What SEMrush gets wrong about B2B keyword data

An agency pitched a client on a new Google Ads strategy. They brought a full deck with projected CPCs, search volumes, and estimated traffic for the target keywords.

Every number was wrong. Not slightly off. Wrong by multiples.

They had used SEMrush as their source of truth for a hyper-niche industrial vertical. SEMrush showed most of their core terms at "0 search volume." They dismissed those terms and built the strategy around the keywords with measurable volume.

Those measurable-volume keywords were consumer-grade. The zero-volume keywords were the ones that actually drove pipeline.

I had to clean it up.

Why SEMrush fails in niche industrial

SEMrush (and Ahrefs, and Moz, and every other third-party tool) builds their keyword data by scraping consumer clickstream data from browser extensions and toolbar plugins.

That clickstream data is dominated by consumer behavior. Millions of people searching Amazon, Google Shopping, lifestyle sites. Industrial procurement engineers are a rounding error in that dataset.

When a term gets searched by 15-20 people a month globally, it falls below the threshold these tools can reliably detect. They report "0 volume" or "< 10."

But in industrial B2B, 15 searches a month for "NFPA 70E arc flash warning label specifications" might represent $500K in annual pipeline. Those 15 searchers aren't casual browsers. They're engineers with procurement authority.

What the tools are actually useful for

SEMrush isn't useless. It's just useful for different things.

Good for: understanding competitor domain authority, tracking organic rankings, finding broad category keywords to use as seeds for research, auditing content gaps.

Bad for: forecasting CPC, estimating search volume in niche verticals, building a B2B keyword strategy from scratch.

If you use SEMrush data to tell a VP what to expect from a Google Ads campaign in industrial manufacturing, you're going to miss by 80%.

The only valid data source for niche B2B

You have to run the campaigns.

Set up a tight exact-match test budget. $1,500-2,500 over four to six weeks. Target your core terms and as many technical variants as you can identify. Let it run. Look at actual CPCs, actual impression volume, actual query data.

That's your baseline. Not a third-party estimate.

Yes, it costs money to gather real data. But it costs more to build a strategy on wrong data and defend why the projections didn't land.

Every client pitch I do for a new industrial account includes this caveat: "The auction data we need doesn't exist in any tool. Here's how we'll go get it."

That's not a weakness. That's honesty. And it's what keeps you from presenting a deck that blows up three months into the engagement.