Why a running loop's sizing does not transfer to a launch
The framework sizes a loop that already turns. It takes a month of sign-ups, follows them round, and reads the share that comes back. At launch there is no month of sign-ups, no history of how users behave and no rate on any arrow. Every number on the map is a guess until people arrive, and the first people who arrive are unusual: they came because of press, a waitlist or a personal connection, not because the loop brought them.
Three further problems follow. Launch attention is concentrated into a short window, while a loop needs time to turn. The early users tend to be enthusiasts who invite more than later users will, which flatters the first reading. And the launch spike puts so many people into the product at once that any loop contribution disappears inside the total. The launch objective hub defines the window and its KPIs; the decisions below fit a loop into that window without letting it take credit for the spike.
Decisions that change for a launch loop
| Loop element | In a running loop | At launch | Decision it forces |
|---|---|---|---|
| Whether to rely on a loop | Already established | Unknown; depends on early shareable output | Whether the plan leans on the loop at all |
| Input | A steady stream of new users | A seeded group, chosen and recruited | Which group, and how dense |
| Temporary loops | None | A referral waitlist before release | What checks and consent it needs |
| Cycle time | A rate to improve | A constraint to set against the window | What carries the spike and what carries the months after |
| Instrumentation | In place | Has to be built before release | What enters the readiness gate |
| Reading the return | Rolling, from the whole base | By launch cohort, after the spike | When and how the loop is judged |
Decide whether the launch can lean on a loop at all
A loop needs an output that reaches new people through normal use: a shared list, an invitation to collaborate, a public profile, a result worth posting. Ask how soon after first use that output appears. If people create something shareable within their first days, the loop can start during the launch's tail. If sharing only happens after months of use, the loop is irrelevant to the launch and belongs in a later plan. Choose honestly; a product that is not better with other people will not grow through invitations however well the prompt is designed, and forcing one into the plan diverts effort from the channels that will carry the launch.
Seed a dense group whose members already know each other
A loop that starts among strangers has nobody to invite. Seed it in a small group with existing ties, such as a set of teams, households, clubs or classes, so that the first users have obvious people to bring in and those people have a reason to join. The diffusion of innovations framework explains why early adopters are the right recruits; for a loop, density matters as much as enthusiasm. The trade-off is reach against connection: a broad launch audience produces more first users but fewer invitations each. Choose density for the seed and breadth for the launch campaign.
Run the referral waitlist as a temporary loop with checks
A waitlist where each confirmed referral moves someone up the queue is a loop that runs before the product exists. It needs guardrails a running loop may not. Use double opt-inA sign-up process in which the subscriber confirms through a link sent to the address given. so only confirmed addresses count, and check for disposable domains and repeated devices, since queue rewards attract fake sign-ups. Keep consent for later marketing separate from the waitlist itself: joining to hear about the launch is not agreement to receive everything afterwards. The Information Commissioner's Office guidance on direct marketing covers consent and refer-a-friend messages; read it before writing the invitation copy. The guide to launching a new app covers waitlist pages and launch-day sends.
Set the cycle time against the launch window
Estimate how long one turn takes, from a user's first session to the first session of someone they invited, and compare it with the window. When the cycle is longer than the window, which is the usual case, the loop cannot carry the spike. Press, paid reach, creators and owned channels must do that, and the loop is planned to carry the months after. The launch planning and tiers guide sets the window; the loop's job is the post-launch phase that plans often end too early. Choose the channels for the spike first, then budget the loop work for the weeks that follow.
Put invite events and attribution into the readiness gate
A loop that is not instrumented on launch day cannot be measured later, because the launch cohorts will already be gone. Before release, the invitation sent, the invitation opened, the invited sign-up and the invited user's first value need events, and every invited sign-up needs to carry who invited them. For apps, a deferred deep link has to survive the install so the invitee lands where they were invited; the deep links and app analytics guide covers the events that confirm it. Where the shareable output is public content, the user-generated content guide covers rights and crediting. Choose to delay release over launching blind.
Read the return per cycle by launch cohort, after the spike
Launch-week totals mix press-driven sign-ups, waitlist conversions and invited users, and they flatter the loop. Group users by when they joined, follow each CohortA group of customers who share a starting event and period, tracked together over time. until it has had a full cycle to invite, and read the return per cycle for each. Expect the seed and launch-week cohorts to differ from later ones. A loop below one is normal; what matters is the direction across cohorts and which arrow is leaking. The growth loops in practice guide covers loop factor and throughput; at launch the extra rule is patience.
Decision points in a worked launch loop
Chorebloom seeds, launches and then reads its loop by cohort
Chorebloom, an invented app for sharing household chores, is only useful once a second housemate joins, so a loop is plausible: the first user creates a household and invites the others within the first day. The team seeds it in 60 shared houses, 230 people, recruited through housing co-operatives. A referral waitlist runs for 8 weeks and collects 4,200 sign-ups; fraud checks remove 600 from disposable domains and repeated devices, and 38% of the remaining 3,600 tick a separate box for marketing after launch.
The median cycle, from a user's first session to an invited housemate's first completed chore, is 19 days against a 14-day launch window, so press and paid social carry launch week, bringing 10,900 installs. Launch-week dashboards show 1,600 invited sign-ups and the team briefly calls the loop a success.
| Cohort | Activated users | Invites sent per activated user | Invites that become activated users | Return per cycle |
|---|---|---|---|---|
| Seed houses | 230 | 1.1 | 60% | 0.66 |
| Launch week (press and paid) | 7,200 | 0.5 | 40% | 0.20 |
| Weeks 3 to 6 | 2,100 | 0.9 | 35% | 0.32 |
| Weeks 7 to 10, after the invitation fix | 1,800 | 1.0 | 55% | 0.55 |
The cohort view shows that launch-week users, mostly people living alone who read the coverage, rarely invited anyone. In weeks 3 to 6 invitations rose but most invitees landed on a generic sign-up screen, not their household. A deferred deep link that opens the inviter's household board lifts invite-to-activation from 35% to 55%. The loop is still below one, but it now lowers the cost of every paid user, and the team funds it for the next quarter.
Without the cohort view, the launch-week total would have been read as a working loop. The table separated the seed houses, which behaved as the plan hoped, from the press audience, which did not fit the product's shared use, and it located the leak in the invitation's landing rather than in the willingness to invite.
Measuring a launch loop: KPIs and cadence
Before release, read the waitlist weekly: confirmed sign-ups, referred share after fraud checks and the share giving separate marketing consent. During the window, the launch's own KPIs lead, and the loop is watched only for broken events. After the window, the loop's KPIA metric tied directly to an objective. set is the return per cycle by launch cohort, the cycle time, invitations per activated user and the share of invitations that become activated users, read every two to four weeks as each cohort completes a cycle. The Activation RateThe share of new users or sign-ups who complete the activation event, the action marking first real value, within a set window. of invited users sits beside it, since an invitation that brings someone who never reaches first value is a leak, not a return. A referral programmeA scheme that rewards existing customers, and often the friends they invite, when a recommendation brings in a new customer. with rewards needs its cost per referred user in the same view.
Where a launch loop misleads
Seed groups are chosen for their connections, so their return overstates what strangers will do. Waitlist referrals are driven by the queue reward and stop when it disappears, so they say little about sharing after launch. Early cohorts are small and noisy, and a single enthusiastic household can swing a reading. And press coverage that reaches the wrong audience lowers the launch cohort's return without saying anything about the loop's design, which is why cohorts are compared by source as well as by date.
Common mistakes
- Planning the launch around a loop that cannot turn in time. The window is shorter than the cycle.
- Seeding with strangers. A first group with nobody to invite starts nothing.
- Counting unconfirmed waitlist referrals. Queue rewards attract fake sign-ups.
- Folding marketing consent into the waitlist. Joining for launch news is not agreement to everything after.
- Launching without invite events. The launch cohorts cannot be measured later.
- Judging the loop on launch-week totals. Read it by cohort once the spike has faded.
