How we get the numbers right

A crawler reads each carrier's own public pricing pages on a schedule. It identifies itself with a contact address, honors each site's robots.txt, and waits out the crawl delay a site declares before asking for the next page.

Some carriers disallow their pricing paths in robots.txt. Where that is the case, we do not crawl — a person enters that carrier's plans by hand, reading the same public pages the crawler would have read. The site does not fetch where it has been asked not to, hand-entered or not.

Every extracted record gets a confidence score built from three things: where the page came from and whether that source can support a price claim at all, whether the carrier's terms page was reachable, and how much of the plan or deal was actually read off the page rather than guessed.

That score sorts into three tiers. High means the source is an official carrier page, the terms were captured, and the numbers came through clean. Medium means the source is official but something is missing — the terms were not reachable, or part of the record was not fully extracted. Low means the source is not an official carrier page at all — a third-party listing, an ad, or a summary of someone else's summary.

A record cannot reach the high tier without its terms captured, whatever else it gets right — a price with no confirmed terms behind it is not a high-confidence price. We do not publish the exact thresholds a record needs to cross for each tier: an operator can tune them, and printing today's number would both go stale the moment it changed and hand away the shape of the automation gate.

Where the table can be sorted by recommended rank, the order it produces comes from one rule set that lives in this site's own code. It is versioned, and it changes only through code review and a version bump — never through a settings form.

Rule set version: 2.0.0.

Price — 65%

Reads the plan's per-line price at the line count being viewed.

Data allowance — 20%

Reads the plan's data tier — light, standard, premium, or unlimited.

Hotspot allowance — 5%

Reads the plan's mobile hotspot data allowance.

Premium data — 4%

Reads the plan's full-speed premium data allowance.

Perks — 3%

Reads the number of perks bundled with the plan.

Autopay discount — 3%

Reads whether the plan's listed price already includes an autopay discount.

Price is measured per line, at the line count being viewed, against a fixed ceiling of $100 a line. A price at or above the ceiling earns none of the price weight. Each dollar below it adds the same amount. The ceiling does not move, so a plan scores the same whatever else is on screen.

Hotspot, premium data, perks and the autopay check together carry 15% of the score. They can separate plans at a similar price. They cannot lift a plan above one with the same data allowance that costs $24 or more a line less.

Carrier identity is not one of those inputs. The scoring function has no parameter for it, and the weights above have no entry for it — there is no carrier a plan could belong to that changes its score. No carrier can pay for position, because there is no code path through which a payment could reach the rank at all.

A worked example: Unlimited Plus (illustrative example) (Not a live listing — for demonstration only), priced at $40 a line at 2 lines.

Price

0.39

Data allowance

0.20

Hotspot allowance

0.05

Premium data

0.02

Perks

0.00

Autopay discount

0.03

Total

0.69

That total is a relative figure, not a percentage and not a grade out of 100 — a rule set that changes through review cannot support more precision than that.

The rank position shown on the comparison table only appears when the table is sorted by recommended rank, so a table sorted by price never asserts an ordering you didn't ask for.

The guided quiz uses this same rule set, but on a set of answers rather than an open browse. When an answer set matches nothing, we don't hand back an empty page — we widen the search one step at a time, in this fixed order, and say which answers were widened:

1. Budget

When widened, stops applying your maximum monthly price.

2. Coverage

When widened, stops restricting to carriers running their own network.

3. Carrier

When widened, stops excluding the carriers you checked.

Each plan on the shortlist carries a one-line reason built from the same factor breakdown. The reason calls a plan priced well only when its price is below the middle price of the plans on that shortlist. It says the plan costs more than most when its price is above that middle price. If fewer than two shortlist plans have a price, it makes no price claim. It never names a factor that added nothing to the score as a strength.