Working · 2014-05PrimaryAcquisitionPaid

Exclude frequent buyers from search retargeting spend

Measure ad lift by customer type and cut spend aimed at people who would buy anyway.

Claim and evidence

The claimresulting in average returns that are negative.

eBay · nber.org · 2014-05

What the numbers show

No published number for this tactic has passed our word-for-word check yet, so there is no chart.

Why it works

Diagram, not data
Audience by purchasehistoryNew and lapsed users:keep adsFrequent buyers:excludeRecalculate ROI bysegment
Audience by purchase historyNew and lapsedusers: keep adsFrequentbuyers: excludeRecalculate ROI by segment

eBay's experiments found new and infrequent users respond to ads while frequent buyers, who account for most spend, don't, so average return was negative.

How to run it

  1. Segment audiences by purchase history
  2. Test ads against holdouts per segment
  3. Exclude loyal customers from broad ads
  4. Concentrate on new and lapsed users
  5. Recalculate ROI by segment

Ways to run it

Design options to test, not promised results.

  1. Option ACustomer list exclusion

    Upload buyer lists and exclude them from broad prospecting and non-brand campaigns.

  2. Option BHoldout per segment

    Test ads against holdouts separately for new, lapsed and loyal customers, as eBay's experiments did.

  3. Option CLapsed customers only

    Keep retargeting only for customers who have not bought in a while.

Dated examples

  1. 2014

    eBay

    Field experiments on non-brand keyword ads.

    Frequent buyers account for most of the spend, giving negative average returns; new and infrequent users respond.

    “resulting in average returns that are negative.”
    ✓ quote checked word-for-word on the page

    nber.org · 2014-05

Does it still work?

Workingas of 2014-05

2014 peer-reviewed field experiment on one large marketplace. source

Risks

  • May not generalise to small brands
  • Short-term measurement

Sources

  1. nber.org · NBER working paper 20171 · Tom Blake, Chris Nosko, Steven Tadelis · 2014-05
    “resulting in average returns that are negative.”
    ✓ quote checked word-for-word on the page

Change log

  • 2026-11-20 — harvest v0.1 (acq-paid)