SIMON HARMS
FIELD REPORT 10PAID SEARCH

FIELD REPORT 10 · PAID SEARCH · MATCH GROUP

The brand
search rebuild

Brand search sat at low spend with flat ROI. I rebuilt the account around search intent until spend and ROI grew 5x together. Every ROI figure on this page writes off half of brand revenue as cannibalization first.

5xspend, before
vs after the rebuild
5xROI, with half of brand
revenue written off
● THE HONEST ROI DIALEXAMPLE · ILLUSTRATIVE INPUTS
−100% 0% 100% 200% 300% LOSS COUNTED ROI 100%
  • COUNTED
  • WRITTEN OFF AS CANNIBALIZATION
  • AS REPORTED BY THE PLATFORM
50%
PLATFORM CREDITS$4.00per $1 of spend · ROI 300%
WRITTEN OFF$2.0050% of the credit
COUNTED$2.00per $1 of spend · ROI 100%

Illustrative example, computed live: the platform credits $4.00 of revenue to each $1 of brand spend. ROI = (credit × (1 − write off) − spend) ÷ spend. 50% is the write off I used on every brand ROI figure in this report.

READ THE REPORT
01 THE BRIEF

Brand search was stuck at low spend and flat ROI.

At Match Group I owned app and web acquisition across Google, Meta and Apple Search Ads, roughly $2,000,000 a month. Brand search was one line in that budget. It sat at low spend, and its ROI had stopped moving.

The usual fix trades one for the other. Push spend and accept a lower ROI, or protect ROI and stay small. I wanted both at once, so I restructured the account.

Brand revenue has a trap in it. Plenty of people who click a brand ad would have clicked the free organic listing right under it. So before I touched a bid, I set the counting rule: half of brand revenue written off as cannibalization, on every ROI number, before and after.

CHANNELPaid search,
brand terms
ROLERestructured and
ran the account
PEAK SPEND~$180K
a month
WHENMatch Group
2019 to 2024
02 THE COUNTING RULE

Half of brand revenue came off the top.

Two things happen on a brand results page. The ad platform credits the ad with all of it.

13SAME PERSON, TWO DOORS

Plenty of people who click our ad would have clicked our free listing two rows down. The platform still credits the ad with their revenue. That share is cannibalization, and it is the part I wrote off.

2WHY BRAND IS STILL WORTH BUYING

Rivals bid on our name. With no ad of ours above theirs, some searchers leave with a rival. The ad protects that revenue, so brand search earns a budget. It just has to be counted honestly.

Why 50%? It is a convention I chose to be conservative, informed by the incrementality work I was running at the time: more than a dozen studies across Google, Meta and YouTube, geo tests included. If a number survives a 50% haircut, nobody has to take my word for it.
03 THE REBUILD

One campaign became four intent lanes.

Same searches, two accounts. Switch views to see where each query went.

● ACCOUNT STRUCTURESCHEMATIC · SAME SEARCHES, TWO ACCOUNTS

After: four brand lanes, each with its own copy, sitelinks and bids. Negative keywords stop generic and competitor searches at the wall.

brandEXACT NAME brand app brand login brand free dating appGENERIC free dating sitesGENERIC rival nameCOMPETITOR
EXACT BRANDtop slot held against rivals
BRAND + APPstore copy, app first
BRAND + LOGINlog in sitelink, lowest bid
BRAND + FREEfree sign up copy
NEGATIVE KEYWORDSBlocked. No longer bought or counted as brand.
ONE BRAND CAMPAIGN
  • one budget, one bid
  • one ad for every search
  • broad matching, few negatives
  • generic and rival traffic counted as brand
  • one blended ROI number

EVERY LANE SPLIT AGAIN BY MARKET AND DEVICEONE BUDGET ACROSS EVERY MARKET AND DEVICE

Schematic of the rebuild. Brand searches are grouped by what the searcher wants next. Generic and competitor searches that had leaked in through broad matching are blocked with negative keyword lists.

WHAT I DID, IN ORDER

01
SET THE COUNTING RULE

Half of brand revenue written off as cannibalization, fixed before the first change and applied to the before numbers too. Both sides of the comparison carry the same haircut.

02
SPLIT QUERIES BY INTENT

Exact brand, brand plus app, brand plus login, brand plus free. Each is a different person with a different next step, so each got its own campaign.

03
CUT BY MARKET AND DEVICE

Every lane split again by country and device, so a strong market could take more budget without a weak one averaging it down.

04
CLOSE THE LEAKS

Broad matching had let generic dating searches and competitor names into the brand campaign, priced like brand and converting like generic. Negative keyword lists shut them out.

05
MATCH COPY, SITELINKS AND BIDS TO INTENT

A log in sitelink for members looking for the door, store copy for app searches, free sign up copy for price searches. Bids ran highest on exact brand, where rivals bid on our name and the top slot had to be held, and lowest on login.

06
SCALE LANE BY LANE

Budget went up one lane at a time while I watched cost per customer. It rose with spend, as it does, and I kept going while it stayed far below what a subscriber was worth.

04 THE RESULT

Spend and ROI grew 5x together.

Before the rebuild vs after. ROI is counted after writing off half of brand revenue as cannibalization, in both periods.

1x3x5x 1x3x5x SPEND · BEFORE = 1x COUNTED ROI · BEFORE = 1x THE REBUILD THE USUAL TRADE SHAPE, NOT DATA BEFORE AFTER 5x · 5x
THE USUAL TRADE

Push a stalled account harder and the extra money buys worse traffic, so ROI falls as spend climbs. The dashed line is that shape, drawn for contrast.

THE REBUILD

Spend and counted ROI each ended at 5x their level before the rebuild. The lanes let money go where the return held.

5xspend, before vs after the rebuild
5xROI, counted after the 50% write off
~$180Ka month in brand search spend at peak
50%of brand revenue written off before any ROI figure

Multiples and spend figures are from my own reporting at Match Group. Every ROI figure writes off 50% of brand revenue as cannibalization. Revenue stayed with Match Group.

05 HOW FAR TO PUSH

Cost per customer rose with spend. It stayed far below value.

Two real points from the rebuilt account as it scaled. Drag across the chart, or use the slider, to move between them. This is a separate measurement from the 5x comparison.

● CUSTOMER ACQUISITION COST VS MONTHLY SPENDREBUILT ACCOUNT · USD

WHAT A SUBSCRIBER WAS WORTH SITS FAR ABOVE THE TOP OF THIS CHART

$1$2$3$4 $0 $25K $50K $75K $100K $125K $150K $175K MONTHLY BRAND SEARCH SPEND ILLUSTRATIVE $1.50 AT $25K $3.25 AT $150K $1$2$3$4$0$50K$100K$150KMONTHLY SPEND ILLUSTRATIVE $1.50AT $25K $3.25 AT $150K
  • REAL POINT
  • ILLUSTRATIVE CURVE THROUGH THE TWO REAL POINTS
MONTHLY SPEND$25K
COST PER CUSTOMER$1.50
SOURCEREAL POINT

The two dots are real: about $1.50 per customer at $25K a month and about $3.25 at $150K a month. The dashed line is an illustrative curve through the two real points, there to show the shape. Readings between the dots are rounded to the nearest 5 cents and are illustrative.

06 WHAT CARRIES OVER

Count brand revenue like a skeptic.

01

Count brand revenue conservatively. Write off the share organic search would have caught, and print the write off next to every ROI number.

02

Split by intent. One campaign forces one bid onto searchers who want different things, and generic traffic hides inside the average.

03

Scale until marginal cost meets value. Cost per customer will rise. Keep buying while the next customer still costs far less than one is worth.

NEXT · FIELD REPORT 11 · APP MEASUREMENTTeaching Google to Bid on ValueI moved conversion measurement from the MMP to Firebase so Google Ads could bid on revenue instead of installs.

Happy to walk through any of these live.

simon@twinsalesinc.com All field reports linkedin.com/in/simon-harms Resume (PDF)