Why We Do Not Trust a Local Search Result

A Failed Measurement Does Not Look Like An Error

Before NexTier builds anything, a market gets measured: how much demand exists, who is already established there, and how hard the established companies would be to sit alongside. That screening decides whether a site gets built at all, and it is the part of this business where being wrong is expensive.

The reason it is hard is not that the data is scarce. It is that when a local measurement goes wrong, it does not fail loudly. It returns a plausible number.

A search performed from the wrong place still returns companies from the city you typed, because the city name was in the query. A keyword tool that misreads its input still returns a tidy figure. A competitor list that quietly stops at the first page still looks like a competitor list. Nothing announces the problem, and everything downstream of it inherits the error while looking entirely ordinary.

What follows is four things that turned out to be true, each of which cost a wrong answer before it was caught. The measurement procedures themselves are not published — market screening is the part of this business that is actually proprietary — but the findings are, because they are the reason we do not take a local search result at face value.

1. The Same Search, From Two Places, Is Two Different Markets

Columbia, South Carolina. The same query, the same hour, measured two ways: once from a verified position inside the city, and once the way almost everyone runs it.

Wrong by a factor of three — but the count is the less interesting half. The unverified version had never shown four of those companies at all, including the one with the highest review total in the market. It was not a noisier measurement of the same thing. It was a measurement of something else that happened to look like an answer, and it said this market is open where the truth was marginal at best.

A market read this way is not slightly optimistic. It is a different market. Screening decisions made on one are being made about a place that does not exist.

2. Trade Words Belong To More Than One Trade

In Florida and along the coastal Carolinas, “paver” means brick hardscape — patios, walkways, pool decks. Read a paving search in those markets literally and it fills with companies that will never quote an asphalt driveway. In one Ocala read the apparent market leader, at 147 reviews, lays brick.

The same collision runs the other way in the demand data. “Asphalt” is a roofing word as well as a paving word, so a raw asphalt keyword set for any city carries shingle-roofing searches mixed in with paving searches. In one metro that was roughly 130 searches a month of people looking for a roofer, sitting inside a set being counted as paving demand.

Both errors point the same direction: they make a market look bigger and more contested than it is. Taken together they can inflate the apparent size of a paving market roughly threefold while putting an entirely different trade at the top of it.

3. Real Search Volume Is Not Always Your Customer

This is the subtle one, because nothing about the number is wrong. The searches are real. The people are real. They are simply not buying what you thought.

The largest keyword in one concrete market carried 150 searches a month, comfortably the biggest figure available there. Looking at who actually ranked for it settled the question: ready-mix suppliers, a gravel yard, a precast manufacturer. Those searchers want to buy material by the truckload. Not one of them is hiring somebody to pour a driveway.

For contrast, a paving hiring query at the same 150 searches a month returns paving contractors, eight out of eight, with advertisers bidding on it. Identical volume, completely different meaning. Volume is a measurement of attention, not of intent, and the two get treated as the same number constantly.

4. Difficulty Scores Cannot See The Part That Matters

Keyword difficulty scores rate the competitiveness of the ordinary list of blue links. For a local trade search, that is not where the customer goes.

One emergency-repair term scores a difficulty of 8 out of 100 — which reads as wide open — while the map results above it are held by a company with more than a thousand reviews, with the third-largest listing sitting near seven hundred. A sibling term in the same trade and city scores 0.

Both scores are accurate and both are beside the point. They measure the list nobody scrolls to, and that list is soft because the map results absorb the clicks. Nobody with a burst pipe reads past the three phone numbers.

This is more dangerous than a tool simply being silent about local results, because a difficulty score is an assertive number. A 0 sitting next to a high advertising cost looks like money somebody left on a table. In practice a difficulty score can rule a market out. It should never be allowed to rule one in.

What These Have In Common

Every one of them produces a market that looks larger, softer and more inviting than the real one. That is not a coincidence, and it is worth sitting with.

The failure log for this portfolio runs entirely in one direction: bad markets that looked good. Not one entry is a market wrongly rejected. Part of that is a selection effect — you only catch an error in something you keep looking at, and a market you discard you never look at again — but part of it is simpler. Nobody talks themselves into a market being harder than it is. The reasoning always finds its way toward the answer that means there is work to do.

Which is why the useful discipline is not cleverness. It is running a control: taking something whose answer you already know and putting it through the identical process. When a suspect market returns nothing and the control returns nothing too, the instrument is broken and the market is unmeasured — a completely different finding from an empty market, and one that looks identical without the control.

That habit has caught more errors here than any rule has, including a case where an entire market read as having no demand at all and the truth was that the query had been malformed. The measured outcome of that screening is published in full, blanks included.

How we measure — FAQ

Why does the same search give different answers? Because a search run from the wrong place still returns businesses from the city you typed, since the city name was in the query. Nothing on the page announces which set you are looking at. In one recorded case the same query in the same hour returned an entry point of 11 one way and 35 the other, and the weaker read had never shown four of the companies at all — including the one with the highest review count in that market.

Is the three-pack the whole competitive picture? No, and it is not close. The three businesses shown are a fraction of the list, and the companies that decide whether a market is worth entering are frequently not among them. In one market the two largest operators by review count sat outside the first page entirely, and an earlier assessment built from one screen of results had never seen the biggest one.

Can search volume be real and still be worthless? Yes, and it is the failure that fools people longest because nothing about the number is wrong. The biggest keyword in one concrete market returned a results page of ready-mix suppliers, a gravel yard and a precast manufacturer. Those searchers want material by the truckload rather than a contractor. The volume was genuine. The customer was somebody else's.

Why not trust a keyword difficulty score? Because it rates the ordinary list of links, and for a local trade search that is not where the customer goes. One emergency-repair term scores 8 out of 100 while the map results above it are held by a company with over a thousand reviews. The score is accurate and beside the point: the links are soft precisely because the map absorbs the clicks. Difficulty can rule a market out. It should never rule one in.

Why are the actual methods not published here? Because which markets are worth entering, and how they get screened, is the genuinely proprietary part of this business. The findings are published because they explain why we treat local data the way we do, and because they are checkable. The procedure that produces them is not, and an earlier draft of this page that published it in full was cut for that reason.

More from NexTier Digital

• How NexTier Digital Works
• What the Paving Sites Actually Rank For