पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 19 · Build
Growth loops, not funnels
A funnel needs a fresh rupee at the top every month. A loop turns its own output into its next input. Map yours, time its cycle and put the money where the compound rate is highest.
Pathshala, The Founder Library · 11 October 2026 · 8 min read
Ask five people in a company how it grows and the answers usually describe a funnel: traffic at the top, sign-ups in the middle, paying customers at the bottom. The companies that grow for years without buying every customer can describe something else. Their customers, content or revenue come back round and produce more of themselves.
That second shape is a loop. The idea was set out most clearly by Brian Balfour, Casey Winters, Kevin Kwok and Andrew Chen in Growth Loops are the New Funnels on the Reforge blog in 2018. Their argument is short. A funnel explains one step of growth well and the whole system badly. A loop explains the system: a closed chain in which the output of one turn is reinvested as the input of the next. This lesson turns the idea into a method a founder can run with a spreadsheet: draw the loop, time it, compute its compound rate and decide where the next rupee goes.
Where the funnel misleads
The funnel is not wrong. Each stage in it is real and each conversion rate is worth knowing; the [funnel lesson](/library/sales-funnel-and-conversion-rates-to-expect) covers what to expect at each. The trouble is what the picture leaves out. The Reforge authors name three faults. A funnel runs in one direction, so it has no compounding in it: every month starts again with a new budget at the top. It splits planning into silos, with acquisition, product and monetisation each owning a layer. And teams optimising their own layer can damage the next one, as when marketing buys cheap sign-ups who never come back.
The practical cost in India is easy to see. A D2C brand spends ₹40 lakh a quarter on Meta and Google, reads a healthy funnel and grows only as fast as the budget does. Stop spending and the top of the funnel empties within a week. Nothing the brand built last quarter is helping it this quarter. That is a funnel. Compare a business-software product whose every invoice carries the product’s name to a buyer who may need the same tool, or a recipe platform whose every published recipe is a page that a stranger finds on Google next month. Each customer or page makes the next one more likely. That is a loop.
What a loop is, and the four common kinds
A loop has an input, a sequence of steps a person or a system takes, and an output that feeds the input again. The Reforge piece walks through Pinterest’s loop in five steps: a user signs up or returns, sees relevant content, saves it, the saves send quality signals that help the content rank in search engines, and new users find that content through search and arrive. Each turn of that loop needs no new marketing budget. It needs users doing what they came to do.
Four kinds cover most early companies. The referral loop: a customer gets value and brings another, as in Dropbox’s two-sided storage reward, which Drew Houston reported in his 2010 Startup Lessons Learned deck permanently increased sign-ups by 60 per cent, after paid search had cost $233 to $388 a customer for a $99 product. The content loop: users or the company create pages, pages rank, searchers arrive and some create more pages. The paid loop: a customer’s first-year margin funds the next customer’s acquisition, so the loop runs only when payback is shorter than the cash you have. The sales loop: a customer becomes a case study, a reference call or a referral inside their industry body, and the next deal closes faster because of it.
Network effects are the loop’s strongest form, because each new user makes the product more valuable to every other user rather than merely bringing one more in. NFX’s Network Effects Bible argues that network effects account for about 70 per cent of the value created in technology since 1994. Most companies will not have one. Almost every company can have a loop.
Map the loop on one page
Draw it as a circle of boxes, each box a step, each arrow labelled with the share of people or units that make it to the next box. Start with what actually happens, not what you hope happens. For a B2B invoicing product the loop might read: a customer sends invoices; each invoice reaches a buyer; some buyers open the link at the bottom; some of those sign up; some sign-ups become customers who send invoices. If the customer sends forty invoices a month to thirty different buyers, 10 per cent open the link, 8 per cent of openers sign up and a quarter of sign-ups become active, each customer yields about 0.06 new active customers a month. That is the loop’s yield: output per unit of input, per turn.
Two tests tell you whether you have drawn a loop or wishful thinking. Is the output of the last box the same thing as the input of the first? If the loop starts with customers and ends with leads, there is a step missing. And can you measure every arrow from your own data this month? An arrow you cannot measure is an arrow you cannot improve. Most companies find, once they draw it honestly, that they have one or two loops that matter and several funnels that feed them. The Reforge authors found the same: the fastest growing products are typically powered by one or two major loops.
Measure the cycle time
Yield is half the number. The other half is how long one turn takes: the cycle time, the days from an input entering the loop to the output being ready to act as input again. For a referral loop it is the days from a customer joining to the friend they invited joining. For a content loop it is the days from a page going live to it ranking and bringing a visitor who contributes. For a paid loop it is the payback period. For a sales loop it is the time from close to the first reference call that produces a lead.
Cycle time matters as much as yield because compounding counts turns, and a short loop turns more often. Paul Graham’s Startup = Growth makes the point in weekly terms: a company growing 1 per cent a week grows 1.7 times in a year, while one growing 5 per cent a week grows 12.6 times. A loop yielding 40 per cent per turn on a sixty-day cycle completes six turns a year. A loop yielding 12 per cent on a ten-day cycle completes thirty-six. Measure the median, not the average, because a few fast cases flatter the number; and measure it from your own logs, by cohort, the same way each month.
Turn yield and cycle into one compound rate
The two numbers combine into a single comparable rate. Multiply yield by the share of each turn’s output that survives to take part in the next, add one, raise it to the power of thirty divided by the cycle in days, and subtract one. That is the loop’s compound rate per month. Every loop in the company can now be put on the same scale and compared with the growth rate the business needs. The figure runs the arithmetic for two loops at once. Set loop A to your slow, high-yield loop and loop B to your fast, low-yield one, and see which one the next quarter’s money should go to.
A funnel is something you fill. A loop is something you turn. Invest in the one that turns fastest for the rupee.
Where to invest
The figure usually surprises in one direction: halving a cycle time does more than a large rise in yield. With the defaults, loop B’s 12 per cent per turn on a ten-day cycle compounds at about 32 per cent a month, more than twice loop A’s 15 per cent on a sixty-day cycle, though A’s yield per turn is over three times larger. The order of work that follows is simple. Shorten the cycle first: ask for the referral at first value rather than after a month, publish the page where it can rank this week, collect the case study at go-live rather than at renewal. Then fix the weakest arrow, the step with the lowest conversion, because a loop’s yield is the product of its steps and the smallest one bounds the rest. Only then widen the mouth by putting more input in, with paid acquisition or sales effort.
Reforge’s own example of the choice is useful. Initiative A brings 500 engaged users once. Initiative B brings twenty in the first week and grows 10 per cent a week. The authors prefer B, and the arithmetic agrees within about thirteen weeks, after which B has produced more users than A ever will. The trap to avoid is choosing between loops on yield alone, or on what a channel produced last month. Compare compound rates. And remember the retention term: a loop whose output churns before the next turn has no compounding at all, which is why the [retention lesson](/library/retention-is-the-growth-engine) comes before this one in the track.
The monthly loop review
On the first working day of each month, take your one or two loops and write one row each. The conversion on every arrow, from last month’s cohort. The yield per turn. The median cycle time in days. The retention through the loop. The compound rate per month that the four produce, and the same figure from last month. Mark the arrow with the lowest conversion and the step that takes the longest; one of them is next month’s project, and only one. Once a quarter compare each loop’s compound rate with the growth rate your plan needs. If no loop reaches it, the plan is relying on paid input, and the [paid acquisition lesson](/library/paid-acquisition-meta-google-and-the-ceiling) explains where that ceiling sits.
The rupee figures and rates above are illustrations; the sources are below. Draw your loop on one page this week.