पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 08 · Start

Referral and word of mouth by design

Word of mouth is not luck. It is a loop with three rates in it, a reward that suits the product and a moment of ask, and each part can be designed, measured and improved.

Pathshala, The Founder Library · 11 October 2026 · 7 min read

Two people hold clay cups of tea side by side by the sea. — Photograph: Jakob Welik · Pexels

Every Indian founder has been told that the best marketing is a happy customer. Most then wait for the happy customers to start talking. A few go and find out which customers talk, when, to whom and why, and build the product so that more of them do it. The second group calls it word of mouth too. It is not luck.

Referral is the channel Indian buyers trust most, because so much buying here already runs through someone you know: the CA who recommends the software, the society group that recommends the tutor, the cousin who recommends the phone. That trust is why it is also the channel founders most often leave to chance. This lesson treats it as a system with parts. A loop with three measurable rates. An incentive designed for the product. A moment in the customer’s life when asking is natural. And a sheet that shows whether the system is working.

Why referred customers are worth more

The best evidence comes from a study that had something founders rarely have: the full records of thousands of customers, referred and not. Philipp Schmitt, Bernd Skiera and Christophe Van den Bulte followed about ten thousand customers acquired by a large German bank in 2006, of whom roughly half came through a referral programme that paid the existing customer a €25 voucher. Their paper, Referral Programs and Customer Value in the Journal of Marketing in 2011, found that referred customers had about 25 per cent higher contribution margin at first, a gap that narrowed over time, and about 18 per cent lower churn, a gap that did not narrow. Over six years a referred customer was worth about 16 per cent more, and about 25 per cent more once their lower acquisition cost was counted.

A follow-up paper by the same group, How Customer Referral Programs Turn Social Capital into Economic Capital in the Journal of Marketing Research in 2018, explained why. Referrers bring in people like themselves, so a good customer tends to refer good customers; and the referred customer stays partly because the referrer does. When the referrer left the bank, the referred customer’s lower churn disappeared and turned higher. Two practical rules follow. Ask your best customers first, not all customers equally. And look after the referrer after the referral, because the new customer’s loyalty is borrowed from theirs.

The loop has three rates

A referral programme is a loop. Some share of new customers refer. Each referrer sends some number of invitations. Some share of those invitations become customers, who can refer in turn. Multiply the three and you have the loop’s coefficient, often written k. Twenty per cent of customers referring, three invitations each and fifteen per cent of invitations converting gives k of 0.09. That sounds small, and on its own it is: a loop below 1 does not grow a company without other channels feeding it. What it does is multiply them. Every hundred customers bought through other channels bring roughly 1 ÷ (1 − k) customers in all, so k of 0.09 brings roughly a tenth more customers for the same spend, and k of 0.4 brings two-thirds more.

The fourth number is time. The cycle is the days from a customer joining to the people they invited joining. A loop with a seven-day cycle completes fifty turns a year; a loop with a ninety-day cycle completes four. Two products with the same k but different cycles grow at visibly different speeds in the first year, which the figure below shows. The design work is therefore on four levers, and each has its own fixes: more customers referring (ask at the right moment), more invitations each (make sharing easy and specific), more invitations converting (make the landing for the invited person excellent), and a shorter cycle (ask earlier and make the first value faster).

The incentive: two-sided and in the product’s own currency

Dropbox is the case every growth team studies, and Drew Houston’s own Startup Lessons Learned deck from 2010 gives the numbers. Paid search had cost the company between $233 and $388 to acquire a customer for a $99 product. The referral programme, inspired by PayPal’s early cash bonus, gave both sides extra storage, and Houston reported that it permanently increased sign-ups by 60 per cent; in the thirty days before the talk users sent 2.8 million direct referral invitations. Three features made it work. The reward was two-sided, so the invitation was a gift rather than a favour. It was paid in the product’s own currency, so it cost Dropbox little and made the referrer a heavier user. And it was built into the product, one click away from where people already were.

For an Indian startup the translation is direct. A software product gives both sides a free month or extra seats. A D2C brand gives both a discount on the next order rather than cash, so the reward brings the referrer back. A services firm gives the referrer a priority slot or a free audit. Cash and UPI cashback work, and they also attract people who refer for the cash, which the bank study suggests is the wrong selection: you want referrers who resemble your best customers. Whatever the reward, cap it, write the terms in a single sentence, and pay it the day the referred customer reaches the point that matters, not the day they sign up.

The moment of ask

Most programmes ask at the wrong time: in the welcome email, before the customer has had any value, or in a footer nobody reads. Fred Reichheld’s work on the question “would you recommend this company to a friend”, published in the Harvard Business Review as The One Number You Need to Grow, rests on the observation that a recommendation is a customer putting their own reputation on the line, which they do only when they feel strongly. The design question is when in the customer’s life with the product that feeling peaks.

A vendor in a blue shirt serves chai in clay cups at a busy outdoor market stall. — The chai stall is where a neighbourhood trades its recommendations. A good ask arrives at the moment the customer is already pleased. Photograph: Anubrata Saha · Pexels

It is usually right after the first real value, the moment the [activation lesson](/library/activation-first-session-that-decides) is about: the first payroll run that went through without a query, the first order that arrived a day early, the first report the customer forwarded to their boss. Find that moment in your own product by asking the last twenty customers who referred someone when they did it and what had just happened. Then put the ask there, specific and small: not “refer a friend” but “who else in your trade body files GST every month? Send them this and you both get a month free.” A pre-written message the customer can forward on WhatsApp in one tap, in their language, converts better than a link they have to explain.

Word of mouth is a customer lending you their reputation. Design for the moment they are proud to, and never make them regret it.

Reading the loop, and what breaks it

Set the figure to your own numbers. With a hundred customers a month from other channels and k of 0.09 the loop adds a few per cent; it is worth running but not worth a team. At k of 0.3 with a three-week cycle the loop is adding about two-fifths again to everything else, and the referral programme deserves a named owner and a weekly number. Move the cycle from twenty-one days to sixty and watch the first year flatten even though k is unchanged: speed of the first value is a referral lever as much as a product one.

Three things break loops. Retention. A referred customer who leaves in a month damages the referrer’s reputation and the referrer stops; read the loop alongside the [retention curve](/library/retention-curves-and-flattening-test) and fix that first if it is falling. Fraud. Rewards paid on sign-up attract fake accounts and self-referrals; pay on the first real use and check for repeated devices, addresses and UPI handles. Fatigue. Asking every customer every week teaches them to ignore the ask; ask once at the moment of value and once more after the next milestone, then stop.

The monthly referral review

On the first working day of each month compute five numbers from the previous month and write them in one row with the date. The share of new customers who sent at least one invitation within one cycle. Invitations per referrer. The share of invitations that became paying or active customers. The median cycle time in days. And the share of all new customers who arrived by referral. Multiply the first three to get k and compare it with last month. Then change one lever only: the moment of ask, the reward, the pre-written message or the landing page for invited people, and note what you changed. Once a quarter, compare the retention and margin of referred customers with everyone else. If referred customers are not better, the programme is buying customers rather than earning them, and the reward needs to change.

The research above comes from one bank and one software company and will not transfer exactly to any other business. The sources are below. Find your moment of ask this week.