Google Keyword Planner reports only 42 distinct search volumes

The problem: your report shows a percentage change that does not exist

You open a monthly SEO report. Volume for a target keyword moved from 1600 to 1300. Someone in the room asks why traffic did not follow. The honest answer is that nothing happened. Google Keyword Planner never reported 1300 or 1600 as measurements. It reported the nearest bucket on a fixed ladder, and the ladder is coarse enough that a one step move is indistinguishable from noise.

The symptom shows up in three places. First, in month over month comparisons, where small percentage changes get presented as trends. Second, in keyword prioritisation, where two keywords a single step apart get ranked as if one had meaningfully more demand. Third, in forecasting, where an average is computed across values that were never continuous to begin with.

If you buy or sell SEO, this matters commercially. A report that says volume is up 23 percent is describing a rounding decision by Google, not a change in the market. The person reading it has no way to tell the difference unless someone explains the scale.

How to diagnose it

You do not need a paid tool to see this. You need to export enough data and look at the distinct values.

Start with the Keyword Planner export itself. Pull monthly volumes, not just the annual average, for a set of keywords across several countries. The annual average hides the ladder because it produces numbers that look continuous.

Once you have the export, count the distinct values. In a spreadsheet:

=SUMPRODUCT((A2:A1393<>"")/COUNTIF(A2:A1393,A2:A1393&""))

Or in a shell, from a CSV where volume is the second column:

cut -d, -f2 volumes.csv | tail -n +2 | sort -n | uniq -c | sort -rn

The second command prints every distinct volume with its frequency. If the data is bucketed, you will see a small number of values each repeating many times, rather than a long tail of unique numbers.

Then check the ratio between consecutive values. Sort the distinct values ascending and divide each by the previous one:

sort -n distinct.txt | awk 'NR>1{printf "%.3f\n", $1/prev} {prev=$1}'

If the output clusters around a single constant, you are looking at a logarithmic scale, not a linear one. That single number is the step size of the buckets.

Finally, run a control test. Pick keywords completely unrelated to your niche, ideally high volume commercial terms, and pull three consecutive months. If the same figure repeats exactly, month after month, you are observing the bucket boundary, not a stable market.

What the data showed

Prospectia measured this in September 2026 from its own exports. The dataset covered 6 keywords from the technical SEO cluster, across 4 countries (US, CA, AU, UK), for 12 months each. That is 1392 monthly volume measurements.

Those 1392 measurements contained only 42 distinct values. The full set, ascending: 10, 20, 30, 40, 50, 70, 90, 110, 140, 170, 210, 260, 320, 390, 480, 590, 720, 880, 1000, 1300, 1600, 1900, 2400, 2900, 3600, 4400, 5400, 6600, 8100, 9900, 12100, 14800, 18100, 22200, 27100, 33100, 40500, 49500, 60500, 90500, 110000, 135000.

The ratio between neighbouring values is constant at about 1.22. That is the whole finding. The scale is logarithmic with a fixed step, which means Google is not measuring demand and reporting it. Google is assigning each keyword to one of 42 bins.

The practical consequence is that a move of one step, for example 1600 to 1300, can be zero real change. The margin is about plus or minus 22 percent on every single data point. Only a difference of two steps or more, roughly 50 percent, means anything at all.

A concrete case from the same dataset makes this visible. The keyword "technical seo specialist" in the US showed an average of 1300 searches per month. That average was inflated by a single spike of 12100 in December 2025. The last three months of the series were 90, 90 and 90. An average of 1300 describes a keyword that is currently sitting in the 90 bucket, with one anomalous month pulling the mean upward by more than an order of magnitude.

The control test removed any doubt. Unrelated keywords behaved the same way. "plumber near me" reported 823000 for three consecutive months. "car insurance quote" reported 673000 for three consecutive months. The exact same figure repeating across consecutive months is bucketing, not real demand. Real demand does not land on the identical integer three times in a row.

The 42 values and what they mean for a report

The fix

Stop treating Keyword Planner volume as a continuous measurement. Treat it as an ordinal bucket index. That single change fixes most of the reporting problems.

Convert every volume to its bucket index before doing any comparison. Build the ladder once, then map values onto it:

LADDER = [10, 20, 30, 40, 50, 70, 90, 110, 140, 170, 210, 260, 320,
          390, 480, 590, 720, 880, 1000, 1300, 1600, 1900, 2400, 2900,
          3600, 4400, 5400, 6600, 8100, 9900, 12100, 14800, 18100,
          22200, 27100, 33100, 40500, 49500, 60500, 90500, 110000, 135000]

def bucket_index(volume):
    return min(range(len(LADDER)), key=lambda i: abs(LADDER[i] - volume))

Then apply two rules in every report and every prioritisation sheet.

• Never report a change of one step as a change. If the index moved by 1 or 0, report it as flat.
• Only flag a movement when the index changed by 2 or more. That corresponds to roughly 50 percent, which is the smallest difference the data can actually support.

For keyword prioritisation, group keywords by bucket index rather than by raw volume. Two keywords at 1300 and 1600 belong in the same tier. Splitting them into different priority levels is an artefact of the tool.

For forecasting, do not average monthly values across a series that contains a spike. The "technical seo specialist" case shows why: an average of 1300 was produced by a series whose recent months were all 90. Use the median of the last three months, mapped to its bucket, and label it as a bucket rather than a point estimate.

If you are building a keyword model for a client, this is the point where the volume layer stops being the weak link. The rest of the work, mapping intent, competition and site fit, is what actually needs judgement. If you need that layer built properly, the technical SEO service covers the keyword and architecture work that sits on top of this data.

How to verify the fix worked

Before the fix, measure two things on your existing reports. First, the number of distinct volume values across your keyword set. Second, the number of reported month over month changes that were a single step.

After the fix, the same two measurements should change in a specific way. The distinct value count should not change at all, because you have not changed the source data. What changes is the count of reported changes. Single step movements should drop to zero in your output, because they are now classified as flat.

A simple check: take your last three monthly reports and re-run them through the bucket index logic. Count how many keyword movements were flagged as changes. Then count how many of those were two steps or more. The gap between those two numbers is the volume of false signal you were previously sending to clients or to your own team.

Also verify the spike handling. For any keyword where the average and the median diverge by more than one bucket, flag it for manual review. In the "technical seo specialist" case, the average sat at 1300 while the recent months sat at 90. That is a divergence of several buckets and should never have been reported as a stable average.

Run this check quarterly. The ladder itself may change over time, and if Google adds or removes bucket boundaries, your index mapping needs to be rebuilt. The 42 values observed in this dataset are the current state, not a permanent constant.

Frequently asked questions