Pinned headlines

Pinned USP headlines were expected to improve CTR by drawing attention and improve CR by pre-qualifying users before the click. The economic target was lower CPA despite possible CPC growth.

A/B testSplit: 50/509 performance snapshots
A/Btest design
Split50/50
KPICTR / CR / CPA-CPO / ROAS-CAC

Experiment context

Hypothesis. Pinned USP headlines were expected to improve CTR by drawing attention and improve CR by pre-qualifying users before the click. The economic target was lower CPA despite possible CPC growth.

How the test ran. Bidding and targeting were held constant to isolate the creative effect; weaker directions were used first to limit risk.

Traffic split. 50/50.

Performance results 9 snapshots · Click to expand

N/A means the metric cannot be calculated because the baseline is zero or the metric does not apply; — means no source value was provided.

project-alpha — seasonal activity
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format96,7294,5104.66%1.97$8,891.39$233.4938.08$5.18%87.57%90.69%103.56%106.04%96.17%87.97%109.32%97.00%
Old ad format110,4594,9734.50%1.86$9,245.65$265.4134.83$5.34%

Result. Unsuccessful / insufficient evidence: traffic fell 13%, CTR was broadly flat, and CPA increased by about 10%; retain the control format.

project-beta — service3
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format6,77198714.58%2.02$1,990.74$180.5611.03$18.29%95.73%89.81%93.81%101.15%90.84%113.24%80.22%126.09%
Old Ad Format7,0731,09915.54%1.99$2,191.51$159.4513.74$14.51%

Result. Successful: CR increased 26% and CPA decreased 20% for service3; roll out the pinned format.

project-beta — service4
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format9,8976186.24%2.00$1,233.81$43.1528.59$6.98%50.73%49.48%97.54%107.50%53.19%79.31%67.07%160.28%
Old Ad Format19,5111,2496.40%1.86$2,319.55$54.4142.63$4.36%

Result. Successful: CR increased 60% and CPA decreased 33% for service4 despite lower traffic; roll out the pinned format.

project-epsilon — service3
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format3,32946513.97%1.71$796.37$62.9712.65$13.54%124.78%151.96%121.79%105.53%160.36%151.19%106.07%99.49%
Old Ad Format2,66830611.47%1.62$496.60$41.6511.92$13.61%

Result. Unsuccessful for service3: CTR increased 21%, but higher CPC left CR flat and CPA 6% higher; retain the control or retest with lower bids.

project-epsilon — service4
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format7,5425397.15%2.16$1,162.45$21.9153.06$4.06%81.15%85.15%104.93%101.63%86.54%120.25%71.97%141.22%
Old Ad Format9,2946336.81%2.12$1,343.24$18.2273.72$2.88%

Result. Successful for service4: CR increased 60% and CPA decreased 33%; roll out selectively.

project-zeta — service5
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format15,4961,92612.43%2.12$4,085.69$109.1937.42$5.67%237.63%269.75%113.51%96.60%260.57%339.94%76.65%126.02%
Old Ad Format6,52171410.95%2.20$1,567.96$32.1248.82$4.50%

Result. Successful for service5: conversions increased 240% and CPA decreased 44%; roll out.

project-zeta — service4
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format7,0044967.08%2.50$1,239.92$35.1935.24$7.09%103.18%80.39%77.91%99.38%79.89%82.53%96.80%102.66%
Old Ad Format6,7886179.09%2.52$1,552.08$42.6436.40$6.91%

Result. Unsuccessful for service4: CTR fell 24%, conversions fell 57%, and CPA increased 45%; retain the control.

project-eta — service5
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format1,54920813.43%1.64$341.23$14.8922.92$7.16%61.01%55.91%91.65%106.11%59.33%33.38%177.76%59.70%
Old Ad Format2,53937214.65%1.55$575.11$44.6112.89$11.99%

Result. Unsuccessful for service5: conversions were 2.2× lower and CPA was 2.7× higher; retain the control.

project-eta — service4
LabelImpr.InteractionsInteraction rateAvg. CPCCostConversionsCost / conv.Conv. rateImpr new/oldClicks new/oldCTR new/oldCPC new/oldCost new/oldConversions new/oldCPA new/oldCR new/old
New Ad Format525529.90%1.44$74.93$0N/A0.00%71.23%65.82%92.40%103.22%67.95%0.00%N/A0.00%
Old Ad Format7377910.72%1.40$110.28$6.5016.97$8.23%

Result. Successful for service4: CR increased 28% and CPA decreased 29%; roll out only for this direction.

Conclusions

Overall verdict: mixed. Pinned headlines should not be rolled out account-wide. They improved conversion efficiency for project-beta service3/service4, project-epsilon service4, project-zeta service5 and project-eta service4, but hurt project-epsilon service3, project-zeta service4 and project-eta service5. Use direction-level rollout decisions and retain the control format where CPA increased.