Cohort metrics and retention of TON wallets 2026

Any public TON metric — active addresses, daily transactions, wallet base growth — looks decent if you only read the aggregates. But ask one simple question, “how many of the people who joined during the May 2024 Notcoin wave are still transacting in May 2026”, and the aggregates go silent. Answering it requires cohort analysis — and that is precisely what most investor-facing reports avoid.

TL;DR

• DAU/MAU on TON are inflated because constant airdrop inflow masks the churn of older cohorts.
• Cohort retention = share of users from a given launch week who remain active N weeks later.
• Notcoin, DOGS, and Hamster Kombat each built huge one-shot cohorts with very steep falloff after the claim.
• Organic wallets coming through Wallet-in-Telegram show a fundamentally different retention shape.
• Without separating “wallet exists” from “wallet transacts”, any estimate of TON’s audience stays inflated.

What cohort retention is and why DAU/MAU mislead

A cohort is a group of users defined by the date of their first action. On TON this usually means the week or month of the first outgoing transaction from an address. Cohort retention in week N is the share of those addresses that performed at least one transaction during week N.

Standard public metrics work very differently. Daily Active Users counts unique addresses transacting in a day, with no reference to when those addresses appeared. Monthly Active Users does the same on a monthly window. The totals look healthy as long as new cohorts arrive faster than old ones leave.

On TON this is a structural problem, not a theoretical one. Between May and December 2024 the chain absorbed three mass airdrop waves, each minting millions of wallets in weeks. Those wallets show up in the next day’s DAU and the next month’s MAU, and averages start to look like organic growth. In reality, each wave is burning down on its own schedule and the next wave is filling the gap.

The cohort view dissolves that illusion. Lock the week of appearance, follow retention row by row, and each airdrop cohort comes out as a sharp step: a big spike around claim, a steep falloff, and a thin tail of survivors.

How to rebuild cohorts from on-chain data (methodology)

A fully public cohort dashboard for TON still does not exist. Re:Doubt and TonStat surface aggregates and some slices by first activity, but the retention matrix has to be assembled by hand. The baseline method looks like this.

Step 1. Define the appearance event. Candidates: first incoming transaction, first outgoing transaction, first jetton claim. The most honest choice is first outgoing, because an inbound airdrop does not prove the address is controlled by a human.

Step 2. Group by appearance week. This yields a table of “week X → addresses born during that week”.

Step 3. Track activity per cohort. For each subsequent week, count the share of the original cohort that transacted. That is the retention curve for one cohort.

Step 4. Compare cohorts. Overlay curves on a shared “weeks since appearance” axis to see which cohorts fade fastest.

Technically this is a query against the TON transaction index — toncenter, tonapi, or a local node via liteclient — aggregated per address. Re:Doubt partially exposes such slices on its activity dashboards, but even there the full “appearance week × activity week” matrix usually stays behind the public surface.

Case: the Notcoin cohort (May-June 2024) — scale and churn

Notcoin built the largest cohort in TON history. The mini-app accumulated 35M+ users by April 2024, and a substantial fraction of them activated wallets in May 2024 around the NOT claim. The deep breakdown lives in our Notcoin retrospective.

The cohort behaves like a textbook tap-to-earn curve. Activity peaks during the claim and listing week of 16 May 2024. Within four to eight weeks, dashboards show a sharp drop — most addresses move NOT to an exchange or swap it once, then never return. The surviving tail is meaningfully smaller than the initial volume.

Important qualifier: precise retention percentages for this cohort are not consolidated in a single public source. Re:Doubt and analytical threads display a steep dropoff visually, but counting methodology differs across authors. The correct framing is: the cohort is huge, the dropoff is dashboard-visible, and the surviving tail is clearly the minority.

Case: DOGS and Hamster Kombat cohorts

DOGS (August 2024) and Hamster Kombat (June 2024) form the next two large cohorts, and each repeated the Notcoin pattern at its own scale.

Hamster Kombat exceeded Notcoin in registered bot users, but conversion into actual claims and retention came in lower. The cohort showed up as a short flash, passed through the claim, and collapsed quickly — Hamster Kombat is barely visible in current active on-chain TON metrics.

The DOGS cohort was wider in reach and steeper in falloff. The initial audience was massive, but the retention curve decays faster than Notcoin’s: the project had fewer post-claim retention mechanics.

Across the three cohorts a pattern is clear: the thinner the post-distribution mechanic, the steeper the dropoff. Notcoin lasted longer thanks to Squad seasons and Not Games. DOGS and Hamster Kombat did not.

Case: the “organic” cohort through Wallet-in-Telegram

Cohorts unattached to a one-off claim behave very differently. Wallet-in-Telegram — the embedded custodial wallet inside the messenger — produces a steady inflow of users driven by on-ramp and P2P features. These users do not arrive for a specific token claim; they arrive to send money or buy USDT-jetton.

The organic cohort’s retention curve is visibly flatter on dashboard observation. There is still a first-week dropoff — some people buy USDT once and never come back — but the tail holds up significantly better. A fraction of these users gradually convert into active DeFi or gaming participants.

Exact percentages are again not available in the open, but the qualitative conclusion is robust: the organic channel delivers fewer users, but of higher quality. For honest sizing of TON’s active audience, organic cohorts weigh more than airdrop waves.

”Wallet exists” is not “wallet active” — the critical gap

The most common analytical mistake in TON reports is equating address count with audience size. After airdrop waves this distortion is severe.

A wallet can sit in one of three states.

• Fully empty. Address derived from a public key, never transacted. TON has many of these because the address is computable from the key without a deploy.
• Got the drop, abandoned. The address has a balance or a single claim record and nothing else. This is the bulk of tap-to-earn cohorts.
• Active. Recurring transactions, swaps, transfers, dApp interactions.

Only the third bucket belongs in any serious audience estimate. When an investor or partner sees “TON has N million wallets”, N usually contains all three. The realistic active base is smaller — sometimes by an order of magnitude.

What this means for builders and investors

For projects on TON, the cohort lens reframes the goal. Pulling new users through an airdrop produces a spike that looks great in a pitch deck, but if the cohort’s retention is no better than baseline noise four to eight weeks later, the campaign did not build an audience — it ran a one-off giveaway. This conclusion is consistent across independent observations of 2024’s tap-to-earn waves.

For investors, the implication is direct. Comparing TON to other ecosystems by DAU without adjusting for the source of that DAU is misleading. If a meaningful share of TON’s active addresses is the long tail of airdrop cohorts, a clean comparison with chains that lack this mechanic loses meaning. A realistic estimate requires a cohort table and explicit separation of organic inflow.

For content teams and TON-project partners, the takeaway is simpler: measuring campaign success in wallets created is almost always wrong. The right metric is cohort retention at 30 to 90 days.

Conclusion

Cohort retention is not an exotic tool. It is the baseline way to separate real growth from the arithmetic illusion produced by consecutive airdrop waves. TON went through several such waves in 2024, and public metrics still carry their fingerprints.

The good news: the data needed to build cohort tables sits in the open on-chain. The bad news: a unified public retention dashboard for TON still does not exist, and every team rolls its own. Until that infrastructure matures, any claim about “N million active TON users” deserves an asterisk — it is probably a DAU/MAU figure with no cohort breakdown, and the real content is smaller than advertised.

For anyone building on TON or considering an investment, the only honest path is to start asking cohort questions and to stop trusting aggregates. See also our breakdown of TON DeFi metrics and the guide to reading on-chain metrics — both are useful methodological groundwork before the cohort layer.

Sources

• redoubt.online — public activity dashboards and user slices for TON.
• tonstat.com — aggregated on-chain metrics, address and transaction dynamics.
• tonscan.org — transaction index for building custom cohort queries.
• toncenter.com — API for extracting first-activity per address.