Case Study · Paid Search · Q3 2021
The 26,715% number, explained.
It is the most eye-catching figure on this site and the easiest one to disbelieve. So here is the account it came from, what was actually wrong with it, what changed, and — just as importantly — how you should and should not read it.
The Setup
What we inherited
The account came over from a previous agency mid-2021. Spend was steady, the reporting looked busy, and by the only measures that mattered — qualified clicks and cost to acquire them — it was not working. The comparison below is that agency’s final full period against the first comparable period after the rebuild.
I am not naming the client. What matters for judging the work is the shape of the account, and that is fully visible in the numbers themselves — you do not need the logo to read them.
What was actually broken
A $464.35 cost per click is not a bidding strategy, it is a symptom. At that number the account was almost certainly serving broadly against expensive, loosely-matched terms and collecting impressions that nobody clicked. That is why the CTR delta is so extreme: the starting point was close to zero, and percentage change against a near-zero baseline produces a very large number.
This is the honest read on 26,715%. It is real, it is arithmetically correct, and it is mostly a measure of how broken the prior setup was. The number worth judging me on is the one underneath it — same objective, 27% less spend, and clicks that actually converted.
A cost per click in that range is almost always the same three things: match types far too broad, a negative keyword list that was never really built, and ad copy written against the product rather than against the searches the account was actually showing for. None of those are exotic problems. They are what happens when an account gets managed by report instead of by search terms.
What changed
The rebuild follows the same order every time. Pull the search terms report first and find out what the account is actually buying, which is almost never what the keyword list claims. Add negatives until the waste stops. Restructure into tight, single-intent ad groups so one ad can honestly answer one search. Rewrite the copy against those searches. Then, and only then, touch bids.
Most of the gain here landed in the first two steps. That is usually where it lives — not in clever bidding, but in refusing to keep paying for clicks that were never going to convert.
The Numbers
Same objective. Same period. Less money.
Click-through rate
+26,715%
Against a baseline that was effectively flat. Large because the starting point was close to zero — see the read above.
Cost per click
$1.27
Down from $464.35. The prior setup was paying premium prices for impressions that were never going to convert.
Spend
$4,406
Against $6,036 in the comparable prior period — a 27% reduction while qualified volume went up.
Efficiency
365×
The ratio between the two cost-per-click figures. Stated as arithmetic, not as a promise of what your account will do.
How To Read This
What this does and does not prove
It proves I can take over a badly-run account and fix it quickly. It does not prove your account will improve 26,715%, and I would be suspicious of anyone who implied otherwise. If your paid search is already competently run, the honest upside is incremental — measured in percentage points, not orders of magnitude.
The reason this case study is on the site is not the size of the number. It is that the same instinct applies everywhere I work: find what is structurally wrong before optimizing what is merely underperforming. Most AI strategy engagements fail for exactly the same reason this ad account did — the foundation was never right, and everyone kept tuning the surface.
Figures are from the client’s ad platform reporting, Q3 2021, compared against the prior agency’s final comparable period. Available for review under NDA.
Want the same look at your account?
The $2,500 audit is exactly this exercise, applied to your operation — what is structurally wrong, what to fix first, and what to leave alone.