How Long to Pause Brand Keywords Before the Incrementality Read Is Trustworthy
A practical duration answer for brand-paid incrementality tests: how many days to pause, why, and when to stop early or extend.
Quick answer
For most sub-€10M stores with 3k–10k weekly branded sessions, pause brand paid for 14–21 days — long enough to cover two full weekday/weekend cycles and detect a ~15% incrementality lift at 80% power. Below 3k weekly branded sessions, extend to 28 days or switch to a geo-holdout. Never run the pause across Black Friday, a product launch, or a TV/OOH burst.
Brand keyword pause duration
The number of days you keep branded paid search off so an incrementality test yields a statistically trustworthy read.
Brand keyword pause duration is the length of the treatment window in a time-based brand incrementality test — you turn branded paid search off for a set number of days, measure what happens to total branded revenue (paid + organic combined), and compare it to a matched pre-period. The right duration is not a fixed number of days; it's whichever is longer between the minimum window needed to reach your target minimum detectable effect (MDE) at 80% power and the minimum window needed to cover two full weekday/weekend cycles. It's also capped by the revenue you're willing to put at risk if the true incrementality is high.
The wrong answer here is picking a round number. "Let's pause for a week" or "let's run it a month" both ignore the four levers that actually decide the read: branded query volume, weekday/weekend seasonality, the effect size you need to detect, and the euro exposure you can absorb.
Lever 1: branded query volume sets the floor
The test is only powered if you accumulate enough branded sessions in both the pre-period and the pause window. Below roughly 3,000 branded sessions per week, most sub-€10M stores can't detect a 15% lift inside 21 days — the daily variance is too wide.
Check your GA4 branded query segment for the trailing 8 weeks. If weekly branded sessions sit under the volume floor for a powered pause test, the honest answer is to extend to 28 days, tighten your MDE target, or run a geo-holdout instead of a time-based pause.
Volume floor rule of thumb
Under 3k weekly branded sessions → 28-day window minimum. 3k–10k → 14–21 days. Over 10k → 10–14 days is usually enough to hit a 10% MDE.
Lever 2: cover two full weekday/weekend cycles
Branded search demand is not flat across the week. Apparel and beauty typically see 30–50% higher weekend query volume; electronics and home goods often invert that pattern. A pause window that captures only one weekend gives you one weekend data point — not a distribution.
Two full weekday/weekend cycles — a hard 14-day minimum — is the operational floor even when statistical power says you could stop sooner. It lets you fit a day-of-week fixed effect and separate cyclical noise from the real pause impact.
Lever 3: the MDE you actually need
A Performance Manager sitting on a €400k/year brand-paid budget doesn't need to detect a 3% lift — the decision threshold is much coarser. Most sub-€10M stores set MDE at 10–20%: below that, brand-paid stays on; above it, spend gets reallocated to prospecting.
Tighter MDE means longer pauses. Halving your MDE roughly quadruples the required sample. If you truly need to detect a 5% incremental effect, plan for a 6–8 week holdout — or accept that a time-based pause is the wrong instrument and move to a geo-holdout design that runs continuously without capping revenue exposure.
Worked example: €5M Shopify apparel store
6,000 weekly branded sessions, 3.8% branded conversion rate, €82 AOV. Brand-paid budget €22k/month. Target MDE 15% at 80% power. The powered window comes out to ~17 days; rounded to a full second weekend, the plan is a 21-day pause starting on a Monday, avoiding any promotional peak within 14 days on either side.
Lever 4: cap by revenue-at-risk, not by days
If your worst-case incrementality is real — say brand paid is genuinely driving 40% incremental revenue — a 21-day pause on that €5M store puts roughly €35k–€45k of branded revenue at risk. Set the ceiling in euros before you start, and stop early if daily branded revenue drops more than the ceiling implies.
This is also where competitor behaviour matters. If a competitor moves in on your brand terms during the pause, your read overstates incrementality — and the organic rank recovery lag after you resume can extend the revenue drag by another 1–2 weeks. Both are reasons to prefer a shorter, tighter window over a leisurely one.
Common questions about brand pause duration
Almost never. A 7-day pause only covers one weekend cycle and rarely accumulates enough branded sessions to detect anything short of a 30%+ lift. Use 7 days only for a directional smoke test, not a decision-grade read.
28 days for most sub-€10M stores. Beyond that, organic rank drift, seasonality, and competitor incursion introduce more noise than the extra statistical power removes. If you need a longer read, switch to a geo-holdout design.
You can, but it turns the analysis into a switchback design that needs more careful modelling. For a straightforward first read, keep the pause continuous across full weekday/weekend cycles.
Don't. Any promotional peak, TV/OOH burst, or launch inside the pause window — or within 14 days on either side — poisons the read. Reschedule the test to a clean period.
Use a geo-holdout instead. You split traffic by region and keep brand paid off in the holdout regions continuously, which gives you a running read without capping revenue exposure or waiting on calendar time.
Typically 3–10 days for well-established brand terms, longer if a competitor bid on your brand during the pause. Factor this recovery lag into your revenue-at-risk calculation, not just the pause window itself.
Pause everything that triggers on your brand token — exact, phrase, and broad — plus any DSA or Performance Max asset groups that harvest branded queries. Otherwise you're measuring a partial pause and understating incrementality.
Most Performance Managers on sub-€10M stores set 15% as the decision threshold. Below that, keep brand paid running; above it, reallocate. Tightening to 5% requires a 6–8 week holdout, which usually isn't worth the revenue exposure.
No. Incrementality drifts as your organic rank, competitor set, and brand awareness change. Re-run at least annually, or after any material shift in prospecting spend, TV budget, or SEO position.
For stores over €10M in revenue with clean regional splits, geo-holdout is the better default: it runs continuously and caps revenue exposure. For smaller stores without enough regional volume, a well-planned 14–21 day time-based pause is the practical choice.
Test ideas before you ship them
Run unlimited A/B tests, attach hypotheses to outcomes, and build a searchable archive of what works — and what doesn't.