1、Measurement and Targeting Unplugged:The Symphony of Strategy with Unified Marketing MeasurementCarl Carter,SVP Media&Analytics,EMEA,CircanaDimitris Dimopolous,VP-Regional Analytics,CircanaConstantinos Kotopoulos,ACE Managing Director,CircanaCircana|Measurement and Targeting Unplugged|Page 2Measureme

2、nt and TargetingUnpluggedThe Challengeprovides a strategic view across investment channels in the short and long term,it fails to provide more frequent and tactical reads of incrementality experiments.Data complexity,data silos,and inconsistent metrics make it difficult to accurately measure marketi

3、ng impact to optimize marketing strategies.Not all industry verticals or categories are the same.Different verticals,brands,and sub-brands have different needs and therefore reliance and use of each methodology varies.Factors such as privacy regulations,data granularity,speed to insight,accuracy lev

4、els,cost and depth of insight,as well as the stakeholder use all influence approaches employed.The ever-evolving media and advertising landscape adds to the complexity of effectively measuring marketing performance and making informed decisions.Capitalizing on in-market,short-term demand versus buil

5、ding the brand,coupled with the advent of retail media networks,is leaving measurement practices out of tune.Privacy changes with 3P cookies,Apples ATT/ITP,and loss of device signals in general make attribution harder,creating a gap in speed of media performance visibility.Fragmented insights and in

6、consistent adoption of from different measurement like Marketing Mix Modeling(MMM),incrementality experiments(Lift),and Multi-touch Attribution(MTA)compete for sound.Where MMM MEDIACircana|Measurement and Targeting Unplugged|Page 3How Unified Marketing Measurement creates a harmonious melodyMEDIALik

7、e a symphony conductor bringing the different orchestra sections together,Unified Marketing Measurement integrates various measurement methods for a comprehensive and holistic view of marketing performance.It consolidates data from multiple sources,harmonizes methodologies,and provides a single sour

8、ce of truth for accurate attribution insights.By streamlining the measurement process,Unified Marketing Measurement enables marketers to perform a masterpiece,understand the true impact of their marketing efforts,and make data-driven decisions.An integrated approach across measurement solutions such

9、 as MMM,Lift,and MTA helps resolve many of our identified challenges:MMM delivers a top-down,strategic view across a broad range of traditional,digital,and evolving investment channels including short-and long-term ROI as well as retail media attribution.Lift provides a flexible way to test new and

10、existing channels and tactics and design incremental experiments at varying scales.They can be based on Geo-Lift approaches,also known as Matched Market testing.Lift studies can also be performed using 1P customer data where Circana has evolved methodology to apply continuous bi-directional priors a

11、cross Mix and MTA use cases,but our focus here is Geo-testing due to its scalability and coverage across retailers and global markets.MMM and Geo-Lift do not require user-level data to decompose marketing effects,resolving many privacy challenges and giving a strategic and tactical view in tandem.MT

12、A enables marketers to use their existing models to leverage automated performance signals to optimize key performance and conversion indicators at a lower investment level.This has historically helped us solve the challenge of speed to make inflight optimizations and marginal gains,though tradition

13、al methods are facing privacy challenges.Circana|Measurement and Targeting Unplugged|Page 4Fine-tuning the performanceWe conducted our own research into the capabilities of each approach independently and their benefits across a range of key measurement metrics.We then compared them in a unified way

14、 to see if we could tune the performance into a harmonious melody.Our key metrics included:MEDIADependencies refers to the data inputs required for each measurement methodology,the providers,and thus the friction that can exist in delivering efficiently.1DependenciesDependenciesThe data granularity

15、is related to the dependencies but is a factor compared and reviewed on its own.The reason is that data granularity can be impacted by data regulation,usage rights,and ease of access and processing.2Data granularityData granularitySpeed measures the time taken to deliver insight through the compared

16、 methodologies.Speed has become more critical than ever for decision-making,especially as many media channels provide an agile and flexible approach to course correction and replanning.3SpeedSpeedMarketing execution is a complex process with many interrelated stakeholders across different organizati

17、ons and platforms.Therefore,flexibility and agility of methodology play an important role in the success of the approach.4Measurement flexibilityMeasurement flexibilityWith each methodology,there is a varied level in which the advertiser,media agency,and publisher are involved in the process.5Requir

18、ed client engagementRequired client engagementApplicability considers how easily the insight can be implemented in the real world.It considers whether the analysis delivers results and recommendations that advertisers,media agencies,and publishers can use.6ApplicabilityApplicabilityDoes the methodol

19、ogy provide actionable insight against the objectives,and do we have confidence in the results to make decisions?Statistically,each analysis has confidence intervals(p-values)that allow us to mathematically represent the robustness,and therefore is easier to quantify and compare.7AccuracyAccuracyThe

20、 depth of insight is also judged by its ability to answer strategic and tactical questions ranging from channel spend,channel mix,and flighting through to audience,formats,placement,creative,and frequency.8Depth of insightDepth of insightCost implications cannot be ignored and must be balanced again

21、st our other comparative measures and the objectives of the analysis.9CostCostCircana|Measurement and Targeting Unplugged|Page 5MEDIAPerforming in harmonyThe results are summarized in the radar chart below comparing each methodologys score on its own and then how a Unified Marketing Measurement appr

22、oach can perform in harmony and provide meaningful outcomes across those scores.Comparison of Measurement ApproachesSPEEDACCURACYDEPTH OFINSIGHTSFLEXIBILITY IN MEASUREMENTDEPENDENCIESDATA REQUIREMENTSCLIENT ENGAGEMENTREQUIREMENTCOST543210MMMLIFTMTASCORES:1=LOW|5=HIGHSPEEDACCURACYDEPTH OFINSIGHTSFLEX

23、IBILITY IN MEASUREMENTDEPENDENCIESDATA REQUIREMENTSCLIENT ENGAGEMENTREQUIREMENTCOST543210UNIFIED MMUnified Marketing MeasurementCircana|Measurement and Targeting Unplugged|Page 6MEDIAAlthough using all methods in conjunction could provide an optimal approach,it is unnecessary to have all three toget

24、her.The right combination largely depends on the objectives of the insight required and the advertiser vertical,channels,and products being measured.For example:A typical use case for CPG providing a top-down and bottom-up approach.MMM caters for optimization of channel mix,spend levels,flighting,an

25、d scenario building,while Lift supports testing of newer channels or investment levels,tactical test,unlocking more frequent calibration of existing model priors.MMM and Lift togetherFor growing D2C brands with broader and diverse investment strategies,this combination can provide a strong holistic

26、pairing.MMM provides the broader channel mix analysis to optimize spend ratios and levels,while privacy-safe evolved attribution methods can fill the gaps with conversion analysis in real time,optimizing off-site and on-site performance-based marketing initiatives.Higher frequency MMM and MTAWhat to

27、 watch when using Unified Marketing MeasurementMarketing practitioners using Unified Marketing Measurement should be cautious about three challenges:Data quality and accuracy1Methodology harmonization2Interpretation and actionability of insights3How Unified Marketing Measurement can solve those chal

28、lengesUnified Marketing Measurement relies heavily on data inputs from various sources.Ensuring the quality and accuracy of data is crucial to obtaining reliable insights.Watch out for data inconsistencies,data gaps,and data errors that may impact the accuracy of measurement results.Regularly valida

29、te and clean data to ensure its reliability and integrity.Working directly with 1P data providers is recommended for example,using any MMM data platforms like Googles MMM data feed,Metas MMM feed,Amazon Ads,and many others.At Circana,we are privileged to have direct access to complete and granular p

30、urchase data from retailers across many markets.This allows us to assess the right purchase data to perform a range of analysis approaches with accuracy and scale such as MMM and Lift.Circana|Measurement and Targeting Unplugged|Page 7MEDIAHarmonizing the methodologys melodyUnified Marketing Measurem

31、ent involves harmonizing different measurement methodologies,such as MMM,Lift,and MTA.Its important to carefully align and reconcile these methodologies to avoid inconsistencies and discrepancies in measurement results and interpretation.Approaches to harmonization are still relatively new and there

32、fore can be approached in a few ways.Practitioners should watch out for inconsistencies in assumptions,modeling techniques,and attribution rules that may affect the accuracy and comparability of results.Circanas experience working with global CPG brands and rich retail sales data has given us many y

33、ears of lessons in approaching the challenge of harmonization,specifically the combination of MMM and Lift experiments.MMM provides a critical tool to help CPG brands understand their media mix and channel response in the longer term but requires historic spend,scale,and variation as it is based on

34、correlation.In some cases,MMM methods are evolving to have more frequent and granular insights;MMM still has shortfalls related to understanding smaller test-and-learn media executions,always-on channels,and search(pull-based media).Incrementality experiments such as Geo-Lift enable practitioners to

35、 design experiments to specially test these MMM shortfalls to identify incremental gains at a tactical level and provide sales lift,conversion,and ROAS.But these are two different methodologies providing insight at different levels the need for harmonization is now critical.Based on our experience,w

36、e recommend the following use cases to close shortfalls.Circana|Measurement and Targeting Unplugged|Page 8MEDIAMMM shortfalls=Future incrementality opportunitiesWhen reviewing MMM results,look for the following outcomes and use incrementality experiments to fill the measurement gaps:Channels and cam

37、paigns where wide intervals are present,thus confidence is reduced.These indicate incrementality experiments can assist.Channels are highly correlated(co-linearity)and therefore insights cant be provided for the channel alone.Budget change forecasting.If youre looking to predict performance of chann

38、el budget changes above 20%,we would recommend employing incrementality experiments with budget increments to produce a curve like the one in Chart A below.MMM relies on historical data;trying to predict a data point for which you have no historical evidence is statistically problematic.How to harmo

39、nizeThe outcomes of incrementality experiments provide KPIs such as iROAS,sales percent lift,and an incremental measure(value,volume,and units).These provide valuable inferred historical inputs into future MMM model runs,ensuring a more accurate representation of the variable.These inputs are used a

40、s estimations,something better known as a“prior”the concept of employing some previously known or learned information to help with model accuracy.In the example below(Chart A),we ran a search incrementality test to isolate search ROAS,and tested increased investment in search for a national food bra

41、nd(2020).The iROAS curve identified room for search spend growth and was used for model inputs to help attribute search more accurately.123Relative Spend Index1151101051009590i ROAS Index90 110 130 150 170 190 210BAUMIDHIGHSearch Spend Variation vs.iROAS250200150100500i ROAS Index99 100 101 102 103

42、104 105 106 107 108 109 110Relative Spend IndexBAUMIDHIGHYouTube Spend Variation vs.iROASA similar test here(Chart B to the left)analysed incremental YouTube spend designed after MMM failed to predict spend levels 1.5x 2x BAU.Results showed increased ROAS at 2x BAU spend,allowing for more accurate f

43、orecasting of investment changes.Chart AChart BMEDIAWorking across many global advertiser clients and marketing technology partners,Circana frequently runs incrementality experiments that often contain channel-and campaign-level experiments that overlap with MMM measurement.This provides the perfect

44、 opportunity to sync the two approaches.Like complementary sections of an orchestra,these two approaches need to meet specific criteria to be harmonized.These include:The same level of reported(or reportable)aggregation from the results such as channel,geography,and dependent variable coverage(e.g.,

45、for CPG sales,all retailers versus one retailer).The same measurement metric(e.g.,sales volume and sales units).Comparable time periods.Choosing non-aligned time periods will introduce bias(e.g.,seasonal variables and influences).123How to harmonizeAs the incrementality experiments and MMM were not

46、designed in tandem,alignment will not be perfect.However,the experiments provide a perfectly valid way to check which model runs provide the best representation of the measured channels.We recommend practitioners review the variations between the two approaches and make decisions on which model is m

47、ost valid based on the following rules:If your MMM results sit within a range of 5%10%or less versus the incremental experiments,then we can consider the MMM already valid compared to the experiments.The MMM model can be selected.If your MMM results fall above that range,it is recommended to use the

48、 experiments as priors and rerun your model.Depending on the recency of your experiments,this could be a good time to run those incrementality tests again for further validation.This is especially true in highly targeted digital channels.Existing incrementality experiments=MMM validationCircana|Meas

49、urement and Targeting Unplugged|Page 9Circana|Measurement and Targeting Unplugged|Page 10MEDIACircanas Incrementality Experiment Benchmarks data contains over 600 experiments across more than five years of media measurement,covering most digital channels and many tactical executions.This has served

50、as a vital tool in the continuous calibration of MMM,resulting in a more accurate view of channel performance.Building a benchmark for experiments provides a solid base to understand a trended view of experimental results and an accurate picture of the latest response in context to other channels an

51、d tactics.It is recommended that practitioners consider how they can work with providers to leverage their benchmarks or consider how to build their own.How to harmonizeThis use case is a clean example of experiments providing direct estimation in the model,also known as a prior.This estimation allo

52、ws you to control the expectations of a channels performance range in the model to reduce over or underestimation,producing a more valid and confident model result.Existing incrementality benchmarks=MMM calibrationHow often should you run incrementality experiments?There is no magic number for this

53、question,as the frequency of experimentation really depends on multiple factors,including:Diversity of your media channels and activity The more diverse your investments,the more likely it is that you will need higher volumes of incremental testing,and at higher frequencies,to account for synergies

54、of channels and changes in channel timing,spending,and execution objectives.Targeting and granularity of your executions With more targeted and granular executions(tactically varied),incremental testing is important as MMMs will typically struggle to decipher their impact.Having specific results for

55、 more narrow campaigns and target sets ensures their representation and value while ensuring other campaigns are compared to them more fairly,that is,contribution is not biased in either direction.123Level of spend Higher spending levels mean there is much greater decision risk at stake,and ensuring

56、 the most recent representation of your spend is vital.Therefore,we recommend incremental tests are run when:Channel spend has changed more than 10%versus prior measure spend.New channels have been introduced which have no previous measurement point.Spend levels are consistently on(always-on),such a

57、s search,and are challenging to measure.Ultimately,the more frequently you experiment,the more budget you will need.But this should be weighed up against the loss of efficiency and effectiveness by investing in the wrong insights,which would mean spending in the wrong places at the wrong levels.Circ

58、ana|Measurement and Targeting Unplugged|Page 11MEDIAConducting the perfect symphonyUnified Marketing Measurement provides a harmonious view of marketing performance so you can perform a masterpiece.But interpreting the insights and translating them into actionable recommendations can be complex.Watc

59、h out for misinterpretation of results or overgeneralization of findings correlation does not equal causation.Ensure a thorough understanding of the insights and their implications for marketing strategies and tactics.Avoid making decisions solely based on measurement results,but also consider busin

60、ess context,marketing objectives,and other factors.In particular,we find leveraging end-of-campaign reports from media partners and agencies helps contextualize the media spend and explains many of the outcomes the models and analysis unearthed.Recommendations to marketers to leverage Unified Market

61、ing Measurement Unified Marketing Measurement will help you gain a holistic understanding of marketing performance and optimize marketing strategies.Understanding your insight and decision-making needs will help you investigate which approaches will work in tandem to fulfill those needs.Embracing th

62、is approach now will put you at an advantage as we hurtle toward a more fragmented media spend allocation future.Consolidate your data sources where possible.This,in turn,will help harmonize methodologies to ensure consistent and accurate measurement of marketing impact.Leverage the insights from Un

63、ified Marketing Measurement to inform marketing budget allocation,channel optimization,and campaign strategies with all your media partners.Transparency and openness will make sounder measurement outcomes,and as the process crescendos,the quality will,too.Continuously monitor and analyze the Unified

64、 Marketing Measurement results to refine marketing strategies and drive better business outcomes.This is not a one-shot scenario;it is a continuous approach and requires consistent tracking over time.Whether working with an external practitioner or internal team developing a process to continuously

65、track and monitor progress and gaps,this will reap huge benefits and lead to innovation in your approach for years to come.Match the data to the approach and expert team.Expert teams and practitioners tend to understand and have experience in their approach and industry or vertical.Ensuring that the

66、 relevant data meets the right team and approach helps reduce the data complexity,interpretation,and implementation(knowledge and experience),allowing results to show a true reflection of observations to then align differences in outcome.Learn More.Reach out to your Circana representative or email u

67、s at .0624About UsCircana is the leading advisor on the complexity of consumer behavior.Through unparalleled technology,advanced analytics,cross-industry data,and deep expertise,we provide clarity that helps almost 7,000 of the worlds leading brands and retailers take action and unlock business growth.We understand more about the complete consumer,the complete store,and the complete wallet so our clients can go beyond the data to apply insights,ignite innovation,meet consumer demand,and outpace the competition.For more information,visit .Follow us on X:WeAreCircana.