How the picks and the figures are made

What is in the table

One row per portable power station currently sold new on the manufacturer's own US store. The brands covered are EcoFlow, Jackery, Bluetti, Anker SOLIX, Goal Zero, DJI, Oupes, Zendure and 4Patriots. Within each brand the rows are the mainstream sizes buyers search for, roughly 250 Wh to 6,500 Wh, with the models a brand leads with on its store page.

Each row carries the same fields: rated capacity (Wh), continuous AC output (W), surge output (W), battery chemistry, rated cycle life and the capacity it is rated to, weight, dimensions, AC outlet count, USB-C port count and highest USB-C output, USB-A count, 12 V outputs, wireless charging, wall recharge time, maximum AC charging input, maximum solar input, the stated solar recharge time, whether extra batteries attach and the maximum expanded capacity, UPS switchover time, app control, release year, warranty, list price and the current official price.

Solar panels are not rows in this table. A 'solar generator' is a power station sold in a bundle with a panel; the station is what stores and delivers the energy, so the station is the entity we track. The solar page explains what the panel adds and how the station's maximum solar input caps it.

Where each figure comes from

Every specification and price comes from the manufacturer's own product page, spec tab, spec sheet or user manual on the manufacturer's domain. Retailer listings, review sites and forums are never the source of a spec. An Amazon listing is used for one thing only: the button that goes to it.

The Sources list at the bottom of each unit page names the manufacturer page and the date it was checked for the spec fields and, separately, for the price, and names every independent review behind the unit's score. The same provenance is in the CSV and JSON exports as spec_source_url, spec_observed_at and price_observed_at. If a figure on this site disagrees with the manufacturer's page today, the manufacturer's page is right and the row needs a correction: write to the address on the contact page with the model and the field.

Prices are the manufacturer's US store list price and the price it showed on the date checked, which the footer of every page states. Manufacturers discount constantly, so a price here is a dated observation, never a promise. Open the product page before buying.

The rubric: seven criteria, scored 1 to 5

12345

Every unit is scored on the seven criteria below. Two kinds of evidence feed the scores. The first is the manufacturer's published specification, which gives the first six criteria a score by the fixed rules stated under each. The second is the consensus of independent published tests, expert reviews and owner reports, collected as described in the next section, which can score all seven. Where a criterion has both, they are combined with the sources' weight growing with their number: one source counts a third, two count half, four count two thirds, and the published figures never count less than a third. Owner-reported reliability comes from independent sources only: no published figure can speak to it.

The overall score is a weighted average: fit for the job 25%, value per watt-hour 20%, recharge speed 15%, and 10% each for portability, chemistry and cycle life, expandability and owner-reported reliability. A criterion with no score is left out and the weights are re-scaled, so a unit is never punished for a figure nobody has established. Every unit's page shows its seven scores with the one-line reason behind each.

Fit for the job

capacity, continuous output and surge against the job's load. For the general score, the share of this site's six jobs the unit clears (from 1 for none to 5 for all six), averaged with the share of the sixteen devices in the device table its inverter can start (a 700 W unit that cannot start a refrigerator compressor scores lower than an 1,800 W unit that can). For a specific job, the unit must clear the job's watt-hours, running watts and start-up watts; it then scores 5 when it holds 1.2 to 2 times the watt-hours needed, 4 when it holds less spare than that or 2 to 3 times, and 3 when it holds more than 3 times, because capacity you carry and never use is weight and money. Independent tests add whether the unit delivered its rated capacity and output in someone's hands.

Value per watt-hour

the current price divided by rated capacity, ranked against the other units in the same size class: the lowest in the class scores 5, the highest 1, the rest in between. Inside a job, value ranks by price instead, because every unit in the running already does the job and the cheaper one is the better buy. A unit with no store price has no value score.

Portability

weight on one scale for every unit, 5 at 12 lb or less, 1 at 110 lb or more, a straight line between (a 24 lb unit scores about 4.5, a 50 lb unit about 3.4, an 85 lb unit about 2). A rank would call a 90 lb unit portable because its neighbours weigh more; the scale does not. Reviews add the handle, the wheels, and whether one person moves it.

Recharge speed

maximum solar input as a share of capacity, ranked within the class, averaged with maximum wall input as a share of capacity (or one divided by the stated wall recharge hours where no input figure is stated), also ranked within the class. Reviews add whether the stated times hold up with a real panel and a real outlet.

Chemistry and cycle life

LiFePO4 scores 4, NMC or an unnamed lithium chemistry 2; a rated cycle count of 3,500 or more adds 1, of 2,500 or more adds 0.5, and under 1,000 takes 0.5 away. The score never leaves the 1 to 5 range.

Expandability

a unit that takes extra batteries scores 4, 4.5 when it grows to at least twice its capacity and 5 at three times; a unit that does not, or whose maker does not say, scores 2. A stated UPS switchover time adds 0.5, because it is what lets the unit sit between the wall and a refrigerator or a CPAP.

Owner-reported reliability

fan noise in use, app and Bluetooth reliability, firmware problems, failures and dead units, how warranty claims were handled, build quality and display readability, from independent reviews and owner reports only.

Independent reviews, and what disagreement does

For every unit, up to four independent public sources are collected: published tests and reviews by outdoor, technology, solar and consumer publications; expert reviews hosted by retailers; and long-form owner reports on forums or in video transcripts. The manufacturer's own pages and store customer-review lists are never sources, and Amazon review pages are never fetched. Each source is recorded with its URL, publisher, date where shown, and whether the writer had the unit in hand.

From each source, each of the seven criteria the source actually speaks to gets a score from 1 to 5 with a one-line reason in this site's own words. A criterion the source says nothing about gets no score from it. No line longer than fifteen words is kept from any source, and no source's wording appears on a reader page.

Where sources disagree, both scores stand. The gap between the highest and lowest score on a criterion is shown on the unit page as 'disagree' when it exceeds two points, and it lowers the unit's confidence rating; it never picks a winner and never tips a verdict. Confidence is high with three or more sources including one with the unit in hand and no wide disagreement, medium with two sources or with wide disagreement, low with one, and none with zero.

The verdict on a card or a unit page is written from the scores and the published figures, never copied or paraphrased from any one source's ranking, and never attributed to a named publication in the body of a page. A reader can see why a unit was chosen without visiting anyone else's site. The sources are citations in the collapsed Sources list at the bottom of the unit page, not the argument.

How a pick earns its label

A unit with fewer than two independent sources carries no award label anywhere on the site. Its scores still show, and it still sits in every table by them; it simply cannot carry an award until independent evidence exists for it. Picks are therefore drawn from the units that clear that floor, and where no unit in the running clears it yet the first-ranked unit still stands as the card, with the situation as a plain heading rather than an award. Every pick card prints how many independent reviews the score rests on, and says so plainly when there are none, so the evidence behind a number is visible without opening this page.

A pick is the highest-scoring unit among the units that qualify for a situation, with the weights below; ties go to the lower price. The label on the card is the buyer's situation, in the words buyers use: 'Best for home backup', 'Best for CPAP', 'Best value'. It is an award, and a unit earns it only when two or more independent sources back its score. A unit with fewer than two sources can still be the pick, because the rubric ranks it first, but its card carries the situation as a plain heading rather than an award, states what is known, and shows the same overall score as the comparison table with a limited-evidence tooltip. The situation weights select the winner; the displayed number is always the overall score. Once the evidence arrives the award label follows.

A situation's pick is always its first-ranked unit, on every page that shows it, even when one unit wins two situations; those labels and reasons share one card within a decision block. The 'best value' and 'easiest to carry' cards that sit beside a main pick name a different unit, so the section reads as a shortlist, and their verdict sentence explains the practical tradeoff. The overall and value picks are drawn from the units one person carries (a stated weight of 40 lb or less) that clear a night of CPAP, because that is the unit most buyers are actually choosing. Job picks are drawn from every unit that holds enough for that job. Size-class picks are drawn from the class. Inside a situation, value and recharge rank each unit against the others in the running for that situation rather than against its size class.

Best overall

fit 25, value 20, recharge 15, portability 10, chemistry 10, expandability 10, reliability 10.

Best value

value 50, fit 20, chemistry 10, reliability 10, recharge 10.

Best for home backup

fit 25, value 20, expandability 20, reliability 15, recharge 10, chemistry 10.

Best for CPAP

portability 30, value 20, reliability 20, fit 15, chemistry 10, recharge 5.

Best for camping

portability 35, recharge 20, fit 15, value 15, reliability 10, chemistry 5.

Best for an RV

fit 20, value 20, recharge 20, expandability 15, reliability 15, chemistry 10.

Best for van life

recharge 30, fit 15, value 15, chemistry 15, reliability 15, expandability 10.

Best for phones and laptops

portability 40, value 30, fit 10, reliability 10, chemistry 10.

Best in a size class

the overall weights, within the class. Best value in a class: the value weights. Easiest to carry: portability 60, value 15, reliability 15, chemistry 10, with portability ranked by weight within the units in the running, because that card asks which of these is easiest, not whether any is easy.

Best for solar, per class

recharge 45, fit 15, value 15, chemistry 10, reliability 10, expandability 5, among units with a stated solar input.

How tables are ordered

The best-portable-power-stations table starts with overall score highest first and exposes a sortable overall-score column. Other full tables list units smallest to largest by capacity, because capacity is the first thing to decide, and says so in one line above the table. Price is a filter and a column, never the default order. The fit tables on the two jobs where the unit is carried, camping and CPAP, list lightest first. The solar page lists by hours of full sun to refill. Every table can be re-ordered by price, price per watt-hour, weight, output or solar input with the control above it.

Which merchant a button points to, whether that merchant pays a commission, and whether this site has joined any program are not inputs to any order, any score or any pick. The affiliate disclosure page states the link state and the same rule.

How figures are made comparable

Capacity is always the rated battery capacity in watt-hours. Where a brand markets a kilowatt-hour figure, it is converted to Wh (1 kWh = 1,000 Wh). Continuous AC output is the inverter's normal rating; a boost mode that runs resistive loads above the rating (EcoFlow X-Boost, Bluetti Power Lifting and similar) is noted on the unit page but never used as the continuous figure. Surge is the peak the manufacturer states for the inverter.

Chemistry is recorded as LiFePO4 when the manufacturer says LiFePO4, LFP or lithium iron phosphate, and as NMC when it says NMC, NCM or ternary lithium. A unit whose page only says 'lithium-ion' is recorded that way and never guessed into a chemistry. Cycle life is recorded together with the capacity it is rated to, because '3,000 cycles to 80%' and '3,000 cycles to 70%' are different claims.

Weight and dimensions are shown in both metric and US units. Where the manufacturer publishes only one, the other is converted (1 lb = 0.45359 kg, 1 in = 25.4 mm) and the conversion is stated in the unit's notes. Price per watt-hour is the current official price divided by rated capacity, or the list price when no current price could be found; a unit with no price has no price per Wh and sorts to the end of a price-ordered table rather than being estimated.

Size classes are four fixed bands of rated capacity, the same on every page: compact under 600 Wh, mid-size 600 to 1,499 Wh, large 1,500 to 3,499 Wh, whole-home 3,500 Wh and up. Value, portability and recharge scores rank a unit against its own class so that a 300 Wh unit is never measured against a 4,000 Wh one.

The sizing calculator uses the arithmetic on the watt-hours page: device average watts times hours, divided by 0.85 for inverter losses, plus 20% headroom. AC efficiency and headroom are fixed planning assumptions shown next to the answer. Custom DC rows use a separately editable efficiency estimate. Custom quantities multiply energy and peak demand; duty percentage changes energy only. All copies in one custom row may start together, while separate rows start one group at a time. Hardwired, split-phase and unknown connections never receive a confirmed fit; DC rows also require a separate port and cable check. Custom entries stay in browser memory and are not included in analytics or URLs.

How fresh the data is

The spec bundle carries a generated_at timestamp and every row carries its own dates. The health endpoint at /api/v1/health.json reports the bundle's age against a 90-day window; after that the site labels the dataset stale rather than pretending a price checked months ago is today's price.

Specs rarely change after launch, so a spec date weeks old is usually still right. Prices change weekly. Treat the price column as a dated observation and the manufacturer's page as the current truth. Review scores carry the date each source was collected.

What this site does not do, and what it does instead

This site does not run the units. There are no runtime tests, no capacity measurements, no efficiency tests and no photographs of its own, and no page claims any. What it does instead is the rubric above: the manufacturers' published figures, scored the same way for every unit, combined with the consensus of independent tests and owner reports, with the fit arithmetic in the open on every page. Where a page shows a product photo it is the retailer's own image for that listing and links to it.

A figure the manufacturer does not publish is shown as a dash. It is never filled from a retailer listing, a review or memory. Some brands publish far less than others, and the table shows that honestly rather than levelling it.

Discontinued units are not rows unless a current unit page needs to explain a rename or a replacement, in which case the older name appears in that page's text. Home battery systems that are not sold as portable units are outside scope.

Corrections and changes

When a row is re-checked, the dates move and the old values are replaced; the CSV and JSON exports always reflect the current bundle. When new independent sources are collected, the scores, the confidence and the labels follow at the next build.

Corrections are welcome from readers and manufacturers alike and are held to the same standard: a public page on the manufacturer's domain that another reader can open. A private assertion can raise a question but cannot change a value. A manufacturer cannot buy a score, a label or a position.

This page was last materially reviewed on 2026-09-07, when the rubric replaced the earlier spec-only shortlists.

Rated capacity, estimated energy and measured results

Rated watt-hours describe the battery. The buying cards multiply that rating by 0.85 to estimate energy delivered through AC outlets. This is the calculator's shared planning assumption, not a measured efficiency for the model. It assumes a full battery, no reserve and no recharging; low loads, idle consumption, temperature and battery age can change the result.

Independent review citations and scores do not establish measured runtime or fan noise. No reusable independent measurement records are held for these cards. The unavailable label means exactly that, not that nobody has tested the product. Manufacturer ratings remain separate from our calculations.

The source records retain the official URLs and original read dates behind each limitation. Return to the power-station picks to compare the tradeoffs, or size your own appliances.