Why Followers Drop and How Refill Guarantees Really Work

You placed an order, the number went up, and a week later 180 of those 1,000 followers are gone. That is what this guide is about: follower drop. Almost everyone who uses a social media marketing panel runs into it eventually, and when they do they ask two questions. First, why did this happen. Second, what happens to my money. The honest answer to both is rarely printed on panel websites, because part of it is uncomfortable from a sales point of view. The uncomfortable part is in here.

The short version: drop is not a malfunction. It is the system working as designed. Instagram, TikTok, YouTube and every other platform periodically delete fake, automated and long-dormant accounts. Meta's sweep on the night of 6 to 7 May 2026, widely reported as the "Great Purge of 2026", was the most visible example yet. Millions of accounts disappeared in a single six-hour window. Most small and mid-sized creators saw roughly a 2 to 5 percent decline, while the largest accounts lost followers in the millions. Crucially, that purge hit accounts that had never bought a thing. So purchased followers are not the only cause of drop, but they are the layer most exposed to it.

The answer to the second question depends entirely on the service you bought. Whether a drop gets replaced comes down to one attribute: does that service support refill or not. On a refill-enabled service, the missing amount is re-delivered when you file a request inside the guarantee window. On a service with no refill, meaning a non-guaranteed service, a drop is not refunded. That is not a policy preference, it is a technical consequence of where those followers come from, and it is written plainly in our terms of service. It is also the main reason cheap services are cheap.

This guide walks through the mechanics of drop, where the quality gap actually comes from, what drip-feed solves and what it categorically does not, what a refill guarantee is, and more importantly what it is not. By the end, even if you never buy anything, you should be able to read a service description and understand exactly what it promises and what it quietly does not. Every listing in the service catalogue reads very differently once you have that lens, and creating a free account keeps your order history and refill requests on one screen.

What drop is, and what it is not

Drop means a delivered order losing count over time. You received 1,000 followers, three weeks later 870 remain, so 130 dropped. Simple enough, except that what people call "drop" is at least four different events, and if you cannot tell them apart you cannot diagnose anything.

Fluctuation during delivery is not drop. While an order is processing, the number moves in both directions. Platform counters are cached, not real-time. Measuring anything while an order sits in Processing is meaningless. Measurement starts the moment the order status reads Completed.

Misreading the start count is not drop. The system records the count at the moment of ordering as the start value. If you are simultaneously losing followers organically, or you placed a parallel order on another panel, you cannot bill the sum of both movements to one service. Ordering the same metric from multiple sources at once makes measurement mathematically impossible.

Platform cleanup is drop, and it is the most common kind. When an account is judged fake, automated or long dormant, the platform deletes it. Your follower disappears, your count falls. This happens to purchased followers at a far higher rate than to organic ones, because the source pool is more exposed by nature.

Source pull-back is also drop. On some low-cost services the follow action is not durable; the accounts unfollow after a while. That is the source's behaviour rather than the platform's, and it is the dominant failure mode at the bottom of the price range.

The distinction matters because a refill only compensates a genuine drop. It does not compensate a measurement error, your own organic churn, or an order you placed somewhere else entirely.

How much drop is normal?

The honest answer: anyone quoting you a fixed number is inventing it. The rate depends on the service, the platform, the state of the account and whatever purge wave happens to be running that month. The broad industry expectation is that well-sourced services stay in single-digit percentage loss over 30 days, while the cheapest tiers can shed 30 percent, 50 percent, or more in a bad month. Those are observation ranges, not promises. The table below is for calibrating expectations, not a commitment.

Service tier	Typical 30-day behaviour	Refill support	Who it suits
Non-guaranteed / cheapest	High, unpredictable loss	None, drops are not refunded	Short-term, one-off tests
Standard refill	Loss happens, replaced within window	Usually 30 days	Reasonable balance for most buyers
Extended guarantee	More stable source, less loss	90 days or more, stated per service	Accounts that need continuity
Labelled "non-drop"	Most durable source, still not zero	Depends on the listing	Critical accounts, higher budget

Pay attention to that "non-drop" label. The industry uses the word as if it meant "never falls", but its real meaning is "sourced to minimise loss". No service, on any panel, can honestly promise permanence. That promise does not belong to the provider, it belongs to the platform's deletion decision, and no provider controls it.

The six real causes of follower drop

1. Periodic bot and dormant account cleanup

This is the single largest cause. Platforms delete fake accounts for two reasons: advertisers pay against real reach, so the metrics have to be clean, and regulatory pressure makes inauthentic engagement increasingly expensive to tolerate. Meta's May 2026 sweep was driven by exactly those two forces, and it used an AI moderation system hunting for coordinated inauthentic behaviour. TikTok operates at a comparable scale: according to the platform's own transparency reporting, it blocked over 700 million fake account creations and prevented more than 36 billion fake likes in the first half of 2024 alone. No source pool is permanent against a machine operating at that scale.

2. The quality of the source pool

Price and durability are almost linearly related. A service selling 1,000 followers for pennies has to source those followers somewhere, and that somewhere is usually a pool of freshly created accounts with no profile picture, no bio and no posts. Those are the easiest possible targets for a detection model. More expensive services draw on older, more "lived-in" account pools that survive longer. This is where the real cost of hunting for the cheapest panel shows up: the unit price falls, and so does the durability per unit.

3. The state of your own account

The same service behaves differently on two different accounts. Followers delivered to a brand-new profile with three posts and no engagement erode faster than the same followers delivered to a three-year-old active account. Platforms evaluate every signal around an account together. The wider the gap between follower count and content engagement, the more scrutiny the connections around that account attract.

4. Multiple simultaneous orders

Sending followers to one account from three panels on the same day makes measurement impossible and produces a sharp, unnatural spike. The sharper the spike, the easier the detection. One source, one plan, one measurable result.

5. Username changes or going private

Making the account private mid-delivery, changing the username or temporarily deactivating breaks delivery. Losses caused this way generally fall outside refill coverage, because the failure is at the target rather than the source. As explained on the how it works page, the account has to stay public and stable for the duration of the order.

6. Time itself

Refill or no refill, purchased followers are not permanent. Look at any pool a year later and a large share of it has eroded. That is not the mark of a bad service, it is the nature of purchased engagement. Panel services are not a real, organic audience; they move a counter, they do not manufacture fans. Accepting that up front beats discovering it later.

How platform bot cleanup actually works

Understanding this is understanding drop. Platforms do not inspect accounts one by one and rule on each. They work on signal clusters, and they flag patterns rather than individuals.

The signals typically weighed include account creation date and activity density since, profile completeness (picture, bio, posts), the ratio between following and followers, how follow actions distribute over time, bulk activity from a shared IP or device fingerprint, and above all coordination. If hundreds of accounts follow the same profiles inside the same minute window, that is not individual behaviour. Meta calls this coordinated inauthentic behaviour, and it is precisely what the AI system used in the 2026 sweep was looking for.

Three practical conclusions follow:

Cleanups arrive in waves, not continuously. Nothing happens for months, then a mass deletion runs overnight. This is why "it has not dropped in two months, so it must be permanent" is a false inference. The wave simply has not arrived yet.
Your account is not the target, the source accounts are. What gets deleted is the account following you, not yours. That is why cleanup usually surfaces as follower loss rather than as a penalty.
No provider can promise to beat that machine. Anyone who does is lying. The only two things a provider can do are use a more durable source and replace what falls. The second one is called refill.

Do these services violate platform rules?

The honest answer is yes, they can. Instagram's terms of use and community guidelines explicitly prohibit artificially collecting likes, followers or shares. TikTok's integrity and authenticity policy prohibits services that artificially inflate engagement or game the recommendation system. YouTube's fake engagement policy prohibits artificially increasing views, likes or subscriber counts and states that channels which break it may be terminated. A panel that does not tell you this is withholding the single most basic fact about what it sells.

In practice, enforcement weight falls on deleting the source accounts rather than banning the buyer, but that is an observation, not a guarantee, and nobody can give you an absolute assurance about account safety. The purpose of this section is not to talk you out of anything. It is to make sure the decision is an informed one. If you find the risk acceptable, at least know what the risk is. Our frequently asked questions take the same line.

Low-quality versus high-quality followers

Two panels both list "1,000 Instagram followers" and one charges five times the other. The difference is almost always the source. Quality here is not an abstract marketing word, it is the sum of a few measurable properties.

Account age. A pool of freshly created accounts is cheap and short-lived. Accounts created months or years ago, with a history behind them, are expensive and last longer. Account age is one of the strongest signals a detection model has.

Profile completeness. An account with no picture, an empty bio and zero posts is trivially easy to classify. Filled-in profiles cost more because preparing them takes work.

Geography. Accounts from a specific country cost more and matter for audience coherence. Mixed, geographically undefined pools are the cheapest thing on the market.

Source behaviour. Some pools unfollow on their own after a while. That is a separate event from platform deletion, and it dominates the bottom price tier.

Property	Economy service	Quality service
Account age	New, days or weeks	Months or years
Profile	No picture, empty	Picture, bio, sometimes posts
Durability	Low, erodes when a wave hits	Higher, still not zero loss
Refill	Usually none	Usually 30 days or more
Unit price	Lowest	Three to ten times higher

Reading that table and concluding "so always buy expensive" is also wrong. The right call depends on the use case. If you need 48 hours of visibility for one campaign, the economy tier can be perfectly rational. If you want a number that stays for months, buying the economy tier means paying twice: once for the service, and again when it falls. Keep that arithmetic in mind when you look at the price range for Instagram follower services.

A glossary for reading quality labels

These labels are not standardised, and their meaning drifts from panel to panel. The general tendency:

Real / HQ / Premium: Older, fuller profiles. It does not mean "real humans". It means "more convincing accounts".
Non-drop: A source engineered to minimise loss. Not zero loss.
Refill 30D / R30: A 30-day refill guarantee. When the window closes, the guarantee closes; the number is not protected afterwards.
No refill / Non-guaranteed: If it falls, there is no replacement and no refund. This is the cheapest tier.
Instant / Fast: Refers to start speed, not quality. Fast does not mean durable.

Look for these in the service description, not the service name. The name is marketing, the description is the contract.

What drip-feed is, and what it does not fix

Drip-feed delivers an order in scheduled portions rather than all at once. Say 1,000 followers spread over ten days at 100 per day. Panels typically ask for three inputs: the amount per run, the number of runs, and the interval between runs in minutes.

A worked example: 1,000 followers total, 100 per run, 10 runs, 1,440-minute interval. Result: 100 followers a day for ten days. The total is amount per run multiplied by run count, and you are charged against that total.

Scenario	Per run	Runs	Interval (min)	Result
New account, cautious	50	20	1440	1,000 over 20 days
Standard spread	100	10	1440	1,000 over 10 days
Campaign support	250	4	720	1,000 over 2 days

What drip-feed fixes: the abrupt, unnatural spike. An account with 200 followers jumping to 10,000 in an hour looks strange both to your visitors and to pattern analysis. Spreading the same volume across three weeks at least removes the extreme edge on the time axis. There is a second benefit too: the count climbs gradually while delivery runs, which reads as less suspicious to a human looking at your profile.

What drip-feed does not fix: source quality. This is the most common misunderstanding in the entire category. Drip-feed is a pacing control, not a quality upgrade. If the source pool is weak, spreading those same accounts over ten days does not make them durable. When the cleanup wave arrives, those accounts get deleted regardless of how slowly they were delivered. Combining drip-feed with a cheap non-guaranteed service and expecting permanence is the most expensive mistake in this industry.

One more thing: do not try to measure an order while drip-feed is still running. Comparing start and current counts before completion produces nothing meaningful. On most panels you also cannot file a refill request against an order that is still delivering; the request only makes sense once the order is complete.

What a refill guarantee actually is

A refill is the free re-delivery of the amount lost from a completed order, within a defined window. The mechanism works like this.

At the moment you order, the system records the target account's current count. That is the start count. When the order completes, the target value equals start count plus order quantity. When you file a refill request, the system reads the current count, compares it against the target value, and re-sends the shortfall. The critical point: a refill tops your order back up, it does not enrich you. If you bought 1,000 and 200 dropped, refill sends 200, not another 1,000.

The window is a property of the service itself. Some carry 30 days, others 60, 90 or longer. It starts counting from the moment the order completes. Once it expires, that order's guarantee is over and no request can be filed for later losses. This is not stinginess on the panel's part: the provider chain upstream runs on the same clock.

Let us be exact about coverage.

Refill covers:

Loss occurring after completion, up to the amount that service delivered
Requests filed inside the guarantee window
Cases where the target account stayed in the same state it was in at order time

Refill does not cover:

Loss after the guarantee window closes
Your own organic follower churn
Drop on an order placed through a different panel
Loss on accounts that were renamed or set to private
Situations where the count fell below the original start value, which are outside the measurable range
Services with no refill support, under any circumstances

On our panel, refill-enabled services are marked in the catalogue, and requests go straight to the provider rather than queuing for manual approval, so nothing sits on a human's desk. The same symmetry applies to cancellation: if a service supports cancellation, the request goes directly to the provider. You can see which listings support refill from the card itself in the catalogue.

A refill is not a refund

This distinction contains one of the most important sentences in this guide, so it gets its own heading. A refill re-sends the missing amount. A refund returns your money. On a refill-enabled service, the remedy for drop is re-delivery, not cash back. On a service without refill support, there is neither re-delivery nor cash back.

Once more, as plainly as it can be put: on a non-guaranteed service, drop is not grounds for a refund. A panel that softens this rule, that says "we will sort something out anyway", is either charging you more elsewhere or will not honour it. We put this in our terms in writing, because information you need before ordering should not be discovered at the moment you complain.

Non-guaranteed services: why they exist and when they make sense

If you have read this far you might reasonably ask why non-guaranteed services exist at all. There is a legitimate answer.

They are cheap because the source costs little, and they are rational in specific scenarios:

Short-term visibility. A one-day campaign, an event, a launch moment. If you do not care what the number looks like in a month, paying for a guarantee is waste.
Testing. Running a small order to observe a service's start speed and source character. A hundred-unit test is cheaper than a thousand-unit mistake.
Crossing the social proof threshold. A profile with zero followers and a profile with three digits read differently to a visitor. A small budget spent to cross that line, with no expectation of permanence, is a defensible call.

There is exactly one scenario where the non-guaranteed tier is irrational: expecting permanence. If you want the number to hold, the price advantage is an illusion. You will buy again when it falls, and the two purchases together cost more than the refill-enabled service would have. The arithmetic:

You buy the non-guaranteed service at unit cost X.
Within three weeks, half of it drops.
You buy again to hold the number, total spend now 1.5X.
The cycle repeats roughly monthly.
If the refill-enabled service costs 3X, after three months the "cheap" route has reached 4.5X and still carries no guarantee.

Saying "I bought the cheapest one" without doing that arithmetic measures the first invoice, not the cost.

How to read a service description before ordering

This may be the highest-return section in the guide. Most problems panel users experience trace back to not reading the description. It is not decorative copy, it is the contract for that specific service.

Before ordering, look for these six things:

Is there a refill, and for how many days? If not, drop is your risk. That single line explains most of the price spread in the catalogue.
Minimum and maximum quantity. On some listings the maximum is 1, which means it is a package: the price covers the whole package, not each 1,000 units. Confusing the two produces serious budgeting errors.
Start time and speed. Phrases like "starts in 0 to 1 hour, 5,000 per day" are estimates, not commitments. They move with provider load.
How the source is described. What does it say about geography, account type, profile completeness? If it says nothing, assume the lowest tier.
Special rules. Lines like "account must be public", "do not change the username", "do not order while another order is running" directly govern your refill rights.
Input format. Some services want a username, others a full link. Wrong format means delivery to the wrong target, and that cannot be undone.

You will never be asked for a password. Hold this as an absolute rule: do not hand account credentials to anything that asks for more than a public username or link. The same principle runs through our explainer on what an SMM panel is.

Delivery to the wrong target cannot be reversed

This deserves its own warning. If you mistype the link or username, the system delivers to that target, and once it completes there is no mechanism to pull it back. Those followers went to somebody else's account. This is neither a refill case nor a refund case. Reading the target field twice on the order screen eliminates this entire class of problem.

How to file a refill request, step by step

If you have spotted a drop on a refill-enabled order, this is the path. Keep the order of steps, because each one makes the next work correctly.

Check the order status. If it does not read Completed, wait. You cannot measure drop on an order still in delivery, and the refill button generally will not activate.
Confirm the service supports refill. It is stated in the order detail and in the service description. If it does not, no request can be filed and no refund is issued for drop.
Confirm the guarantee window is still open. If more time has passed since completion than the service's stated window, the right has expired.
Note the current count. Start count, order quantity, current count. Those three numbers are the basis of the request.
Find the order in your list and press Refill. The request goes straight to the provider without waiting on manual approval.
Wait. Refills are not instant. Depending on the provider's queue, they can take from a few hours to a few days. Do not file repeatedly during that time; there is a 24-hour cooldown and a second request inside it will not be accepted.
Measure again once the refill lands. If you are back at the target value, you are done. If not, and the window is still open, you can file again.

Manually processed orders, meaning those not tied to a provider, follow a different path: those requests are reviewed by the team and move through support. When a refill request is rejected, the reason is almost always one of the exclusions listed above: the window expired, the service is non-guaranteed, the count fell below the original start value, or the account changed after the order.

Details that speed up a request

When a case goes through support, supplying these up front turns a two-day thread into a two-hour one: order ID, service name, completion date, start count, order quantity, and current count. "My followers dropped" on its own triggers nothing, because there is no way to determine which order lost how much.

Drop behaves differently on each platform

People talk about follower drop as if it were one phenomenon, but platforms are independent systems and their behaviour diverges noticeably. Knowing which platform you are working on is the fastest way to calibrate expectations.

Instagram is the harshest and most visible actor here. A follow is a persistent relationship, so when an account is deleted your count falls immediately. The May 2026 sweep demonstrated the scale to everyone: overnight, with no prior notice, hitting buyers and non-buyers alike. On Instagram the drop risk is highest on followers and lower on likes and views, which makes checking the refill label on Instagram services more critical than elsewhere.

TikTok runs automated moderation at enormous scale and reports fake-engagement removals in the billions. Views tend to be relatively stable there while followers and likes are more volatile. TikTok also sometimes expresses detection as reach suppression rather than count loss: the number holds, but the content reaches fewer people. That is a real cost even though it is hard to measure. This is why slower, more even delivery is generally preferred across TikTok service categories.

YouTube has one of the strictest positions on subscribers, and its fake engagement policy states outright that channels may be terminated. Subscriber loss on YouTube tends to be slow but continuous. On the view side, the platform can recount what it considers a valid view after the fact; that is a counter correction rather than a drop, and it may sit outside refill coverage. Anyone buying YouTube services should read that distinction carefully.

Twitter (X), Telegram and Facebook are less predictable. Telegram channel membership tends to be comparatively stable, which is why Telegram services often carry longer guarantees. On the Twitter side, policy and API changes have repeatedly upended source pools in recent years, so drop rates on Twitter services fluctuate by period.

The rule that falls out of this: do not carry a result from one platform to another. A service quality that holds for three months on Instagram will not produce the same outcome on TikTok. Measure each platform separately.

Measuring drop properly: the three data point method

Most people evaluate drop by feel. They glance at the number, it looks low, they get annoyed. That method cannot compare services and cannot support a legitimate refill request. The alternative is a three-minute record-keeping habit.

On the day the order completes, note three things: the date, the count at that moment, and the order quantity. Then take only two more measurements: day 7 and day 30. Those three data points tell you the following.

Measurement	What it shows	How to read it
Day 0 (completion)	Target value	Start count plus order quantity
Day 7	Early erosion	The source pulling back on its own
Day 30	Real durability	State as the window closes, decision point

Loss by day 7 generally comes from the source pool's own character; the accounts unfollowed. Loss by day 30 is largely the fingerprint of platform cleanup. Telling the two apart directly determines which service you buy next: early erosion means a weak source, late erosion means a reasonable source hit by a wave.

The second benefit is refill timing. On an order with a 30-day guarantee, measuring on day 30 and filing on day 31 loses you the right entirely. The correct rhythm: measure around day 25, file immediately if there is a shortfall, and never wait for the last day of the window. Refills sitting in a provider queue for a few days is normal, and that wait does not extend the guarantee.

The third benefit is your buying decision. Measure two different services the same way and you will know, from your own data, which one actually holds. No panel's advertising carries that number and none can, because the outcome varies by account, platform and period. When you are comparing panels, the only trustworthy source is your own measurement log.

What purchased engagement cannot do

This section is commercially unnecessary and editorially essential. The real issue underneath every drop argument is expectation, and when the expectation is set wrong, the outcome is disappointment even if the service performed perfectly.

A purchased follower is a number. Numbers influence the snap judgement a human makes when they land on your profile, which is a genuine effect known as social proof. But there are several things a number cannot do, and they deserve to be stated plainly.

Purchased followers do not bring customers. Those accounts will not buy your product, book your service or message you. If you want sales, what produces sales remains the content and the offer.

Purchased followers do not increase reach, and often reduce it. Distribution algorithms show your content to a slice of your existing followers first, then widen based on the response. An audience of accounts that never engage corrupts that first test. If 900 of your 1,000 followers never react, your engagement rate falls, and that rate is an input to the distribution decision. This is the hidden cost most buyers never price in.

Purchased followers do not survive scrutiny. A brand checks your account before working with you. The gap between follower count and like count is the first metric anyone in the industry looks at. Official programmes like TikTok's Creator Marketplace can reject an application outright when they detect fake engagement.

Purchased followers are not permanent. Even with a refill guarantee, the guarantee covers a window, not forever.

So where does reasonable use survive? Here: getting a profile past its first visibility threshold, not looking empty on launch day, supporting social proof inside a campaign window. Those are limited, tactical, honest uses. Treating a panel service as a growth strategy, on the other hand, is like heating a thermometer to warm a room. The gauge moves, the room stays cold.

A note for resellers and agencies

If you resell to clients, drop is not a technical topic for you, it is a client-relationship topic. The only sustainable strategy is setting the expectation correctly up front.

Three things work in practice. First, do not sell "permanent followers"; put in writing what the client is buying and how long the guarantee window runs. Second, price the margin on non-guaranteed services against the cost of the complaint that follows; a margin that looks healthy on paper goes negative after two refund arguments. Third, automate measurement: querying order status and remaining quantity through the reseller API lets you see a drop before your client does. If you run your own white-label storefront, the child panel setup passes the same service data and guarantee labels straight through to the end user, and that transparency protects you.

There is one extra point on the agency side: reporting. Showing follower count as a single line in a monthly report creates an unexplainable crater when a purge lands. Reporting purchased and organic movement as separate items is both more honest and far easier to defend in a room.

Frequently asked questions

Why are my followers dropping?

Three main reasons: the platform periodically deletes fake and dormant accounts, the service's source pool consists of short-lived accounts, or something on your account changed after the order. The first is by far the most common, and it affects purchased followers at a much higher rate than organic ones.

Does a refill guarantee mean my followers will never drop?

No. A refill does not prevent drop; it re-delivers the lost amount inside the guarantee window. The drop still happens, it just gets compensated. Once the window closes, nothing can be done about later losses.

My followers dropped on a non-guaranteed service. Can I get a refund?

No. Services without refill support offer neither re-delivery nor a refund when drop occurs. That is precisely why they are cheap, and it is stated openly in our terms of service. If you need the number to hold, you have to choose a refill-enabled service.

How long does a refill take?

It is not instant. Depending on the provider's queue it can range from a few hours to a few days. Do not file repeatedly during that time; there is a 24-hour cooldown. Once the request is in, the only thing left to do is wait.

If I use drip-feed, will my followers stop dropping?

They can still drop. Drip-feed is a delivery pacing control, not a quality guarantee. It smooths the spike and makes delivery look more natural, but if the source pool is weak those accounts still get removed in the next cleanup. Drip-feed is meaningful paired with a quality service; it does not rescue a cheap one.

Can I order from several panels at the same time?

Technically yes, but do not. Parallel orders corrupt the start count measurement, make it impossible to attribute a drop to a specific order, and your refill request will be rejected on that basis. The combined spike is also sharper, which works against you.

Can these services get my account banned?

Nobody can give you an absolute guarantee, and you should distrust anyone who tries. Platform rules prohibit artificially collecting followers and likes; YouTube's fake engagement policy explicitly lists channel termination as a consequence. In practice most enforcement lands on deleting the source accounts, but the risk is not zero and you should take it knowingly.

Do I have to give my password when ordering?

No, never. Your public username or post link is enough. Do not give account credentials to any service that asks for them. This rule has no exceptions in this industry.

Conclusion

Drop is a normal phenomenon that everyone using a panel encounters sooner or later, and the cause is usually not a mistake you made. It is platforms routinely cleaning out fake accounts. Those cleanups arrive in waves, without warning, and no provider can prevent them. A provider can do exactly two things: use a more durable source, and replace what falls inside a guarantee window. The second is called refill, and it exists only on refill-enabled services. When a drop occurs on a service without refill support, there is neither re-delivery nor a refund. That is the single most important sentence in this guide, and it needs to be known before you order, not after.

The practical lesson is simple: read the description, not the price. Is there a refill, for how many days, how is the source described, what rules are attached. If you can answer those four questions, you know what you are buying and you will not be surprised a month later. A cheap service bought blind almost always costs more than an expensive service bought deliberately. Open the service catalogue and read the full description of whatever you are considering before you look at its price; that is the single most concrete habit this guide can leave you with.