decides whether you get considered. Performance buys the click. Growth owns what happens between the click and the third purchase, and whether it compounds.

All three come out of one budget, so a working split helps. Stops producing the day you stop paying: performance. Changes how many customers each customer brings you: growth. Changes what people think before they search: brand.

Growth work needs a paid channel underneath it to generate enough traffic to test on, which is why the two sit together in most teams. Our performance marketing team runs that layer where a brand wants experiments and media buying in one place, with ad spend always separate from the management fee.

Which growth lever to test, and how long until you can read the result

Seven levers cover most of what a growth team touches. The last column matters most: teams kill good experiments on day four and keep bad ones for a quarter because nobody agreed a run length.

Google's guidance on custom experiments also asks that "your base campaign meets minimum requirements, such as more than 100 daily conversions" before you trust the read. Most Indian SMB accounts do not clear that bar, which is a reason to test offers and pages rather than bid settings.

Holdout and lift design is its own subject: our incrementality testing guide owns it on this site, and this page does not repeat that method.

How to run one growth experiment end to end

One experiment, start to finish, is the unit of work. The rest is scheduling:

1. Write the loop sentence it belongs to: "more reviews on product pages → higher conversion → more buyers → more reviews." Cannot write it? You have a task, not an experiment.
2. State the belief and one primary metric, chosen before launch, plus one guardrail you refuse to damage (refund rate, unsubscribe rate, CAC).
3. Size it before you build it. The arithmetic below says whether the test is possible on your traffic.
4. Build the smallest version that tests the belief. A hand-written email beats a three-week automation build.
5. Instrument it first. The event fires in GA4 or your back end before one user sees the variant. Untracked launches are how a quarter vanishes.
6. Run whole weeks. Tuesday buyers and Sunday buyers differ, and stopping mid-week bakes the weekday mix into the result.
7. Read it against the number you declared, not the metric that happened to move, and check the guardrail before you celebrate.
8. Log one line, win or lose, then decide: ship, kill, or iterate. A year of that log is the difference between a practice and a run of campaigns.

Worked calculation (our arithmetic, formula shown). The rule of thumb for comparing two conversion rates at 95% confidence and 80% power: visitors per variant ≈ 16 × p × (1 − p) ÷ Δ², where p is the current conversion rate and Δ the absolute improvement you want to detect.

Kaveri Roasters converts at 2% and wants a 20% relative lift, so p = 0.02 and Δ = 0.004: 16 × 0.02 × 0.98 ÷ 0.000016 = 19,600 sessions per variant, about 98 days at 400 sessions a day. Ask for a 50% lift instead, to 3%, and Δ = 0.01 drops it to 3,136 per variant, roughly 16 days. Only the size of the swing changed. That is the arithmetic case against colour tests at small scale: test offers, prices and page structure, and see our A/B testing guide for the split mechanics.

What a growth marketer actually does in a week

Less inventing than the job description implies. Far more instrumenting and waiting.

• Monday, 60–90 minutes: read last week's numbers against the declared metrics, and decide which running tests stop.
• Tuesday and Wednesday: ship. Variant, copy, developer brief, tracking. This is where the week goes.
• Thursday: check instrumentation. Events firing, variants splitting evenly, nothing broken on mobile.
• Friday: write up one finished experiment and re-rank next week's queue.
• Throughout: five unscripted customer conversations a month. Most ideas that worked for us started in something a customer said, not a benchmark report.

A realistic cadence is two to four experiments running at once and one or two finished reads a week.

The team and tooling a small Indian company needs

One person who owns the growth number, six to eight developer hours a week, and part-time design or copywriting. At 10–50 people that is the whole function.

The tooling is thinner than vendors suggest. GA4 and Microsoft Clarity are free and cover measurement and session recording. Your email platform sends the sequence, and your ad platforms already run splits: Google Ads custom experiments cover Search, Display, Demand Gen and Video, and Meta has its own A/B test in Ads Manager.

One budgeting note: there is no free Google testing tool any more. Optimize and Optimize 360 stopped working on 30 September 2023, and Google now points users at third parties, naming AB Tasty, Optimizely and VWO. At a few lakh rupees of monthly revenue, a platform-level split plus a spreadsheet beats a testing-suite licence.

Our view (opinion, not data): most work sold as "growth marketing" by Indian agencies is paid media with a new name on the invoice. The tell is the deliverable list. If the scope is campaigns, creatives and a monthly report, that is media buying, and it should be priced and judged as media buying. Growth work produces a different artefact: a log of experiments with dates, declared metrics and outcomes, failures included. We say this as an agency that also sells media buying. Both are worth paying for, but the relabelling happens because "growth" clears a procurement conversation that "ads management" does not.

The experiments worth running early, and the ones that waste a quarter

Stage-by-stage sequencing lives on our digital marketing for startups page. This is only about which experiments pay back first.

Worth running in your first quarter:

• Offer and price framing on the page that already has traffic. Biggest swing available, cheapest to build.
• The first-purchase experience. Fixing activation raises the value of every acquisition rupee you spend afterwards.
• The thank-you page and the parcel insert: free space, already-converted audience.
• A win-back sequence to buyers who went quiet 30 to 60 days ago, with a holdout so you can tell whether it worked.

What tends to eat a quarter:

• Micro-copy and colour tests at a 2% baseline. The maths above says you cannot read them.
• A referral programme built before you have a few hundred monthly customers to do the referring.
• Five channels at ₹20,000 each. Nothing clears its learning threshold, and the quarter ends with five inconclusive reads instead of one answer.

How to know the growth practice is working

Judge throughput and decisions first, outcomes second. Outcomes arrive late and noisy; throughput shows by week three. Four signals:

• Experiments finished, not started. One abandoned halfway costs what a failure costs and teaches nothing.
• Blended CAC against contribution margin, across two quarters. One month of movement is seasonality.
• Share of new customers arriving through routes you do not pay per click for. That is the number that says a loop exists.
• Whether anyone cites the log in a meeting: "we tried that in March, here is what happened."

Expect most experiments to come back flat. We quote no hit-rate percentage here because we have no citable figure for one. The few that move pay for the programme, which is why the log matters more than the ideas. Stage mapping for content and spend sits in our marketing funnel guide.