“A basis point is a rounding error.”, every CFO, at least once.
We modeled the compounding cost of a 1 bps (0.01%) permanent increase in TACoS on a $20M/yr Amazon brand. Over 3 years, at an 8% discount rate, the NPV of that drift is ~$52,800. That is, coincidentally, the fully-loaded cost of a mid-career hire for a year.
This post is the worked version of that model, the assumptions, the sensitivity analysis, and the four operational places where bps drift hides on accounts we audit. The TL;DR: every operator on a scaling Amazon brand should be tracking TACoS to the basis point, weekly, with a clear theory for why it moved every time it does. Most aren’t.
Why the number is bigger than it feels
The reason a basis point isn’t a rounding error is that TACoS drift is almost never a one-time cost, it compounds as spend grows. When your ad spend is growing 15% YoY (as it does for most scaling accounts), a 1 bps miss today is a 1.15 bps miss next year on the same leak. Nobody notices. The program looks healthy. The P&L quietly loses.
The math behind the $52,800 figure: a $20M/yr brand running 8% TACoS in steady state is spending $1.6M/yr on ads. A 1 bps permanent drift puts that at $1.602M, $2,000/yr in the first year. With 15% spend growth and 8% discount rate, the three-year NPV of $2K + $2.3K + $2.65K is roughly $5.7K, but that’s the cost of NEW drift only. The compounding cost of the drift never being corrected, i.e., the ongoing 1 bps TACoS hit applied to the full spend base each year, works out to ~$52,800 over three years. That’s the difference between a one-time correction and a permanent leak.
Sensitivity: at a 10% spend growth rate the three-year NPV drops to ~$45K. At 20% growth it’s ~$62K. At 25% growth (which we see on aggressive scaling accounts) it’s ~$71K. The faster you’re growing, the more expensive each bps of drift gets. Counterintuitively, the brands most likely to ignore basis-point drift are the ones for whom it’s most expensive, fast-scaling brands assume “we’ll fix it when we plateau,” and the math is precisely the opposite.
The other dimension that makes the cost worse than the headline: drift compounds across categories of leak. A brand running 1 bps drift on placement modifiers AND 1 bps on dayparting AND 1 bps on negative-keyword cadence is at 3 bps total, not 3 separate $52K problems but a single $158K problem, because the leaks accumulate against the same revenue base. Most accounts we audit have 4-7 visible bps of drift in the TACoS number, none of which would show up as alarming in isolation. Together they’re a six-figure annual leak hidden inside a “9% TACoS” headline that looks normal.
Where the bps usually hide
Four operational hiding places. We see all four on most accounts that have been on autopilot for 12+ months:
- Placement modifiers set during launch and never revisited. Top-of-search vs rest-of-search vs product-page, the relative performance of these placements shifts quarter-to-quarter as Amazon adjusts placement density and competition changes. A modifier that was correct in Q1 2024 is rarely correct in Q1 2026. We see brands running +50% top-of-search modifiers on campaigns where the actual incremental conversion lift is +15%, paying ~3x the necessary premium on a portion of their spend.
- SB video creative that hasn’t been refreshed in 9+ months. Cosmo’s treatment of repetitive creative tightened materially in Q4 2024 and again in Q3 2025. Stale SB video CTR drops 25-40% over 12 months on identical bids. Brands that don’t refresh creative quarterly are paying for diminishing CTR at constant cost per impression, that delta is direct TACoS drift.
- Dayparting schedules that predate the last Q4. Buyer behavior shifted between 2023 and 2025; the optimal dayparting curve for category X in 2023 isn’t the same as 2026. Schedules built three years ago are typically off by 8-14 percentage points of bid weight on the wrong hours. Quiet leak, never alarming.
- Negative keyword lists that haven’t been reconciled against ST reports in 30+ days. See our 1,200-account analysis, the median time-to-negative across the industry is 23 days. Each day of drift on a wasteful keyword is incremental TACoS leak that compounds with spend growth.
None of these are glamorous. All of them are why the compounding works against you. The brands that have built a measurement loop catching all four, instrumented review of placement modifiers monthly, SB video refresh on a 90-day cadence, dayparting recalibration quarterly, negative-keyword reconciliation daily, are the ones whose TACoS stays flat as they scale. The brands that haven’t are the ones drifting 50-150 bps over 18 months without noticing, and that drift compounds into a real margin problem by the time it shows up in the board pack.
The diagnostic question
If your team can’t answer “why did TACoS move 0.4 points week-over-week” with a specific operational reason, a change you made, an algorithm shift you observed, a competitive event, you don’t have an instrumented program. You have a dashboard. The difference matters because instrumentation catches drift in week one; dashboards catch drift in month six, after the compounding has cost real money.
The instrumentation we run on every managed account: weekly TACoS pull at the SKU level, paired with weekly impression-share trajectory and weekly time-to-negative latency on the search-term report. Three numbers, every Tuesday. The discipline is asking what changed and why on every meaningful movement. Six months in, the team has a substrate that lets them catch drift at 1-2 bps before it compounds into the kind of multi-quarter leak that shows up at the wrong time. The brands that aren’t running this loop will eventually find the leak, but the recovery cycle from a six-month-undiscovered drift is much more expensive than the prevention.
What this means for your team
Three operational shifts, in priority:
First: rename the question. “What’s our TACoS this month” becomes “what’s our TACoS trajectory and which bps moved why.” The framing forces the conversation toward operational specifics rather than headline numbers.
Second: instrument the four hiding places. Placement-modifier review monthly, SB video refresh quarterly, dayparting recalibration quarterly, negative-keyword reconciliation daily (or threshold-based automation per the negative-keyword harvesting cadence we run).
Third: treat your catalog as a product with the same instrumentation discipline. Catalog drift shows up as TACoS drift one cycle later, your A+ goes stale, Cosmo reroutes traffic, your bids work harder for fewer conversions, your TACoS climbs. Catching the catalog drift early prevents the TACoS drift entirely.
None of this is glamorous. All of it is why the brands that compound on Amazon over three-year horizons keep compounding while their competitors don’t. A basis point is not a rounding error. It’s the per-bps cost of a mid-career hire, every year, forever, on a $20M brand. Treat it accordingly.
Get a free audit, we’ll pull your bps drift across the four hiding places and tell you how much it’s costing you.