Most Social Accounts Are Shrinking: What 6 Million Daily Snapshots Reveal About Follower Decay
We tracked 6.1 million daily snapshots for 52 days: 63.3% of X accounts and 62.2% of Telegram channels shrank. What follower decay does to price.
Follower decay is the default state of a social media account, not the exception, and almost nobody in this market prices for it. Across 5,810,419 daily Telegram snapshots and 323,072 daily X snapshots taken between June 9 and July 31, 2026, the median tracked account ended the window smaller than it started. Not a fringe of dying accounts. The median. Among X accounts with at least 30 days of tracking, 63.3% shrank and 36.7% grew. Among Telegram channels, 62.2% shrank and 37.7% grew. That finding reframes the entire buying decision: an account is a depreciating asset unless the seller can prove otherwise, and the depreciation rate varies by a factor of twenty depending on which category it sits in.
The short answer, before the evidence. On Telegram the median tracked channel lost 0.55% of its subscribers over 52 days, which compounds to roughly 3.8% a year, and the damage is wildly uneven by category. Telegram crypto channels are the largest category in our panel at 1,952 channels, and only 14.1% of them grew; the median crypto channel lost 3.02% in 52 days, a pace that removes close to half a subscriber base in three years. On X the median tracked account lost 0.06%, close to flat, but the daily picture is uglier than the net: accounts above 100,000 followers lost followers on 63.9% of observed days, with a median daily change of -23. The rest of this article is the evidence, the mechanics behind it, the arithmetic that converts a decay rate into a discount, and an honest account of what a 52-day window cannot establish.
The median tracked account shrank, and that changes what an account is worth
Two panels, two platforms, one direction. Restricting to entities with at least 30 days of tracking inside the window, so that a single missed crawl cannot swing a result, here is the 52-day net change.
| Platform | Entities with 30+ tracked days | Share that grew | Share that shrank | Median net change |
|---|---|---|---|---|
| X (Twitter) | 7,901 | 36.7% | 63.3% | -0.06% |
| Telegram | 43,252 | 37.7% | 62.2% | -0.55% |
The tails matter as much as the middle. On X the 10th percentile account lost 0.22% and the 90th percentile gained 1.13%, a tight distribution consistent with a panel of very large, slow-moving accounts. On Telegram the spread is an order of magnitude wider: the 10th percentile channel lost 5.15% and the 90th percentile gained 5.11%. Telegram is a higher-variance asset class in both directions, which is why buying one on the headline subscriber number is a mistake and why the category tables below are the most useful thing in this article.
Notice what these numbers do not say. They do not say social media is dying. Roughly 37% of tracked entities on both platforms grew, and some grew a lot. What they say is that decline is the base case, so growth is a claim that requires evidence rather than an assumption you get for free. The burden of proof has been sitting on the wrong side of the table.
How we measured follower decay: 6.1 million daily snapshots
Every figure here comes from the crawler index that powers our public directories, queried on July 31, 2026. The methodology matters more than usual, because the headline claim is unusual and you should be able to check the reasoning.
The corpus is 5,810,419 Telegram daily snapshots and 323,072 X daily snapshots, both covering June 9 to July 31, 2026, a 52-day window. A snapshot is one entity observed on one day with its follower or subscriber count recorded. Daily tracking on X covers 48,962 accounts. On Telegram it covers a much wider set, of which the per-tier cut below draws on 3,431,529 day-observations across 289,313 distinct channels.
The trajectory analysis, meaning the 52-day start-to-end change, requires at least 30 tracked days per entity. That threshold cuts the panel hard: 7,901 X accounts qualify out of 48,962 tracked, which is 16.1%, and 43,252 Telegram channels qualify. We use the threshold anyway, because a two-point comparison built on sparse coverage is not a trajectory, it is noise with a direction attached.
The most important caveat is about X specifically. Our X daily tracker prioritises large accounts: the median tracked X account carries 1,811,025 followers, and 84.8% of all X day-observations in the window come from accounts above 100,000 followers. Every X finding here therefore describes large X accounts, not X as a whole. If you own an account with 4,000 followers, read the Telegram sections and the mechanics section for the transferable part, and use our X follower count benchmarks study, which covers the full 17.6 million account index.
One row we threw away, and why
The raw per-tier output for X contained a size-tier row with six day-observations whose reported daily changes were larger in magnitude than the follower totals of the accounts involved. A daily delta cannot exceed the count it is a delta of, so those values are impossible and describe a data artefact, not an audience event. We dropped the entire row. Six observations out of 274,104 is 0.002% of the tier analysis, far too few to move a median, but leaving them in would have poisoned the percentile tails with a physically impossible number. The rule we follow: discard values that cannot exist, disclose that you discarded them, and confirm the conclusion holds either way. It does.
Do accounts lose followers on ordinary days? Most of them do
The 52-day net change is the number that matters for pricing, but the day-level view explains what is happening underneath it. For X accounts above 100,000 followers, our panel holds 274,104 day-observations across 14,476 accounts. The median daily change is -23 followers. The share of days ending in a net loss is 63.9%. The 90th percentile day is +778 followers.
Those three numbers describe a distribution that is negative in the middle and skewed positive at the top. The typical day is a small bleed. The good days are far bigger than the bad days, which is why a large account can lose followers on nearly two days in three and still finish 52 days roughly flat. A viral post, a mention by a bigger account, or a news cycle delivers a lump of new followers that offsets weeks of drip.
A useful consistency check: -23 followers a day for 52 days is -1,196 followers, and against the median tracked account of 1,811,025 followers that is -0.066%. The observed median 52-day net change is -0.06%. Medians do not add, so this is a sanity check rather than a derivation, but two independent cuts landing in the same place is reassuring.
The practical consequence for a buyer: a single-day or single-week measurement is close to worthless. If you check a large X account on two consecutive days and it went down, you have learned nothing, because that happens 63.9% of the time to perfectly healthy accounts. Across 30 days, a downward trend means something. That is the difference between a data point and a trajectory.
The scale paradox: the bigger the channel, the more days it shrinks
On Telegram the probability that a channel shrinks on any given day rises monotonically with its size. This is the opposite of the intuition most buyers carry, which is that large channels are established and therefore stable.
| Subscriber tier | Day-observations | Channels | Median daily change | Share of days shrinking |
|---|---|---|---|---|
| Under 1K | 19,401 | 16,983 | 0 | 31.1% |
| 1K to 10K | 1,079,876 | 178,456 | -1 | 55.7% |
| 10K to 100K | 2,038,564 | 83,417 | -5 | 65.2% |
| 100K and above | 293,688 | 10,457 | -65 | 74.0% |
A channel above 100,000 subscribers shrinks on roughly three days out of four, and the median day costs it 65 subscribers. Sustained without offsetting inflow, 65 a day is 23,725 subscribers a year. For a 200,000-subscriber channel that is an 11.9% annual loss on the day-level median alone.
Before concluding that big channels have simply gone quiet, check the posting data. Across the full Telegram index, channels above 100,000 subscribers post a median of 4.43 times a day, against 0.43 times a day for channels under 1,000. The large channels in this dataset are ten times more active than the small ones and they are still the ones losing subscribers most consistently. Whatever drives the decay, it is not that the operators stopped working.
One caution: the tier table and the 52-day trajectory panel are different populations. The trajectory panel of 43,252 channels is dominated by mid-sized channels with continuous coverage, and its median is -0.55%. The 100K-plus tier is a smaller, more heavily crawled slice with much larger losses. Both are true. If you are buying at the top of the size range, the tier numbers describe your risk, and the Telegram channel benchmarks study carries the matching view rate and cadence data.
Why follower counts fall when nothing has gone wrong
Decay is not a symptom of a broken account. It is the arithmetic of any audience that accumulates over time. Four mechanisms produce almost all of it.
Dormant subscribers get deleted out from under you
Telegram states in its own FAQ that an account which does not come online for 18 months is deleted along with all of its data. That is not an edge case, it is a rolling background process applied to the entire user base. A channel that acquired a burst of subscribers three years ago is still absorbing the deletion tail from that burst today, and the operator did nothing wrong to earn it. Every platform has some version of this, whether deletion, deactivation or enforcement removal, and each removal quietly decrements somebody's follower count.
The inflow and outflow equation, and why it guarantees the scale paradox
Model a channel with two parameters: it adds roughly A new subscribers a day from whatever acquisition it does, and it loses roughly a fraction c of its existing base each day to unsubscribes, deletions and removals. The net daily change is A minus c times B, where B is the current base. Set that to zero and the equilibrium size is A divided by c. Below that size the channel grows. Above it the channel shrinks. Every channel has an equilibrium size set not by how big it is but by how fast it acquires and how fast it churns.
This is why the scale paradox is not a paradox. The loss term scales with the base while the acquisition term does not. A channel that reached 300,000 subscribers through a viral moment or paid traffic has been pushed above the equilibrium its ordinary acquisition rate can sustain, and it will slide back toward that equilibrium no matter how diligently it posts. We observe only the net change, not A and c separately, so treat this as a way of reading the data rather than a measurement of it. It predicts the tier gradient above exactly, and no story about lazy operators does.
The buying implication is sharp. You are not purchasing a follower count. You are purchasing an acquisition rate and a churn rate, and the follower count is only the current position between them.
The dormant tail shows up in reach before it shows up in the count
There is independent corroboration for the dormant-subscriber mechanism in a completely different measurement. Across the Telegram index, the computed view rate, meaning average views divided by subscribers, collapses as channels scale: a median 31.3% at 1,000 to 5,000 subscribers, 12.7% at 10,000 to 50,000, and 3.3% above 500,000. A large channel reaches a far smaller fraction of its own list than a small one does. That is precisely the signature you would expect if large lists carry a heavy tail of accounts that are technically subscribed and functionally gone.
Purchased followers evaporate on their own schedule
An account inflated with purchased followers carries a liability that matures without warning, because the accounts sold to it are exactly the accounts most likely to be removed. If an account's growth history shows a vertical step followed by a long slide, that shape is the fingerprint. Read how to check if an X account has real followers for the checks that catch it before you make an offer.
Category is destiny on X: only tech and news grew
Splitting the X trajectory panel by category gives the first cut a buyer can act on. Category labels exist for 2,208 of the 7,901 tracked X accounts, or 27.9%, so read these as directional for the categories with meaningful counts and ignore the thin ones.
| Category | Accounts | Share that grew | Median change | P90 change |
|---|---|---|---|---|
| tech | 33 | 60.6% | +0.171% | +32.49% |
| news | 560 | 57.3% | +0.039% | +2.51% |
| sports | 379 | 45.4% | -0.041% | +2.42% |
| politics | 272 | 39.7% | -0.050% | +1.78% |
| science | 57 | 36.8% | -0.064% | +0.73% |
| gaming | 77 | 44.2% | -0.076% | +4.96% |
| education | 90 | 31.1% | -0.083% | +0.52% |
| movies | 155 | 27.7% | -0.087% | +0.84% |
| business | 157 | 31.8% | -0.088% | +1.70% |
| music | 217 | 27.6% | -0.094% | +1.02% |
| design | 30 | 33.3% | -0.107% | +0.46% |
| memes | 61 | 31.1% | -0.112% | +1.87% |
| crypto | 120 | 26.7% | -0.132% | +2.53% |
Only two categories posted a positive median: tech at +0.171% and news at +0.039%. Tech is a 33-account cell, thin enough that the median is a hint rather than a fact, though +32.49% at the 90th percentile says something real about the upside available in that vertical. News is the robust cell, with 560 accounts and 57.3% growing, and it is the only category in the table that is both large and positive.
Crypto is the worst X category on both measures: 26.7% grew and the median lost 0.132%. Hold that thought, because the Telegram version of the same finding is far more violent. If you want to see the current shape of a vertical before you shop in it, the directory hubs run on the same index: the tech X leaderboard shows what a healthy category looks like at the top end.
Telegram crypto channels are the worst performers in the entire dataset
This is the finding that should change how the market prices Telegram. Category labels exist for 8,652 of the 43,252 tracked channels, 20.0% of the panel, and unlike the X cut several of these cells are large enough to be genuinely reliable.
| Category | Channels | Share that grew | Median change | P90 change |
|---|---|---|---|---|
| ai | 165 | 50.3% | +0.07% | +5.35% |
| programming | 213 | 48.8% | -0.03% | +4.07% |
| music | 404 | 48.3% | -0.14% | +7.30% |
| tech | 301 | 45.8% | -0.22% | +4.67% |
| health | 96 | 41.7% | -0.28% | +4.22% |
| movies | 187 | 42.8% | -0.35% | +5.81% |
| design | 170 | 42.9% | -0.39% | +2.87% |
| education | 1,203 | 40.4% | -0.39% | +3.75% |
| travel | 79 | 35.4% | -0.46% | +3.60% |
| science | 184 | 41.3% | -0.50% | +3.60% |
| business | 205 | 36.6% | -0.69% | +3.35% |
| sports | 272 | 39.7% | -0.69% | +12.70% |
| news | 1,874 | 32.3% | -0.71% | +3.37% |
| anime | 169 | 41.4% | -0.72% | +3.34% |
| politics | 486 | 34.2% | -0.76% | +3.40% |
| gaming | 162 | 37.0% | -1.08% | +10.21% |
| memes | 293 | 23.9% | -1.19% | +2.47% |
| finance | 171 | 31.0% | -1.35% | +5.44% |
| deals | 66 | 33.3% | -1.43% | +5.71% |
| crypto | 1,952 | 14.1% | -3.02% | +1.44% |
Crypto is the largest tracked category at 1,952 channels, 4.5% of the entire trajectory panel, and it is worse than every other category by a wide margin on every measure in the table. Only 14.1% of crypto channels grew, roughly 275 out of 1,952. The median crypto channel lost 3.02% of its subscribers in 52 days, more than double the next worst category. The most damaging detail sits at the other end of the distribution: crypto's 90th percentile is +1.44%, the lowest P90 in the table and the only one below 2%. The next lowest is memes at +2.47%, and most categories land between +3% and +13%. Crypto does not merely have a bad median. Its upside tail is missing. In every other category a buyer can at least hope to land in a top decile that grew several percent in seven weeks; in crypto the top decile barely moved.
AI is the only Telegram category with a positive median at +0.07%, with 50.3% of its 165 channels growing. Programming and music are essentially flat. At the other end, memes stands out for a different reason than crypto: only 23.9% grew, but the median loss of 1.19% is a third of crypto's, describing a category where almost everyone drifts down slowly rather than one in retreat.
Compounding the 52-day medians out to twelve months, which we do by raising the 52-day factor to the power of 7.02, gives crypto at about -19% a year, deals at -9.6%, finance at -9.1%, memes at -8.1%, gaming at -7.3%, politics at -5.2%, news at -4.9% and AI at +0.5%. At the crypto rate a channel keeps about 52% of its subscribers after three years. We compound medians only: extrapolating a 90th percentile 52-day run out to a year is not defensible and we do not do it. If you shop in this vertical, compare individual channels against the category median using the live crypto Telegram directory, and see what the listings themselves look like on our crypto Telegram channel page.
What separates the categories that grow from the ones that bleed
The pattern across both platforms is consistent enough to state as a rule: categories organised around a recurring practical need hold their audiences, and categories organised around a speculative cycle do not. AI, programming and tech sit at the top of the Telegram table; tech and news are the only positives on X; crypto, finance, deals and memes anchor the bottom on Telegram, and crypto is last on X too.
The mechanism is churn, not interest. A crypto channel acquires subscribers during a price move, when acquisition is cheap and abundant, which is exactly when it is being pushed furthest above its sustainable equilibrium. When the move ends, acquisition collapses toward zero while the churn term keeps grinding against a base that was inflated by a temporary condition. The channel does not have to do anything wrong. The equation does the rest.
What makes this economically interesting is that the market has not repriced it. Crypto is the most expensive category to be in by several other measures in our data: 34.0% of crypto X accounts carry Blue, the highest adoption of any category in a 2% random sample of our 17.6 million account X index, and crypto carries the highest median follower count of any category on Bluesky at 440 despite being only the 13th largest there. Across 85 crypto X listings on our marketplace the median asking price works out to $34.92 per 1,000 followers, comfortably mid-pack. The market charges a normal price for the fastest-depreciating asset in the dataset. That gap is the actionable part of this study.
What follower decay does to price: turning a decay rate into a discount
A decay rate is only useful if you can convert it into a number you subtract from an asking price. There are two correct ways to do that and they give different answers, because they answer different questions.
If you are buying to resell, what matters is the follower count on the day you exit, because the resale price tracks the size then, not the size now. The haircut equals the full projected decay over your holding period. Buy at 100,000 followers, hold twelve months at a 19% annual decay, and you are selling roughly 81,000 followers at whatever the unit price is then. Decay is a straight carrying cost on your principal.
If you are buying for cash flow, meaning you intend to monetise the audience continuously, what matters is the average audience across the period rather than the endpoint. That number is better than the endpoint, because you own the larger audience for the earlier part of the period. As a rule of thumb, the cash-flow haircut lands at about half the resale haircut over a twelve-month horizon.
| Annual decay rate | Audience left after 12 months per 100,000 | Resale haircut | Cash-flow haircut |
|---|---|---|---|
| 0.0% (stable) | 100,000 | 0.0% | 0.0% |
| -0.4% (X panel median) | 99,600 | 0.4% | 0.2% |
| -3.8% (Telegram median) | 96,200 | 3.8% | 1.9% |
| -4.9% (Telegram news median) | 95,100 | 4.9% | 2.5% |
| -8.1% (Telegram memes median) | 91,900 | 8.1% | 4.1% |
| -19.4% (Telegram crypto median) | 80,600 | 19.4% | 10.0% |
That table is derived arithmetic, not measured data. It assumes decay is smooth, that value accrues evenly across the year, and that the price per 1,000 followers is unchanged at exit. The last assumption is the shakiest, since a category in decline usually sees unit prices fall too, which compounds the loss rather than offsetting it. Treat the table as a floor on the discount you should ask for, not a ceiling.
The practical version fits in one sentence: subtract the projected twelve-month decay from the headline follower count, then run your usual price-per-thousand math on the reduced number. The free growth simulator lets you project a trajectory forward so you can see the size you would actually own at exit.
Three worked valuations at real marketplace prices
All prices below are anchored to aggregate data from our live listing database on July 31, 2026. Across 345 X listings the median asking price is $100, with a 25th percentile of $25 and a 75th percentile of $300. Across 32 Telegram listings the median is $50 at a median $29.94 per 1,000 subscribers. Thirty-two listings is a thin sample and we say so rather than dressing it up.
A 40,000-subscriber Telegram crypto channel asking $1,200
That is $30.00 per 1,000, essentially the marketplace median unit price for Telegram. The category median decay is -3.02% over 52 days, about -19.4% a year. Twelve months out the channel holds roughly 32,240 subscribers, worth $965 at the same unit price. Your negotiating position is therefore $965, the asking price with 20% removed, unless the seller can demonstrate this specific channel is one of the 14.1% that grew. Ask for the 90-day history. If it is not forthcoming, the decay assumption stands and so does the discount.
A 150,000-follower X tech account asking $5,000
That is $33.33 per 1,000, close to the $34.61 median for the 100K-plus tier across 21 listings, which again is a thin cell. The tech category median is +0.171% over 52 days, about +1.2% annualised, and 60.6% of tracked tech accounts grew. The decay discount here is approximately zero and arguing for one would be wrong. The risk in this deal is not depreciation, it is variance: 39.4% of tracked tech accounts still shrank, and a 33-account sample is too small to price confidently. The right response is not a discount, it is a 14-day observation window before funding. Watch the count daily, and if it trends down against a category whose median is positive, walk or reprice.
A 120,000-subscriber Telegram news channel asking $3,500
That is $29.17 per 1,000. News is the second largest Telegram category at 1,874 channels, 32.3% grew, and the median change is -0.71% over 52 days, about -4.9% a year. Twelve months out the channel holds roughly 114,100 subscribers, worth about $3,329. Over three years it holds roughly 103,300 subscribers, worth about $3,012, a 14% erosion of the purchase price before anything else is considered. Layer on the tier data, since a 120,000-subscriber channel sits in the band that shrinks on 74.0% of days, and the case for paying the full ask gets thin. Cross-check against our Telegram channel valuation guide, which handles the view rate and cadence side of the same calculation.
Bluesky proves a network can grow while its accounts do not
The obvious objection to everything above is that these are mature platforms and a growing network would look different. Bluesky is the test case, since it added users at enormous speed over the period our index covers. Our Bluesky index holds 3,927,915 accounts, and the cohort view is instructive.
| Created | Accounts | Median followers | P90 followers | Median posts |
|---|---|---|---|---|
| 2022 | 46 | 787 | 34,228 | 208 |
| 2023 | 855,317 | 263 | 2,048 | 134 |
| 2024 | 2,095,736 | 139 | 1,146 | 36 |
| 2025 | 666,772 | 65 | 561 | 11 |
| 2026 | 139,069 | 26 | 214 | 7 |
The 2024 cohort alone is 2,095,736 accounts, more than the 2023, 2025 and 2026 cohorts combined. The network grew spectacularly. And yet the median follower count roughly halves with every successive cohort: 263, then 139, then 65, then 26. Median posts collapse in parallel, from 134 for 2023 signups to 7 for 2026 signups. Each new wave of arrivals is smaller and quieter than the one before it.
Be careful about what this proves. It is a cross-section, not a trajectory, and newer accounts have had less time to accumulate followers, so part of the decline is simply age. We cannot separate the two effects with a cohort table, and anyone who tells you otherwise is overselling. What it does establish is the point that matters for buyers: the headline growth of a network does not flow through to the median account inside it. A platform can add two million users in a year while the typical account on it stays tiny. If you are weighing an early position on a new network, our Bluesky by the numbers study covers what those accounts actually sell for, and the Bluesky directory shows the live distribution.
How to check follower decay before you buy: a seven-step procedure
Turning the finding into a workflow. Steps 1 through 7, in order, before any money moves.
- 1. Ask for 90 days of history, not a screenshot of a total. A follower count is a state; you need a series. Analytics exports, a screenshot history, or third-party tracking all work. A seller running the account seriously will have something.
- 2. Take your own baseline and let it run 14 days. Record the count when you first make contact, then again at day 7 and day 14. Two weeks is the minimum that beats daily noise, given that a healthy large X account loses followers on 63.9% of individual days. Escrow makes the wait free: nothing is committed until you fund, so you can negotiate for two weeks without exposing money. See how PlayerSells escrow works.
- 3. Compare the trend against the category median in this article. A Telegram crypto channel that is flat over two weeks is performing well above its category. A Telegram AI channel that is flat is performing below its category. The same raw number means opposite things depending on the vertical.
- 4. Check audience quality, not just size. Run the handle through the follower audit and the engagement calculator. An account whose engagement sits far below its size band usually has a dormant tail, and a dormant tail is future decay that has not been recognised yet.
- 5. Look up comparable accounts in the directory. Our public hubs run on the same crawler index behind this study, so a comparable set is one click away. Start at the Telegram channel directory and find three channels of similar size in the same category.
- 6. Price the decay in explicitly. Use the discount table above and state the number to the seller as a calculation rather than a haggle. A seller who understands the arithmetic will often accept it, and a seller who refuses to engage with it is telling you something.
- 7. Treat these four patterns as red flags. A seller who will not agree to a 14-day window. A count that is flat to the exact digit across weeks, which usually means a stale screenshot rather than a stable account. A vertical step in the last 60 days that has not consolidated. And a subscriber count that has held while the view rate has fallen, which is decay that has already happened and has not yet reached the number you are being asked to pay for.
What a shrinking account is worth to the person selling it
If you are on the sell side, this data is better news than it looks. Roughly 63% of the market is shrinking, which makes documented growth a genuinely scarce attribute, and scarce attributes command premiums. If your account is in the 36.7% of X accounts or 37.7% of Telegram channels that grew, the highest-return thing you can do before listing is prepare the evidence: a clean 90-day trajectory, exported or screenshotted, presented up front rather than produced under questioning. Most sellers cannot produce it, and that asymmetry is worth real money.
If your account is shrinking, the arithmetic argues for selling sooner rather than later, and the marketplace mechanics support that. Across 144 completed deals in our database the median time from deal creation to completion is 24.0 hours, with a 90th percentile of 70.8 hours. Liquidity is not the constraint on your timing. Decay is. A Telegram crypto channel held another six months at the category median loses about a tenth of its subscribers, and at an unchanged unit price that is a tenth of your sale value gone while you decide.
Do not conceal the trend. A buyer running the checks above will find it, and finding it after you denied it converts a price negotiation into a trust problem, which is a far more expensive conversation. Price the decay in yourself, say what it is, and let the fair price be the fair price. Set your starting number with the valuation tool and the method in our X account valuation guide, then adjust for your own measured trajectory. Listing is free from the seller dashboard.
What a 52-day window cannot tell you
The strength of this dataset is breadth: 6.1 million observations is enough that the medians will not move much. Its weakness is depth in time, and being honest about that is the only way the breadth is worth anything.
Fifty-two days is about 1.7 months. The window runs from June 9 to July 31, northern hemisphere summer, when engagement is seasonally softer in several categories. We cannot separate a seasonal dip from a structural decline with a single window, and will not know which we measured until we have a full year of comparable data. If the categories that look worst here are simply the most seasonal, some of that decay reverses in the autumn.
Second, this is a statement about medians and distributions, not about any individual account. A single viral post, an enforcement action, a mention from a large account, or a purge of inauthentic followers can move one account by more than an entire category median. Nothing here predicts what your account will do next month. It describes what the typical account did last month.
Third, there is a survivorship issue that biases our numbers toward optimism. The trajectory panels only include entities with at least 30 tracked days. Accounts deleted, suspended, or made private partway through the window fall out of the panel entirely. The true decay across everything that existed on June 9 is therefore worse than what we report, not better. We cannot say by how much.
Fourth, a negative median does not mean a category lost audience in aggregate. Medians and totals are different things, and the X tech P90 of +32.49% shows how a small number of accounts can absorb enormous gains while the typical account in the same category drifts down. If you are measuring the health of a vertical rather than the risk on a purchase, the median is the wrong statistic.
Finally, a note on the marketplace cuts used in the pricing sections. Our listing data shows accounts sold with the original email attached at a lower price per thousand followers than accounts without it, and older accounts at a lower price per thousand than newer ones. Neither is a causal discount. Both are composition effects, because sellers holding original emails and sellers of older accounts both tend to list larger accounts, and price per thousand falls as size rises. The relationship between age and value is real, but it works through different channels than a raw unit-price table shows, and our piece on why account age beats follower count makes that argument properly. This study is the data proof underneath it: if follower counts decay and account age does not, then age is the more durable half of what you are buying.
Frequently asked questions
Why am I losing followers on X even though I post every day?
Because losing followers on most days is the normal condition of an X account at scale, not a symptom of a problem. In our panel of accounts above 100,000 followers, 63.9% of observed days ended with a net loss and the median day was -23 followers. Posting frequency does not exempt you: Telegram channels above 100,000 subscribers post ten times more often than channels below 1,000 and still shrink on 74.0% of days. Watch your 30-day trend, not any individual day, and investigate only if that trend is negative while your category median is positive.
Is my Telegram channel dying if it loses subscribers every day?
Probably not. A channel between 10,000 and 100,000 subscribers loses subscribers on 65.2% of days, and above 100,000 the figure is 74.0%. Daily losses are the baseline. The meaningful comparison is your 52-day change against your category: a crypto channel at -0.55% is dramatically outperforming its category median of -3.02%, while an AI channel at the same -0.55% is underperforming a median of +0.07%. Also check your view rate. A subscriber count that holds while views fall is the early warning; a count that falls while views hold is usually dormant accounts being cleared out.
Do all social media accounts lose followers over time?
No, but most tracked accounts did over our window: 63.3% of X accounts and 62.2% of Telegram channels with at least 30 days of history. Roughly 37% grew on both platforms, and the growth categories are identifiable rather than random. Telegram AI channels had a positive median, and X tech and news accounts had positive medians. The useful framing is that decline is the base rate and growth is the exception that has to be demonstrated with evidence.
How many followers is it normal to lose per day?
It scales with size. On Telegram the median daily change is 0 below 1,000 subscribers, -1 between 1,000 and 10,000, -5 between 10,000 and 100,000, and -65 above 100,000. On X, among accounts above 100,000 followers, the median daily change is -23. Those are medians of a skewed distribution, so on a good day the same account might add hundreds: the 90th percentile day is +226 for large Telegram channels and +778 for large X accounts.
What is a good follower growth benchmark in 2026?
Beating your category median is the benchmark that means something, and in most categories that median is negative, so flat is above average almost everywhere. On Telegram, a 52-day change above -0.55% beats the typical tracked channel and above +0.07% beats the best-performing category. On X, in our large-account panel, anything above -0.06% beats the typical tracked account. Compare against category first, then against tier. A benchmark from a growth-marketing blog that assumes every account grows is not measuring the same world these numbers came from.
Should I buy an account that is shrinking?
Yes, at the right price. Most accounts on the market are shrinking, so refusing to buy any of them means not buying. The correct response is to quantify the decay and subtract it: project the audience out to your intended exit, reprice at that size, and offer accordingly. The account to avoid is not the shrinking one, it is the one whose seller will not let you measure the trend before you fund. That refusal is the signal, not the decline.
How do I know if a seller's growth screenshot is real?
Take your own measurement instead of relying on theirs. Record the public follower count yourself at day 0, day 7 and day 14, then compare the deltas against the screenshot you were given. If the account grew 4% last month by their evidence but is flat across your two weeks, the screenshot describes a moment rather than a trend. A step change in followers with no matching change in engagement is the standard signature of an inflated count, and inflated counts decay faster than real ones.
Where to go next
Start by measuring rather than assuming. Pick the account or channel you are considering, take a baseline count today, and set a reminder for day 7 and day 14. While that runs, price it twice: once at today's size and once at the size the category median projects for your exit date. The gap between those two numbers is your negotiating room, and it is wider than most sellers expect. Browse live listings on the marketplace, and if you are on the other side of the trade, prepare the trajectory evidence before you list, because in a market where 63% of accounts are shrinking, proof of growth is the most underpriced asset on the table.
Related Articles
Continue learning with these related guides.
Telegram Channel Benchmarks 2026: View Rates, Size, and Growth from 2.4 Million Channels
Telegram view rate benchmarks from 2.4 million indexed channels: 31.3% median at 1K-5K subscribers, 3.3% above 500K, plus 52-day growth by category.
X Handle Marketplace Explained: What X Charges, and Why You Still Cannot Resell Yours
X sells handles from five to seven figures but bans you from reselling yours. Here is what the license actually says, and why the account is the real asset.
