Editorial Standards, Sourcing and Privacy at Espresso Tested

How does an espresso machine get reviewed here?

Espresso Tested is a desk-research operation, and you should know that before you weigh anything we say. We do not run a coffee lab. There is no room full of machines, no staff barista pulling calibrated shots, and no instrumented rig measuring brew pressure or extraction yield. What we do instead, for every machine in a guide:

• We read the documentation. Spec sheets, manuals, parts diagrams and warranty terms. Manuals are the most under-used source in this category — they tell you basket sizes, boiler type, descaling intervals and which parts the maker expects you to replace yourself.
• We pull the retailer listing data. Current Amazon price, listing specifications, star rating and the distribution behind it, plus availability and variant confusion — plenty of listings bundle a grinder or a different tank under one product page.
• We read owner reviews in volume and look for repetition. One angry review is noise. Forty owners independently describing the same leaking group gasket at eight months is a finding. We weight verified purchases and pay most attention to what people say after the first month, not on unboxing day.
• We compare like against like. A sub-$200 thermoblock machine is not judged against a dual boiler, and a capsule machine is not judged as though it were a semi-automatic.
• We add hands-on notes only where we genuinely have the machine. Where a team member owns or has lived with a unit we say so and mark it as first-hand. Where we have not, we do not imply that we have.

That method has real limits, and naming them is more useful to you than hiding them. We cannot tell you the true temperature stability of a thermoblock across a thirty-second shot. We can tell you what the maker claims, whether there is a PID, and whether owners repeatedly report sour, under-extracted shots that point at recovery problems. Read our conclusions as well-sourced judgement, not as measurement.

What goes into a score?

Scored guides use a ten-point scale built from four weighted dimensions. The weights are fixed across the site, so a 7.8 in one guide means roughly what a 7.8 means in another, and they tilt towards the two things that decide whether a machine survives its first year in your kitchen: what it puts in the cup, and whether it can be kept alive.

Shot quality — 35%

The largest share, because everything else is a support act. This covers the machine's capacity to hold stable pressure and temperature through an extraction: pump type and rated pressure, whether that pressure is genuinely regulated or merely advertised, boiler or thermoblock design, temperature control such as a PID, pre-infusion, and basket geometry — 51mm, 54mm and 58mm groups behave differently and take different accessories. Pressurised baskets earn credit for forgiveness and a deduction for ceiling: they flatter a coarse grind and cap what a competent user can eventually achieve. Where a machine has an integrated grinder, its burr type and step count count here too. This score is built from documented specification plus the pattern of owner reports about extraction consistency; it is not a measured extraction figure.

Build and serviceability — 25%

The second-largest share, and the one most guides ignore. Materials matter — a metal group and a brass or stainless boiler age differently from an all-plastic assembly — but serviceability matters more. Can you replace a group gasket and shower screen yourself? Are those parts sold at all, and at what price? Is the water path reachable for descaling, or sealed? What does the warranty cover, and does the maker run service in your country? We then read owner reports for documented failure patterns: pumps that die at a predictable age, over-pressure valves that weep, control boards that quit. A machine that breaks in a way you can fix scores above one that fails in a way you cannot.

Steam and milk texturing — 20%

Most home espresso ends up under milk. We look at steam pressure and dryness, wand type and articulation, and above all the machine's architecture: a single boiler makes you wait between brewing and steaming, a heat exchanger or dual boiler does not, and that wait is the most common frustration in owner reports on entry-level machines. Time to steam readiness, whether the wand reaches into a jug at a usable angle, and whether you get an open wand or a panarello-style assisted frother all feed the number. Assisted frothers are scored for what they are: a reliable route to warm textured milk, with a low ceiling for latte art.

Daily usability — 20%

What the machine is like on a weekday morning. Footprint and height under a wall cabinet, tank capacity and whether you can refill it without pulling the machine forward, drip tray size before it floods, control layout, warm-up time, noise, the cleaning routine, and how tolerant the machine is of a grind that is slightly off. Usability failures are why machines end up in the cupboard, and owner reviews are unusually reliable here: people describe daily annoyance vividly and accurately.

What does a score not mean?

It is a comparative summary within a category and a price band. It is not a prediction that your unit will last, not a substitute for reading the section on who a machine suits, and not a claim of laboratory measurement. Two machines a tenth of a point apart are, for practical purposes, equivalent, and we would rather you read the reasoning than the digit. Some guides carry no scores and use ranked picks with written justification instead; where the honest answer is "it depends on your milk habit and your counter depth", a number would only launder that uncertainty into false precision.

Can a brand buy a place in a ranking?

No, and there is no rate card to ask about. No manufacturer, distributor or agency has paid for inclusion, position, a score, or the removal of criticism, and we do not accept sponsored placements inside buying guides. Nobody outside the editorial team sees a guide before it publishes, and no brand gets a right of reply beforehand or a veto afterwards. A manufacturer who writes to us about a published guide is treated exactly like a reader: we check the claim, and we change the page only if the claim is right. We do not solicit review samples, and where a team member owns a machine it was bought normally.

What happens when we get something wrong?

We will get things wrong. Specifications get transcribed incorrectly, makers revise a model without renaming it, and a listing occasionally describes a variant that is not the one we assessed. The commitment we can make is about what happens next.

1. Tell us. Email reviews@espressotested.com with the page URL, the sentence or figure you think is wrong, and a source if you have one. A link to the maker's manual settles most disputes in one message.
2. We acknowledge it — we aim to reply within three working days, even if only to say we are still checking.
3. We verify against a primary source. Manufacturer documentation first, retailer listing second. Where the two disagree, we say so on the page rather than quietly picking one.
4. We fix it, and we show the fix. A material error is corrected and noted at the foot of the article with the date and a line describing what changed. If the correction moves a machine up or down a ranking, the note says so. Typos and dead links are repaired without a note.
5. We tell you what we decided. If we cannot verify your report, or we check it and still disagree, you get a reply explaining why rather than silence.

We do not remove a correction note later to tidy up the page; the record of having been wrong stays. Questions that are not corrections are better sent through the contact page, and many are already answered in the FAQs.

How does this site make money?

Through the Amazon Associates programme. Espresso Tested participates as an Amazon Associate, and product links here are affiliate links: if you follow one and buy something, Amazon pays us a small percentage of the transaction. Your price is exactly what it would have been had you typed the product name into Amazon yourself — the commission comes out of the retailer's margin, and following our links never adds a fee or a condition to your purchase.

The boundary that matters is this one: commission rates play no part in which machines we recommend, what order they appear in, what score they receive, or how the write-up is worded. We do not look up what a product pays before ranking it. Where the best machine for a reader is unavailable through the programme, out of stock, or sold only direct by the maker, we still name it and say so. If you would rather not send us the commission, search for the model yourself; the recommendation is unchanged either way.

Where does AI fit into how these pages are made?

Directly, and we would rather describe it precisely than hide behind a vague line about "tools". Two distinct things happen here.

Drafting is AI-assisted. Language models are used to produce first drafts, restructure sections and tighten prose. No page reaches this site in the state a model produced it. Every published page is read, checked and revised by a human editor before it goes live, and that person is responsible for the claims on it — the specifications, the reasoning behind a ranking, and anything stated as fact. When a draft asserts something the editor cannot confirm against documentation or owner reports, the assertion is cut rather than softened.

Product data is collected automatically. Prices, star ratings, rating counts, listing specifications and summaries of owner reviews are gathered by automated processes from retailer listings. That material is a snapshot taken at a particular moment, not a live feed: it is not re-fetched when you open the page, and nothing here updates itself in your browser. Automated summarisation can also flatten nuance or over-weight a vivid minority complaint, which is one reason review sentiment informs our scores but never decides them alone.

Why is the price on this page not the price at checkout?

Because the figure here was captured at a point in time, and Amazon pricing moves with promotions, seller rotation, storefront and stock level, sometimes several times in a day. Ratings drift for the same reason, and listing specifications change when a maker refreshes a model or a third-party seller edits a page. So treat every price, rating and specification here as an indication of where a machine sits, not a live quote. The standing instruction on this site is simple and we mean it: before you buy anything, open the retailer listing and confirm the current price, the exact model and variant, what is in the box, the seller, and the return window. If the listing and this page disagree, the listing is right — and we would like to hear about it. Nothing here is a binding offer.

What does Espresso Tested collect about you?

Less than most sites, mainly because there is very little here to collect it with. Reading a guide involves:

• Server logs. Ordinary web-server entries for each request — IP address, timestamp, the URL requested, the referring page and your browser's user-agent string. They keep the site running and help us spot abuse, scraping and outages.
• Basic analytics. Aggregate measurement of which pages get read, roughly how visitors arrive, and which country-level region they come from. We use it to decide what to write next, not to profile you, and we do not attempt to identify readers.
• Cookies and similar storage. A small number set by this site for functional reasons and for the analytics above. Separately, when you click an affiliate link, Amazon sets its own cookies and tracking parameters so a resulting purchase can be credited to us. Those cookies are Amazon's, live under Amazon's domains, and are governed by Amazon's privacy notice rather than this page — we can neither read nor delete them for you. Blocking non-essential cookies in your browser costs you nothing here.
• Whatever you put in an email. If you write to us, we hold the message and your address for as long as it takes to deal with it and to record any correction that resulted.

What does not happen here is worth stating too. There are no user accounts, no logins and no passwords. We take no payments and never see a card number — every transaction happens on the retailer's site, under the retailer's terms. There is no comment system and no user profile. We do not sell, rent or trade personal information, we do not run behavioural advertising, and we do not knowingly collect anything from children.

How long is it kept, and how do you get it deleted?

Server logs rotate and are discarded on a short cycle measured in weeks rather than years. Analytics are retained in aggregate form, which by its nature is not tied back to an individual reader. Email correspondence is kept while it is useful and then deleted. We do not maintain a long-term database of readers, because we never build one.

To ask what we hold about you, to have it corrected, or to have it deleted, write to reviews@espressotested.com and describe your request. That single address handles every kind of data request here; there is no separate privacy desk and no form to fill in. Give us enough detail to find the records — usually the address you wrote from and an approximate date. We may ask a follow-up question to confirm the request comes from you, and we aim to respond within thirty days. For a reader who has never emailed us, the honest answer will normally be that we hold nothing identifiable about you at all, and we will say so rather than manufacture a file.

What if you are in California?

California residents have statutory rights over personal information, and we will honour them for any reader who asks, whether or not the thresholds that trigger the law apply to a site this size. You may ask us to confirm what categories of information we hold and where they came from, to correct anything inaccurate, and to delete what we hold. We will not treat you differently for asking.

On the question that usually matters most: we do not sell personal information, and we do not share it for cross-context behavioural advertising. There is consequently nothing to opt out of, and you will not find a "do not sell my information" toggle here because there is no sale to switch off. Affiliate attribution through Amazon is the one third-party flow on this site, and it exists to credit a purchase, not to profile your reading.

What if you are in the EU or the UK?

Readers covered by the GDPR or the UK GDPR have the rights you would expect: access to what we hold, correction of inaccuracies, erasure, restriction of or objection to certain processing, and portability of data you gave us. Use the same address to exercise any of them.

Our lawful bases are narrow because our processing is narrow. Server logs and security monitoring rest on legitimate interests — keeping a public website available and defending it from abuse, balanced against the minimal and short-lived nature of what is logged. Non-essential cookies, including analytics, rest on your consent, which you can withdraw at any time through your browser without losing access to anything here. Correspondence is processed to answer the message you sent.

This site is operated from the United States, so if you read it from Europe or the United Kingdom your data reaches servers there. We have not appointed a data protection officer and do not claim to have one; requests go to the editorial address and are handled by the people who run the site. If you are unhappy with how we handle a request, you may complain to your national supervisory authority, or to the Information Commissioner's Office in the UK.

Is the site accessible, and what if it is not?

We build towards WCAG 2.1 Level AA and treat it as a working standard rather than a badge. We have not commissioned a formal third-party audit, so we will not claim conformance. Since you would rather know the weak spots than discover them, here is where we currently fall short:

• Wide comparison tables scroll horizontally on narrow screens, which is awkward with a screen reader or by keyboard alone.
• Alternative text on some older product images is thinner than it should be, and is rewritten as guides are revised.
• Contrast on a few secondary labels and muted captions sits close to the threshold and has not all been re-checked.

If something here blocks you, email reviews@espressotested.com with the page address, what you were trying to do, and the browser and assistive technology you were using. We treat access barriers with the same priority as factual corrections, and if a fix will take time we will tell you how to get the information you needed in the meantime.

Which law governs this page?

Espresso Tested is operated from the United States, and this page is governed by the laws of the State of Oregon, without regard to conflict-of-laws principles. Any dispute arising out of it belongs in the state or federal courts sitting in Multnomah County, Portland, Oregon. Nothing here limits rights you hold under the mandatory consumer or data-protection law of your own country. The fuller legal terms live in the Terms & Conditions.

Who do you write to?

One address covers all of it: reviews@espressotested.com. Corrections, privacy requests, accessibility barriers, questions about a score and messages from manufacturers all arrive in the same inbox. The contact page feeds the same place, and more about who we are is on the about page.

Revised 27 July 2026. This page supersedes any earlier statement of our editorial or privacy practices, and we update it when the practice changes rather than on a schedule.