Comparing three panels fairly on the same order

Three panels, one order each, the same service and the same quantity. It sounds like a fair test. It usually is not, because the three orders run on different links, at different times, against different baselines, and the buyer reads a difference in the queue as a difference in quality. The result is a ranking built on three unrelated events.

This guide sets up a trial that can actually answer a question. The question is not which panel to use in general. It is which of these three panels can take one order, deliver it, explain a shortfall and answer a question, on the same service, in the same window. That is a smaller question, and it is the one a buyer can answer with three minimum orders.

What one order can prove

One order per panel can prove that a panel can accept an order, start it, deliver the bulk of it and settle the rest. It can prove that support answers a pre-sales question and that the order page shows a state. It cannot prove a drop rate, because one order is one observation. It cannot prove a start time, because the queue on that day is not the queue next week. It cannot rank three panels by quality, because the differences you are trying to measure are smaller than the noise in a single delivery.

What it can do is eliminate. A panel that never starts, delivers less than the order and refuses to settle the difference, or ignores the pre-sales question has failed a basic step. That failure is evidence about a process, and it is enough to stop before a larger order on the same row. Elimination is a modest goal, and it is the right goal for a three-order test.

Hold six things constant

A fair trial compares like with like. Six fields decide whether two orders are the same product. If any of them differs, the comparison is measuring the field, not the panel.

Hold constant	Why it changes the result
Platform and category	Followers and views have different pools and different queues
Quality tier	A standard row and a real row are different products
Country	A country-targeted pool costs more and behaves differently
Refill window	A replacement promise is part of the product
Quantity	The effective rate depends on the quantity and the minimum
Link format	A profile link and a post link are different delivery targets

A comparison checklist. It contains no market figures.

The link format is the field buyers forget. A follower order on a profile and a follower order on a post are not the same product, even when the service name is the same. A view order on a long video and a view order on a short are not the same product either. Write the full variant before you place the first order, then copy it into the other two.

The six records

For each panel, record six things. The record is the trial. Without it, the memory of the delivery replaces the delivery.

Record	Panel A	Panel B	Panel C
Charge and currency
Paid at
Started at
Delivered quantity
Count at 24 hours
Count at seven days
Refund or refill outcome
Support reply time

A scorecard to copy before the first order. It contains no market figures.

Add the order id and the start count for each panel. The order id is the only key that lets you compare a support answer with a delivery. The start count is the only baseline that lets you compare a count at seven days with the count before the order. If you skip either, the record is a story rather than a measurement.

What the market spread tells you

Before you run the trial, look at the spread in the market table. It tells you how large a difference the market contains, and therefore how small a difference one order can resolve. When the market table has those products, the table below shows them at the time this page is rendered: the number of panels quoting each, the median rate, the minimum listed rate and the gap between them.

Product	Panels	Median	Minimum	Spread
Instagram followers	28	$0.735	$0.2957	59.8%
TikTok views	26	$0.0205	$0.0048	76.6%
YouTube views	29	$0.3675	$0.161	56.2%
Telegram members	24	$0.2117	$0.0601	71.6%

Rates are per 1,000 units in USD, read from public panel price lists when the page is rendered. Market edition: 7 November 2026.

Price 5,000 Instagram followers at the median and at the minimum listed rate: $3.675 against $1.4785. The gap is $2.1965, which is 59.8% of the median charge. That is the size of the gap the market contains. If your three orders land within a few percent of each other, the trial has not found a quality difference; it has found three panels that price the same product similarly.

The spread is also a warning about single observations. A delivery that starts late can be a queue, a supplier top-up or a platform limit. A delivery that drops can be a cleanup or a pool that was already fragile. One order cannot separate those causes. The trial is a screen, and the spread is the reason to keep the screen small.

Reading the result: eliminate, do not rank

Set the pass marks before the first order. Write them down, so the result cannot move to fit the panel you liked. A workable set is: the order starts within the row's estimate, delivers at or above the quantity you ordered, the count at seven days stays above the start count, and support answers the pre-sales question before you pay. Those are four yes or no checks. A panel that fails one is a candidate for elimination on that row.

The pass marks are yours, not the market's. If the account is a business asset, a drop below the start count inside the refill window is a fail even if the panel offers a refill, because the refill is a second order and a second wait. If the account is a test, a small drop is acceptable and the settlement process matters more than the count. Write the rule for the job, not for the panel.

Do not average the three results into a score. A panel that starts quickly but delivers short is not the same as a panel that starts slowly and delivers in full. The first has a queue problem; the second has a speed problem. A single number hides the difference and produces a ranking that cannot be used for the next order.

When two panels pass

A tie is the normal result of a three-panel screen. Two panels accept the order, deliver it and answer the question, and the difference between them is smaller than the noise in one delivery. The tie is not a failure of the trial. It is the trial telling you that both panels can do the basic job, and that the decision now belongs to a different set of questions.

Break the tie with three things, in this order. First, settlement: which panel refunds a short delivery without a ticket, and how quickly the credit appears. Second, refill execution: which panel can name the window, the executor and the count the refill is measured against. Third, queue transparency: which panel can tell you the current depth and the timeout before an automatic refund. The rate comes after those three, because the rate is the part of the deal you can see before you buy and the other three are the parts you discover after.

Do not break a tie with a review page. A review is a claim about a delivery you did not see, written by a buyer whose variant, quantity and link may have been different from yours. The trial gave you three deliveries you did see. Use them, and keep the scorecard for the next decision.

Cost after refunds

Compare the delivered cost, not the headline charge. If a panel delivers part of an order and refunds the rest, the cost of the delivered units is the charge minus the refund, divided by the delivered quantity, multiplied by 1,000. That number is comparable across panels. The headline rate is not, because it assumes every unit arrived.

The arithmetic is the same as the price guide, with one extra step. Price the order at the quantity you bought. Subtract the refund. Divide by the delivered quantity. Multiply by 1,000. If a panel does not refund the missing share, the delivered cost is the full charge divided by the delivered quantity, which is a higher number than the headline rate suggests. That difference is the price of a weak settlement policy, and it belongs in the comparison.

The worked example in the refund guide uses the same logic: a short delivery has a delivered cost based on the units that arrived, not on the units that were ordered. The panel that settles the shortfall is less expensive at the same headline rate, because you are paying for what arrived.

The same-link mistake

Running three orders on one link at the same time is a common way to ruin a trial. The counts overlap, so a change cannot be attributed to one panel. The platform sees a spike that no single order explains, which adds a risk that has nothing to do with the panel's quality. And if the link's count drops, you cannot tell which delivery dropped.

Use three comparable links, or stagger the orders by the delivery window. Comparable means the same platform, category, quality and country, with a similar starting count. If you cannot find three comparable links, run the trial on one link across three weeks, one panel per week, and record the baseline each time. That is slower, and it is still a fair test.

A five-minute setup

1. Write the variant. Platform, category, quality, country, refill window, quantity. One line, copied to all three orders.
2. Choose three comparable links. Similar start count, same format, no order running on any of them.
3. Set the pass marks. Start, delivery, seven-day count, support reply. Write them before you pay.
4. Record the start count and a screenshot. For each link, before the first order.
5. Place the orders. One per panel, same quantity, same window if possible.
6. Fill the six records. Charge, paid, started, delivered, 24 hours, seven days. Add the order id and the support reply.
7. Eliminate, do not rank. Keep the panels that passed. For the ones that failed, write the failed step and the order id, then move on.

The trial is small on purpose. It costs three minimum orders and one week, and it replaces a ranking built on impressions with a record built on orders. The panel that passes is not proven good. It is proven able to do the basic job on the row you tested, which is the only claim three orders can support.

For the single-order version of the same test, read how to test a new panel with a minimum order. For the arithmetic behind the charges, read the price formula. To see the spread the trial is trying to resolve, open the public market tables.