How we research, and how comparisons are ordered
Every material market input on this site comes from a named source with the date we read it, and you can open that evidence next to the figure. Where we calculate something from those inputs, we say so and show the assumptions. This page sets out what the register covers, where figures come from, which source wins when they disagree, what happens when we cannot resolve them, how often everything is rechecked, and the rules that decide the order providers appear in.
The one that matters
Commercial payments never affect the order of a comparison. If we could not keep that promise for a table, that table would carry the reason instead of the promise.
That is stated flatly because it is a promise about our own conduct, which we control. Claims about the market are written differently: what a provider publishes, and what is current, are scoped to the sources we examined and the date we examined them. We would rather be precise about what we checked than vague in a way that cannot be checked at all.
What the register covers
The register is broader than the handful of providers most of our comparisons show. It holds every proposition we have identified and researched, sorted into segments that decide how each one may be used. A specialist or private-banking proposition is not ranked against a mainstream one, because they serve materially different customer scopes and are not part of the same default comparison.
| Segment | Count | How it is used |
|---|---|---|
| mainstream | 3 | Regulated owner-occupied HPP for ordinary UK applicants. The default comparison. |
| specialist | 3 | Regulated, but different customer scope (residency, wealth, geography). Shown separately, never ranked against mainstream. |
| private-bank | 1 | Relationship-banking product. Directory listing only, no rate comparison. |
| legacy | 1 | Existing-customer servicing; no verified new-business route. |
| alternative | 2 | Not a regulated HPP. Separate section with explicit warnings. |
| not-relevant | 3 | No verified UK retail owner-occupied proposition. Excluded from provider counts. |
We do not claim the register is exhaustive. It is every proposition we have found and been able to evidence, which is not the same as every proposition that exists. New entrants, broker-only ranges and products we have not yet located are the obvious gaps. If you know of one we are missing, that is a correction we want.
Inclusion is not for sale either. A proposition is in the register because it exists and we can evidence it, not because of any relationship with us. The ordering rules below govern position; this governs presence, and the answer is the same for both.
Availability is recorded separately from segment. Segment says which comparison a proposition belongs in; availability says whether we have established a current route into it. There are four availability classes, and the ordering rules below use them: open, where we have verified a current route for the relevant customer and product scope; waiting list, where a route exists but access is queued or constrained; not verified, where we could not establish a current new-business or retail route; and unconfirmed, where the position is unclear enough that we will not rank it as open. A proposition can be specialist and open at the same time; the two classifications answer different questions.
These counts come from the records that build the site rather than being typed into this page. If the register changes, they change with it.
- 13
- propositions tracked, including ones not open to ordinary UK buyers
- 58
- registered sources, across 15 organisations
- 44
- canonical facts, each with a source and a checked date
- 8
- facts where a provider's own sources disagree
- 22
- open questions we have not resolved
- 86
- dated entries in the change log
On those open questions. None of the 22 has yet been raised with the provider concerned. We record what we could not establish and why; we do not yet operate a provider-response process, and we will not describe one until we do.
Where figures come from
Provider primary documents first: rate cards, product guides, criteria guides, tariffs of charges, certificates. Then provider public pages, then the FCA Register and Companies House for regulatory and corporate facts, then official statistics such as Bank of England series where we benchmark against the conventional market. For every material figure we record the source, the date we read it, and enough of the wording to check us.
We do not use other comparison sites as sources for figures. A number copied from a competitor arrives without its evidence and carries their mistakes into our records. If a figure is not in a provider document or an official register, it does not go on the page as a figure.
A rule learned the hard way. A search result summarising a document is a pointer to a source, never a source. More than once a summary has attributed a provider's buy-to-let rule to its home purchase plan, or reported a page as silent when it was not. Everything published here is read from the document itself, including the ones that are scanned images and have to be rendered before they can be read at all.
Published figures, derived findings and worked examples
Not everything on this site is a number a provider printed. Some of our most useful work is calculated from several sourced records, and saying "every figure comes from a source" would be a tidy sentence that misdescribes it. There are three classes, and they are labelled differently because they carry different weight.
- Published fact
- A provider or official source states it directly. We record the source, the date we read it and enough of the wording to check us. A rate, a fee, a deposit tier, a criterion.
- Derived finding
- We calculate it from two or more sourced records. Every input keeps its own source and date, the comparison has to satisfy the matching rules before it is allowed to exist, and the result is ours rather than the provider's. A gap against a benchmark series, a count of products meeting a condition, the difference between two deposit tiers.
- Worked example
- Arithmetic on stated assumptions: a property price, a term, a finance amount. The assumptions are shown with the result. It is an illustration of how something behaves, not a quote, not an affordability decision and not a claim about the market.
Why a derived finding needs more than good inputs
Two correctly sourced numbers can still make a false comparison. A provider rate observed in August set against a benchmark observed in April is a gap between two different months, not a difference between two products. So a derived finding carries its inputs, both observation dates and a vintage classification, and one of the build checks below fails if any of those dependencies is missing. Comparing like with like is part of the derivation, not a courtesy applied afterwards.
What a figure can be
Most sites have two states for a fact: published, or absent. That is not enough to describe this market honestly, so figures here hold one of 5 states. The counts are live.
| State | Count | What it means |
|---|---|---|
| Confirmed as published | 32 | One source, read directly, saying the thing plainly on the date we read it. It means the provider published this, not that it is guaranteed to be the decision an underwriter reaches on an individual application. |
| Disputed | 8 | Two or more first-party sources publish incompatible values, conditions or rules for the same fact. We hold every published version and treat none as governing. |
| Unresolved, multiple observations | 1 | Several published statements that are not competing readings of one number but observations about different things. Each keeps its own scope. |
| Not found in reviewed sources | 2 | We did not locate it in the sources we reviewed, on the date stated. Finding one document proves a thing is published; not finding it does not prove it was never published anywhere, so this state says what we can support rather than more. It is still never treated as a permission. |
| Blocked | 1 | The figure exists but we could not read it, for example a rate that renders as 0% to an automated reader. We publish the obstacle, not a guess. |
Why "disputed" is a state and not a decision
When a provider publishes two different values, the tempting move is to take the newer one, or the one that reads like a summary, or the lower one because it is safer. All three are guesses dressed as editorial judgement. A disputed figure is shown with every value the provider publishes, and separate code decides what a reader is shown and what an eligibility check is allowed to rely on, so a display can never quietly become a decision.
Product scope, and why it has its own rule
Providers publish more than one product. The same company will run a home purchase plan, a buy-to-let range and sometimes a bridge product, each with its own criteria, and the buy-to-let criteria are often the most detailed and the easiest to find. That makes a specific, current, first-party number available for exactly the wrong product.
So every source carries the product family it speaks for, and a page about home purchase plans cannot be built on buy-to-let sources. The permitted exception is deliberate: where a tempting figure exists for another product, we show the reader the number and say why it is not carried across, rather than omitting it and leaving them to find it elsewhere without the caveat.
Scopes currently in use: residential_hpp, provider_general, mixed, unclassified, buy_to_let. A source whose scope has never been assessed cannot be cited by a scoped dataset; the build fails until someone classifies it.
How often figures are rechecked
Different figures go stale at different speeds, so they carry different review windows.
| Figure type | Review window |
|---|---|
| Rates and fees | 14 days |
| Availability status | 30 days |
| Eligibility criteria | 60 days |
| Market statistics | 180 days |
What is automated and what is not. Pages compare a figure's date against its window and mark it as due for review rather than presenting it as current. The recheck itself is manual research, not a scheduled job, so the window is the standard we hold ourselves to and the date on the page is what actually happened. Where those two diverge, the date is the honest one and it is the one you should read. A review window is also not a promise that a figure has not moved: providers can and do change products between our checks, which is why the date sits next to the figure rather than in the small print.
How comparisons are ordered
Comparisons are ordered by stated, objective criteria only. Providers we have verified as open come first, then waiting-list ones, then those we could not verify or confirm, using the availability classes defined above. Within that, ordering is by the figure the table is about: for a rates view, the lowest published initial rate; for a deposit view, the lowest published minimum deposit. Ties break alphabetically.
Order is not a recommendation. It is the display order for the one figure that table is comparing, after scope and eligibility rules have been applied. The lowest initial rate is not necessarily the lowest total cost, and the lowest deposit is not necessarily the best product for you.
Commercial payments never affect ordering. Where a provider or partner relationship exists it is disclosed where it applies, and the rules above still decide position. We explain how the site is funded on how we make money, including the rule that you never pay us anything.
This is enforced, not just promised. One of the build checks below fails the build if commercial configuration can reach the code that ranks, matches or prices anything. A promise that depends on nobody making a mistake is not a promise.
What the build enforces
14 checks run on every build and the site does not publish if any of them fails. Each exists because a specific mistake reached a page once and we did not want it to be possible twice. 9 of them protect the research itself.
| Check | What it prevents |
|---|---|
check-selection | Commercial configuration cannot reach ranking, matching or pricing. |
check-boundaries | The separations between decision, display, research and engine code hold. |
check-criteria | Eligibility matches conjunctively, so a provider is never shown as a match on a route it does not publish. |
check-pairs | Every rate comparison is one a single real customer could actually choose between. |
check-disputed | A disputed figure cannot be flattened, in either the optimistic or the conservative direction. |
check-ftv-duplication | A disputed figure stays disputed everywhere it is copied, including datasets that mirror it. |
check-product-scope | A dataset about one product cannot be built on another product's sources. |
check-benchmark | External comparisons are vintage-matched and carry every dependency they rest on. |
check-insights | Published findings are re-verified against current data on every build. |
The other 5 protect the site rather than the research: internal links, structured data, house voice, layout collisions, and a check that internal editorial notes never reach a published page. Necessary, less interesting.
Every one is tested by injection. A guard nobody has watched fail is a guard nobody should trust, so each is proved by deliberately introducing the error it exists to catch and confirming the build stops. Two known problems that no check can catch are recorded openly rather than left implicit: a build check can only see what has been registered, so it can tell you a source is the wrong product, and never that you failed to find the right one.
How software is used, and where it stops
This site is heavily automated and it would be strange not to say so. Software holds the records, calculates the derived findings, flags figures that have passed their review window, and runs every check above on each build. The comparison tables and the figures in the copy are generated from the records rather than typed, which is why a stale number on one page cannot survive a correction made on another.
Where it stops is evidence. A search result or a generated summary is a pointer to a source, never a source. Nothing enters the record because a tool reported it: the underlying provider document or official register is opened and read, including the ones published as scanned images that have to be rendered before they can be read at all. More than one figure on this site was nearly recorded wrongly because a summary attributed a provider's buy-to-let rule to its home purchase plan, and the only thing that caught it was opening the document.
What we do not do
- We do not issue religious rulings. Where scholars differ we report published positions with attribution, and record who certified what. The certification page sets out that boundary in full.
- We do not name a best provider. Suitability depends on circumstances we cannot see, and that judgement belongs to an FCA-authorised adviser.
- We do not give advice. We are not a lender, a broker or an adviser, and nothing here is a personal recommendation.
- We do not publish figures we could not verify. Where a check failed we say so, with the date we tried, and the figure stays out.
- We do not infer a provider's reasoning. Where a provider publishes an explanation for a criterion we can report it, with attribution. Where it does not, we report the rule and leave the rationale alone rather than inventing a plausible one.
Corrections
When we get something wrong we correct it visibly: the corrected figure on the page, and a dated entry in the change log stating what we said before and what the source actually shows. Corrections are never silently overwritten.
That includes corrections nobody would have noticed. Several findings on this site were withdrawn after review because the evidence turned out not to support them: a deposit route inferred from one document's silence, a criterion attributed to the wrong product, a provider recorded as publishing nothing when its criteria guide sat behind a different document hub. A site that never corrects itself in public is not more accurate than one that does. It is just quieter about it.
Reporting an error. If a figure here is wrong, the fastest fix is to tell us which page, which figure and what the provider's own document says. We do not yet publish a dedicated address for that, and we are not going to describe a process we are not running, so it goes live with the site rather than before it. The correction policy above already operates: every change we have made is in the log.
Common questions
- Where does Islamic Mortgage Guide get its figures?
- From provider primary documents, provider public pages, the FCA Register, Companies House and official statistics such as Bank of England series. Every material figure carries the source it came from and the date we read it. We do not use other comparison sites as sources for figures, because a figure copied from a competitor inherits their errors without their evidence.
- What happens when two sources disagree?
- We publish the disagreement. Where a provider's own sources give different values for the same criterion, the figure is marked disputed, every published value is shown, and none is treated as governing until we can establish which does. Silently choosing the tidier number would make the site easier to read and less true.
- Do providers pay to appear higher in your comparisons?
- No. Ordering is decided by stated, objective rules: open providers before restricted or unconfirmed ones, then by the figure the table is about, with ties broken alphabetically. Where a commercial relationship exists it is disclosed, and it does not change position. If we could not keep that promise for a table, the table would carry the reason instead.
- How do you decide what counts as a material figure?
- Anything you could act on or compare: a rate, a fee, a deposit minimum, an eligibility rule, a provider's regulatory status. Material figures carry source trails. Worked examples and illustrations are labelled as illustrations and do not, because they are arithmetic on assumptions rather than claims about the market.