Methodology
Every number on this site, and where it came from.
Providers quote per day, per three days, per five-day box, with and without snacks, with and without delivery. Putting those on one axis takes arithmetic, and arithmetic you cannot see is just a claim. So here is all of it, including the parts we get wrong.
Prices last read 2 Nov 2026•no corrections yet
01 — The arithmetic
How a quote becomes a per-day figure
Two divisions, done the same way for all 303 plans. The provider’s own wording stays on the row so anyone can check us in their head.
What Diet Diva PH publishes
₱450 for the 5-day Diet Diva Signature Meal Plan - No restrictions box, 3 meals + 1 snack a dayTheir price page, read 2 Nov 2026
What we do to it
450 ÷ 5 days = 450450 ÷ 3 meals = 150The snack is not a meal, so it is not in the divisor.A delivery fee applies in Makati, Taguig & BGC, Quezon City, Manila, Parañaque, Pasig & Ortigas, Mandaluyong, San Juan, Muntinlupa & Alabang and Las Piñas; it is printed on its own line and never added here.
What appears on the row
- A day is a day of food, not a calendar day. A five-day box eaten Monday to Friday is five days. Weekends a provider does not deliver are not counted against them.
- Snacks are never counted as meals. If they were, a provider could halve their per-meal figure by adding a biscuit, and the axis would reward packaging instead of food.
- A month is 4.345 weeks, not 4. Where a provider quotes monthly we divide by that, because 365.25 ÷ 7 ÷ 12 is 4.348 and rounding it to 4 overstates every monthly plan’s weekly price by about eight per cent.
- Delivery fees stay out of the price. Most providers deliver free inside a core radius and charge a flat weekly fee to the outer cities they name — so a fee belongs to a provider and a city, not to a plan. It never enters the per-day figure, because the ranked number has to mean the same thing for every visitor. What you get instead is the arithmetic: name your city and the row prints their weekly fee, our division of it, and the total beside the ranked figure. Before you name a city, the row flags that a fee applies in some of their cities and says how many.
- We use the list price, never the promo price. A code that expires in nine days is not what the plan costs. Codes are shown separately, with the date they die.
- “Starting at” is not a price. Where a provider publishes only a range, we list the cheapest plan that actually exists and name it. If we cannot establish a real plan, the provider is not listed.
- We round to the peso, up or down as the maths falls. No figure is nudged to make a plan look like it clears a round number, and nothing is rounded before it is divided.
Every plan on the site carries the quote it was derived from. If our division is wrong, it is wrong in public.
02 — Where numbers come from
Sources, in the order they beat each other
When two sources disagree, the higher one wins outright. There is no averaging and no judgement call.
A correction from the provider
Sent to us by the business itself. We verify it is really them once, then it publishes and outranks everything else we hold.
A structured, machine-readable feed
A price a machine can read is a price that cannot go stale between visits. No provider we list publishes one today, so this rank is a standing offer rather than a source in use.
Their public price page, read by hand
The fallback, and the reason a price can be up to a week stale. All seven of the providers we list are read this way. Every row prints the date it was last read.
And what we deliberately do not read:
- Prices scraped from social posts. A price in a caption or a carousel image cannot be parsed reliably, it is stale within a week, and harvesting it is against both platforms’ terms. Where a provider publishes prices nowhere else, a person reads them and the listing says so.
- Viber and Messenger group chats. Not public, whatever a screenshot suggests.
- Other people’s round-up articles. Copying a number we cannot trace is how a wrong price survives for two years.
- Anything a provider tells us about a competitor. We check it against the competitor’s own page, or we do not publish it.
03 — The ranking rule
The sort is the number you picked
The whole algorithm
Plans are ordered by whichever figure you chose in the sort control — per day, per meal, or calories. Nothing is weighted, boosted, blended, or placed by hand.
- Ties break alphabetically by provider. Arbitrary, therefore fair, and the same on every reload. Not by who pays, because nobody pays.
- Filters remove rows. They never reorder them. A goal filter shows fewer plans in the same order.
- Sponsored placement is one labelled box above the results. It is visually separated, never mixed in, and the rate is published. Nought sold to date.
Why it is this dull
A relevance score is where a comparison site hides its commercial interests. Once a sort has a dozen inputs, nobody outside can tell a helpful weighting from a paid one — including, eventually, the person who built it.
A single visible division is worse at flattering anyone and much better at being checked. If a plan is at the top, it is because it is the cheapest thing matching what you asked for.
04 — The on-time figure
The one fact no article can print
Every delivery window on the site is the provider’s own claim, copied from their order page — which is why the row reads “says”. Whether they hit it is a separate fact, and it only appears once diners have reported it back.
What makes a report count
- The ask only goes to someone who left this site for the provider’s own order page, and the report only counts when it carries a delivery photo or an order confirmation. So a figure cannot be padded by the provider, or against them by a rival.
- Nothing publishes below 30 verified reports. Seven of the seven providers listed show no figure at all today, and saying so is the point.
- The base is always printed beside the percentage. A percentage from thirty answers is a different claim from the same percentage from four hundred, and both should look different.
- Reports older than twelve months drop out, so a kitchen that has fixed its logistics is not judged on last year. The window is printed beside the figure it produced, and it is the same length for every provider — never shortened to whatever span makes a number publishable.
What a report asks
Six fixed questions, each answerable in one tap. No free text is ever published, and there is no field on the record it could be stored in.
The last box exists because diners want to tell someone, and because it is how we learn what to ask next. It is not a review field, and nothing typed into it goes on the site.
Right of reply. A provider who believes a figure is wrong gets the rule, the base, and the window the reports covered, and up to 300 unedited characters published under the figure on their own page. What we will not do is delete a verified report because it is unflattering.
05 — Taste, and no reviews
Taste is measured here, not reviewed
For a while this page said food quality was out of scope because opinion cannot be checked. That was wrong, and it is worth saying so rather than quietly editing it out. An aggregate of a fixed question is a statistic — it is what every survey ever run is made of. Taste is most of why anyone buys a meal plan, so it gets the same instrument as delivery: one tap, a printed base, a right of reply.
What we ask, and what it cannot tell you
“Was the food good?” in three options, plus a fixed vocabulary of things to flag — repetitive, bland, not enough sauce for the rice, soggy reheated, vegetables overcooked, portion smaller than the photos. The flags are asked of everyone rather than only the diners who disliked it, because you can enjoy a plan all week and still be sick of it by Friday.
- It measures acceptable, not excellent. A dull plan that is cheap and turns up at 4am will score well. Where the reorder figure and the food figure disagree, both are printed so the gap is visible instead of averaged away.
- Palates differ and the flags are not weighted. Whether a flag ends your interest in a plan is yours to decide, which is why we publish the breakdown rather than a single score.
- Star ratings are still refused. A 4.3 blends taste, courier punctuality, packaging and price into one number that cannot be acted on or checked. Six separate figures are less tidy and far more use.
The Philippine reason
Under the Cybercrime Prevention Act, publishing an unverifiable claim about a named business here carries real exposure — for the diner who wrote it and for the site that put it up.
A delivery photo timestamped against a stated window is a fact. “This kitchen is filthy” is not, however true it may be. We are not equipped to defend the second, so we do not host it, and we say so plainly rather than moderating quietly.
06 — What this gets wrong
The known holes, before you find them
A methodology page that only lists strengths is marketing. These are the limits we would raise ourselves if we were reading this site sceptically.
- A hand-read price can be seven days stale. Every one of the seven providers we list is read by hand. Their rows carry the date of the last read; treat anything older than a few days as needing a check on the provider’s own page.
- Coverage is claimed, not tested. The cities on a row are the ones the provider says they deliver to. We have not ordered to the edge of a zone to confirm it, and a provider quietly dropping one is exactly the kind of change we would learn about late.
- Seven providers is not the market. It is the providers whose prices we can establish. Dozens of kitchens sell entirely through direct messages at prices that change per customer, and we would rather omit them than invent a figure.
- Calories are the provider’s own number. Nothing here is lab-tested, and nobody in this market is weighing macros for us.
- The on-time figure only sees diners who came through this site. That is what makes it verifiable and also what makes it a sample. It is not the provider’s true rate; it is the rate among people we can trace.
- Two people maintain this. If a week is missed, the read dates will say so rather than quietly rolling forward.
07 — Corrections
Every number we have had to fix
Published in full, including the ones nobody else noticed. A comparison site that has never corrected anything has not been checked — so an empty table here is a statement about our age, not our accuracy.
Nothing to publish yet
No correction has been raised against a published figure. The site has been live for a short time and the arithmetic has not yet been tested by the people it describes, which is the only test that counts.
When one arrives it appears here with the date it published, what was wrong, who found it, and how long the wrong number was live — including the ones we find ourselves, and the ones we decide still stand. A log of only the flattering half is not a log.
Found something? [email protected]. Corrections from providers publish the same day; ours publish the day we notice.
08 — Page history
When these rules changed
A methodology that can be edited silently is not one. Every change to the rules above belongs here, dated, with what it did to the numbers.
This section is not yet a record. The rule changes it describes are real and are documented in the build plan, but their version numbers and dates have never been checked against what actually happened, so publishing them as a history would be exactly the kind of unchecked claim the rest of this page refuses.
It is a launch gate rather than a to-do: the log fills from the same place the corrections log does, once there is an operator queue writing to it. Until then this band stays, and nothing here is presented as a dated fact.