Why the published numbers disagree by a factor of four

Put the question into a search engine and you will be given confident figures for this metro that sit thousands of dollars apart. Every one of them is presented as the answer. They cannot all be the answer, and understanding why is more useful than picking one.

Failure one: averaging unlike houses

An average is only meaningful when the things being averaged are alike. Roof replacements are not alike. The distribution runs from a small single storey ranch with one shallow simple plane to a large older two storey with a complicated steep roof, two chimneys, plank decking and difficult access, and the second is genuinely several times the job the first is.

Take the mean of that distribution and you get a number that sits in a valley between two humps and describes almost nobody. Publish it as "the average cost in St. Louis" and every reader anchors to it, then feels cheated when their actual roof is priced.

Failure two: measuring different things

One source is quoting material only. Another is material plus labour. A third includes tear off and disposal, which is a substantial fraction of the job. A fourth is quoting a national figure adjusted by a regional cost index rather than anything observed here. A fifth is reporting what people told a survey they paid, which is memory, sometimes years old, sometimes for a different scope entirely.

Failure three: the calculator is not an estimator

A calculator that asks for your address, roof age and email before producing a range is not primarily a measuring instrument. It is a lead form with an arithmetic animation attached. That is not automatically dishonest, and some of them are built on real aerial measurement, but it is worth being clear about what it can and cannot know.

• It can approximate roof area from imagery reasonably well.
• It infers pitch from shadow, parallax or a property record, and the multiplier it produces is applied to every square foot.
• It cannot see layer count at all.
• It cannot see decking condition at all.
• It cannot see flashing condition at a chimney or a sidewall.
• It cannot see access, where a dumpster fits, or how far bundles must be carried.

So it has guessed most of the inputs, and the output is a range wide enough to be technically defensible and narrow enough to feel like an answer. The number's real function is to be interesting enough that you give up a phone number.

Failure four: the anchor is doing the selling

Once a figure is in a reader's head it does work whether or not it is right. A low published average makes every honest local quote look like an overcharge, which is useful to whoever comes in slightly under it. A high published average makes an ordinary quote feel like a bargain. Neither effect has anything to do with your roof.

What to do with the quotes you have

Stop comparing totals first. Compare inputs first. Line up the square count, the pitch each company measured, the layer count each expects, the decking unit rate, the penetration list and the exact material. If those six agree, the totals are comparable and the difference is about the business. If they disagree, the totals are not comparable at all, and the cheapest one may simply be the one that assumed the most.

Then ask each company how the number changes if the decking is worse than expected, and get that answer in writing. It is the single most predictive question you can ask, because it separates companies that have thought about the unknown from companies that intend to discover it in front of you.

A note on what we do publish

Everything numeric on this site is either geometry you can verify with a calculator, or a public record with the source named on the page: NOAA for hail counts, the National Weather Service for storm ratings, the Missouri Department of Commerce and Insurance for statewide claim totals, the Missouri Revised Statutes for anything legal. If a figure appears here without a source next to it, that is a mistake and we would like to know about it.