Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Growth Marketing › Intermediate

Designing a Referral Loop for a Freelance Invoicing App

An exercise in turning something users already do into a loop that brings in new users, then sizing it honestly. You map the current loop from data, design the trigger, incentive and first step of a better one, estimate its return per cycle and cost, and set the measures and holdout that will show whether it compounds and whether it adds users the business would not have won anyway.
ExerciseIntermediate levelModule 3: Loops, models and experimentation at scale5 minUpdated

Why this matters

Paid acquisition stops the day the budget does. A loop turns the output of one cycle into the input of the next, and the growth loops framework insists that every arrow carries a number, so a team can see whether its loop compounds or is a funnel drawn as a circle. A referral loop only works where users reach value alone and can bring others in without a salesperson, the conditions the comparison of product-led and sales-led growth sets out. A freelance invoicing app meets them: the price fits on a card, value arrives when an invoice is paid, and every invoice already reaches someone new.

The app and its users

Tallyroot is an invented invoicing app for freelancers. Users create invoices, send them by email and get paid through a hosted payment page. Every invoice carries a small "Sent with Tallyroot" footer link. All deliberate acquisition is paid search. Some recipients are freelancers who send invoices too.

Illustrative example — not a real result

Tallyroot's figures for one month

MeasureValue
Paying users3,000
Invoices sent24,000
Distinct recipients15,000
Payment page views12,000
Footer link clicks144 (1.2% of page views)
Sign-ups from the footer40
Of those, sent a first invoice within 14 days22
New paying users from paid search300
Paid search cost, including team time£15,000
Price£8 a month
Contribution per paying month£6.50
Average paying months in the first twelve, paid search users9

Per 100 newly activated users, each sends invoices to about five distinct clients in their first cycle, and about 70 have their first invoice paid within it. Cycle time, from one user's activation to a referred user's activation, is about six weeks.

Designing the loop

Answer the loop framework's questions in order.

  • Input. Which users, at which moment?
  • Action. What do they do, and what do recipients see?
  • Incentive. What does it reward, and does the reward depend on the referred user reaching value?
  • First step. What must a referred user do first, and how is that made fast?
  • Return. What share of the output comes back as activated users, and how long does a cycle take?
  • Worth. Are the loop's customers worth having, and would they have come anyway?

Task

  1. Size the current footer loop as a return per cycle per activated user.
  2. Design the trigger and the recipient-facing change, and say why each sits where it does.
  3. Choose an incentive, rejecting at least one alternative with a reason.
  4. Define the referred user's first step and the activation event.
  5. Estimate the redesigned loop's return per cycle and what it adds per month, with your assumptions labelled.
  6. Compare the loop's cost per activated user with paid search CAC.
  7. List the measures, holdout and guardrails that will show whether it works.

Worked answer

The current loop barely turns. The footer returns well under a hundredth of an activated user per cycle.

Trigger: the first paid invoice. When a user's first invoice is paid, show a prompt to invite other freelancers, with a ready-made message. Being paid is the value the product promises. On the payment page, replace the footer with a short line aimed at recipients who invoice too: "Send invoices like this one", carrying the sender's referral link.

Incentive: a free month for both, released when the referred user's first invoice is paid. Tying the reward to a paid invoice rewards real users rather than sign-ups and makes fake accounts pointless. Cash was rejected because it attracts referral hunters.

First step: send a real invoice. The referred user lands on a sign-up page that names the referrer, starts from a template with the payment link built in and is asked for one thing: a client's email address. Activation is a first invoice sent within the window; quality is that invoice being paid.

Illustrative example — not a real result

Sizing the loops per 100 newly activated users

ArrowCurrent footerRedesigned loop (planning assumptions)
Invite promptNone70 paid → 20% invite → 14 users × 2 invites = 28 invites → 25% sign up = 7 → 60% activate = 4.2
Payment page500 recipients → 400 views → 1.2% click = 4.8 → 28% sign up = 1.33 → 55% activate = 0.73400 views → 2% click = 8 → 35% sign up = 2.8 → 55% activate = 1.54
Activated users returned0.735.74
Return per cycle0.0070.057

Added per month at 300 paid users: 300 × 0.057 ÷ (1 − 0.057) ≈ 18 activated users, across all cycles.

Incentive cost per rewarded referral: the referrer's free month (£8 of revenue forgone) plus serving the referred user's free month (£1.50), so £9.50, against a paid search CAC of £15,000 ÷ 300 = £50.

Paid search twelve-month LTV: 9 × £6.50 = £58.50.

The verdict. Even on hopeful assumptions the loop returns far below one per cycle. It is a supplement to paid search, not a replacement, adding a modest stream of users at a fraction of paid cost. Date the map and replace each planning figure with a measured rate as soon as one exists.

Measures. Return per cycle and cycle time, read by cohort so decay shows; the rate on every arrow; the retention and LTV of referred users against paid search users; incentive cost per activated referred user; and self-referral checks. Some referred users would have found Tallyroot anyway, so the loop's Incrementality needs a holdout: a share of new users who never see the prompt, compared on the activated users their invoices and contacts bring in. Guardrails: complaints about invite emails and any fall in the payment page's completion rate. Run it through the experimentation programme playbook with a decision rule.

Common mistakes

  • Drawing the intended loop. Map today's loop from data before designing the new one.
  • Rewarding sign-ups. Incentives paid before value is reached attract fake accounts.
  • Calling a small loop an engine. A return far below one cannot replace paid acquisition.
  • Counting every referred user as new. Without a holdout, the loop takes credit for users who were coming anyway.

Next, apply everything in this level to the growth strategy capstone, or return to the intermediate growth level.

Pages this lesson applies

Terms used on this page

Also related

Related