Leveraging Marketing Investments with Marketing Mix Modeling

2,329 views
2,097 views

Published on

In the fourth installment of Copernicus’ Marketing Planning 3.0 webcast series, special guest Irina Pessin of Data2Decisions illustrated how traditional marketing mix modeling has evolved to accommodate bought, owned, and earned media channels.

She described how to ensure marketing mix modeling answers the questions most critical to improving marketing effectiveness.

Throughout her webcast, she emphasized how marketers can use this comprehensive and powerful evaluative tool to build a stronger business case for marketing investments to senior management.

Published in: Business

Leveraging Marketing Investments with Marketing Mix Modeling

  1. 1. Leveraging Marketing Investment With MarketingMix ModelingJune 26, 2013Data2DecisionsIrina PessinManaging Partner, Data2Decisions US+1 347 406 0247Irina.Pessin@d2dlimited.com
  2. 2. Welcome to Our Fourth Installment!2 weeks, 5 webcasts, improved marketing effectiveness
  3. 3. Series Schedule1. Transformational Marketing Mix Optimization Using a Virtual MarketplacePresenter: Jeff Maloy, Senior Vice President and Chief Marketing OfficerAvailable On Demand: brighttalk.com/webcast/1336/763292. Using a Virtual Marketplace to Evaluate Your Marketing StrategyPresenter: Eric Paquette, Senior Vice President ‘Available On Demand: brighttalk.com/webcast/1336/763313. Optimizing Your Media Plan for the Bought-Owned-Earned WorldPresenter: Rolf Olsen, Vice PresidentAvailable On Demand: brighttalk.com/webcast/1336/763334. Leveraging Marketing Investments with Marketing Mix ModelingDate: Wednesday, June 26Time: 1 pm EDTPresenter: Irina Pessin, Managing Partner, Data2Decisions US5. Marketing Analytics: 5 Things Every CMO Should KnowDate: Thursday, June 27Time: 1 pm EDTPresenter: Peter Krieg, President and CEO
  4. 4. CONFIDENTIAL | SLIDE 4Irina PessinManaging Partner, Data2Decisions US- Data2Decisions -Building a business case for marketingAbout us:• We partner with clients on a journey to deliver moreprofit from marketing• Analysis in 20+ markets since the start of 2001• A comprehensive technology suite, from data throughto optimisation• Evaluation of all marketing levers• Fast modelling approach supported by technology• Advertising’s long term effect captured for accurateoptimisation• Part of the Dentsu Aegis NetworkContact:Irina.Pessin@d2dlimited.com
  5. 5. CONFIDENTIAL | SLIDE 5Econometric modeling: a “white box” approach bringingtogether client knowledge with rigorous analysis• We use multivariate non-linear regression• Regression – straightforward technique –easier for clients to understand andcontribute to• Multivariate – ideal technique fordisentangling many factors changing at thesame time• Non-linear – important to understandthresholds/saturation levels for marketingactivityUpliftAdvertisingLinear models arepoor at predictingPre-modeling interviews• Tap into client knowledge to define datalist and approachInterim models• Meet half-way through modeling processto get client feedback and input• Models need to fit statistically and logicallyFinal models• Jointly developed models help buy-in andacceptance within the business• Work with client to help implement results
  6. 6. CONFIDENTIAL | SLIDE 6Standard data requirements
  7. 7. CONFIDENTIAL | SLIDE 7• Impacts can be measured over time and then aggregated across yearsto disentangle the key drivers of volumeDrivers can be disentangled and quantified
  8. 8. CONFIDENTIAL | SLIDE 8• Off trade price changes and media have driven the growth in volume• Change in impact from 2009 -2011, reflects both changes in efficiency andlevels of activityChanges in performance over time can be illustrated
  9. 9. CONFIDENTIAL | SLIDE 9Three elements to take into account when measuring theimpact of mediaSalesEffectGRPsResponse curveThe sales uplift for different investmentlevels and point of saturationSalesEffectTimeDecay rateHow long the effect lasts after theadvertising has finished0.80.3 0.25 0.200.20.40.60.81AdvertisedProductHaloProduct 1HaloProduct 2HaloProduct 3ROIHalo effectsOnto other products – ROIs candouble once halos are taken intoaccount
  10. 10. CONFIDENTIAL | SLIDE 10Capturing the long term impact of advertising key foraccurate optimizationSTLTSalesupliftJanFebMarAprMayJunJulAugSepOctNovDecJanGRPsSTLTShort and long term responsiveness to advertising• When the short term return on advertising is calculated it is oftenunprofitable for CPG brands- Strong evidence that marketing does more than just push up short termsales - key to understand how advertising impacts sales in the long term• Long term impact can be up to 3-4 times bigger than short term effect- Depends on repeat purchase, purchase cycle and advertising strategyLong term to short term ratio varies by category
  11. 11. CONFIDENTIAL | SLIDE 1101234SponsorshipCampaign3NPDCampaign2SponsorshipCampaign2SponsorshipCampaign1NPDCampaign3BrandCampaign3BrandCampaign2NPDCampaign1NPDCampaign4BrandCampaign1ROICampaign ROI and media spendsCampaign level advertising performance (ROI) can bemeasured and analyzed across message types2.4 1.3 2.9 4.8 2.9 1.9 3.9 0.8 0.9 2.9Average annualspend ($m)SponsorshipNPDBrandNOTE: ROIs have been up-weighted to estimate the long term impact (52weeks postmodelling period) and the proportion of sales not covered in the models. ROIs do notinclude halo effects on non-Brand products.
  12. 12. CONFIDENTIAL | SLIDE 12012340 50 100 150 200 250ROIBrand Size/Media CostBenchmarks from database of learnings can be used toisolate under and over performance for each brandDrop in ROIefficiency in 2011and 2012Brand size (m local currency) divided by the cost per GRP (‘000s local currency)* Quintile averages across databaseROI relative to brand size and media costBrand 2012Brand 2011Brand 2010Brand 2009Data2Decisions average*
  13. 13. CONFIDENTIAL | SLIDE 13Modeling allows us to quantify the direct sales impact ofadvertising campaigns on brands/products advertised….BrandCampaign1BrandCampaign2BrandCampaign3NPDCampaign1NPDCampaign2NPDCampaign3NPDCampaign4Spend ($m) 2.9 3.9 5.2 0.8 2.3 2.9 1.6Core Product 1 0.8 0.1 0.5Core Product 2 1.1 0.7 0.7Product 3 0.6Product 4Product 5 1.0Product 6 0.2Product 7 0.8Product 8 1.6NOTE: ROIs have been up-weighted to estimate the long term impact (52weeks postmodelling period) and the proportion of sales not covered in the models. ROIs do notinclude halo effects on non-Brand products.
  14. 14. CONFIDENTIAL | SLIDE 14…but also quantify the halo effects across otherbrands/products in the portfolioBrandCampaign1BrandCampaign2BrandCampaign3NPDCampaign1NPDCampaign2NPDCampaign3NPDCampaign4SponsorshipCampaign1SponsorshipCampaign2SponsorshipCampaign3Spend ($m) 2.9 3.9 5.2 0.8 2.3 2.9 1.6 4.8 2.9 2.4Core Product 1 0.8 0.1 0.5 0.4 0.4 0.2 0.5 0.4 0.5 0.4Core Product 2 1.1 0.7 0.7 0.2 0.1 0.1 0.2 0.4 0.5 0.2Product 3 0.6Product 4 0.5 0.2Product 5 0.1 0.1 1.0 0.1 0.1 0.1 0.1Product 6 0.2 0.1Product 7 0.8Product 8 0.1 0.1 0.1 0.1 0.1 1.6 0.1 0.1NOTE: ROIs have been up-weighted to estimate the long term impact (52weeks postmodelling period) and the proportion of sales not covered in the models. ROIs do notinclude halo effects on non-Brand products.
  15. 15. CONFIDENTIAL | SLIDE 15This ultimately allows us to quantify the total ROI of mediaacross the group – vital for optimising portfolio allocationBrandCampaign1BrandCampaign2BrandCampaign3NPDCampaign1NPDCampaign2NPDCampaign3NPDCampaign4SponsorshipCampaign1SponsorshipCampaign2SponsorshipCampaign3Spend ($m) 2.9 3.9 5.2 0.8 2.3 2.9 1.6 4.8 2.9 2.4Core Product 1 0.8 0.1 0.5 0.4 0.4 0.2 0.5 0.4 0.5 0.4Core Product 2 1.1 0.7 0.7 0.2 0.1 0.1 0.2 0.4 0.5 0.2Product 3 0.6Product 4 0.5 0.2Product 5 0.1 0.1 1.0 0.1 0.1 0.1 0.1Product 6 0.2 0.1Product 7 0.8Product 8 0.1 0.1 0.1 0.1 0.1 1.6 0.1 0.1Direct ROI 1.9 1.4 1.2 1.0 0.2 0.8 1.6Halo ROI 0.7 0.1 0.1 0.7 0.7 0.4 0.9Total ROI 2.6 1.5 1.3 1.7 0.9 1.2 2.5 1.2 1.2 0.6% halo effect 15% 47%NOTE: ROIs have been up-weighted to estimate the long term impact (52weeks postmodelling period) and the proportion of sales not covered in the models. ROIs do notinclude halo effects on non-Brand products.
  16. 16. CONFIDENTIAL | SLIDE 16• Advert Saliency was identified as a key driverof campaign ROI once the size of the brandhad been accounted for• This provided an explanation for why tworecent campaigns lower ROIs then expectedCreative diagnostic measures can provide insight into whysome campaigns performed better than others• A relationship was determined acrossseveral drinks products between base sales(from the econometric models) and ‘BrandLove’ scores• Products with a larger endorsement of thismetric tended to have higher base salesCase Study: Soft Drinks Manufacturer Case Study: Confectionery Brand50%70%90%110%130%150%0% 50% 100% 150% 200%Indexed%upliftper100TVRsIndexed Ad salience - Top box0.50.70.91.11.31.51.70.0 0.5 1.0 1.5 2.0 2.5IndexedBaseVolumeBrand I love IndexProduct 1Product 2Product 3Product 4Product 5Product 6Product 7Product 8
  17. 17. CONFIDENTIAL | SLIDE 17Exploring the interactions between paid, owned and earnedmedia and the overall impact on brand healthPaidMediaPhase 1:How does paid media driveearned WoM offline andonline?Earned Media(Offline and onlineWoM)DirecteffectPhase 2:How do paid and earnedactivity drive owned?Owned (online)Web activityDirecteffectIndirecteffectPhase 3:How do paid, owned andearned impact long termbrand health?Brand HealthCorporate reputationDirect effectIndirecteffects
  18. 18. CONFIDENTIAL | SLIDE 18• Model at the level at which the marketing activities happen and decisionsare made- Distribution, price and promotions at the SKU and account level andadvertising at the national level• Drastically increase number of observations in model• Avoid aggregation bias• More accurate and actionable outputModeling the impact of marketing at the right level is keySKU 1WalmartSKU 2WalmartSKU 3WalmartSKU 1TargetSKU 2TargetSKU 3TargetSKU 1KrogerSKU 2KrogerSKU 3KrogerAccountsSKUs
  19. 19. CONFIDENTIAL | SLIDE 19• Price elasticities in the range -1.4 to -2.8• Account 1 and Account 2 are most price elastic• Lowest price elasticity for Other accountsModels allow price elasticities to be measuredAccount Min.BasePriceMax.BasePricePrice E.Std.Price E.Product1Price E.Product2Price E.Product3Price E.Product4Price E.Product5Price E.AverageAccount 1 $1.91 $2.47 -2.81 -2.63 -2.75 -2.77 -2.34 -3.24 -2.76Account 2 $1.95 $2.49 -2.83 -2.37 -2.83 -2.50 -1.30 -4.04 -2.65Account 3 $2.09 $2.49 -2.78 -1.86 -2.12 -1.43 -0.99 -2.71 -1.98Account 4 $2.09 $2.49 -2.08 -1.13 -2.16 -1.37 -1.80 -3.33 -1.98Account 5 $2.21 $2.65 -1.99 -1.96 -2.22 -1.92 -1.73 -3.77 -2.27Other $2.36 $2.88 -1.03 -0.74 -1.33 -1.06 -2.17 -2.05 -1.40Change in base price and price elasticity
  20. 20. CONFIDENTIAL | SLIDE 20• Applying financial information (profit margins) allows us to forecast theimpact of a price change on sales volume and profit per unit• This can be used to determine price that optimises profit• In the example below increasing the retail price from its current positionwill cause a drop in profit, optimal price is $1.95Modeling can reveal optimal price pointsRetailpriceWeeklyunitsProfit perunitWeeklyprofit$1.00 1m $0.02 $20k$1.50 0.75m $0.41 $310k$2.00 0.6m $0.53 $320k$2.50 0.4m $0.75 $300k$3.00 0.15m $1.87 $280kCurrent pricepointWeeklyprofit($)Retail price ($/unit)
  21. 21. Base Volume per annum(Units)Brand RSP ($)-10% -5% Current +5% +10%$ 10.00 $ 10.50 $ 11.00 $ 11.50 $ 12.00CompetitorRSP ($)-10% $10.00 146,606 118,653 90,699 62,746 34,793-5% $10.50 168,227 140,274 112,321 84,368 56,415Current $11.00 189,849 161,895 133,942 105,989 78,036+5% $11.50 211,470 183,517 155,564 127,610 99,657+10% $12.00 233,091 205,138 177,185 149,232 121,279Profit change per annum ($) Brand RSP ($)-10% -5% Current +5% +10%$ 10.00 $ 10.50 $ 11.00 $ 11.50 $ 12.00CompetitorRSP($)-10% $10.00 -$0.97m -$1.14m -$1.48m -$1.97m -$2.62m-5% $10.50 -$0.35m -$0.47m -$0.74m -$1.17m -$1.76mCurrent $11.00 $0.26m $0.21m $0.00m -$0.37m -$0.90m+5% $11.50 $0.88m $0.89m $0.74m $0.43m -$0.04m+10% $12.00 $1.49m $1.56m $1.48m $1.23m $0.82m• If the market increases at the same rate then a price rise would be profitable• Increasing price would have a significant impact on profit if the market did not followGame plans can be run relative to key competitors in themarket
  22. 22. CONFIDENTIAL | SLIDE 22• Similar uplifts across the big four accounts- Account 2 drives the best percentage uplifts, but from a lower base level• The products with the largest base levels deliver the best volume uplifts- Product 1 deliver the 2nd highest volume uplift despite having the secondlowest percentage upliftImpact from price promotions can be quantified acrossproducts and key accountsTotalTotal
  23. 23. CONFIDENTIAL | SLIDE 23Promotions analysis could provide required outputs at bothstrategic and tactical levelsAccount 3Account 6Account 5Account 4Account 2Account 10%50%100%150%200%250%300%350%0% 10% 20% 30% 40% 50%AverageUplift(%)Average Discount (%)Average Uplift and Discount byAccount0%50%100%150%200%250%Retailer A Retailer B Retailer C Retailer DAveVolumeUplift(%)Average Uplifts from $1.50 and 2for $3 Offers£1.5 2 for £3StrategicPutting account 2 on the trend linewould grow value share by 1.5%Tactical$1.50 mechanic delivers strongeruplifts than 2 for $3 mechanic$1.5 2 for $3
  24. 24. CONFIDENTIAL | SLIDE 24Using the models to predict and optimizeBudget allocationBespoke softwareBudget settingWhat-if forecastSales with TVSales without TVTimeSalesBudgetIncrementalprofitOptimalbudgetSpendIncrementalrevenueTVPress
  25. 25. +£32m
  26. 26. CONFIDENTIAL | SLIDE 27Database systems can host key metrics and visualisationdashboards to facilitate insight generation
  27. 27. CONFIDENTIAL | SLIDE 28A technology platform to aid decision-makingModellingPlanningReportingData sources
  28. 28. CONFIDENTIAL | SLIDE 29Build preliminarymodels andanalysisFinalize modelsand analysisCollect data andassess modelingstructureReport findingsand strategyimplications anddeliver optimizerBuildoptimizationand forecastingtoolTypical project breakdownPre-modelinginterviewsWeek 1-5 Week 11-12Deliverable: Datavalidation deckDeliverable:Interim findingspresentation forclient feedbackDeliverable:Final presentationand optimizationsoftware deliveryWeek 1-5Pharmaceutical DTC/Promotion Assessment and AllocationWeek 5-8 Week 12Week 9-11
  29. 29. Marketing Analytics:5 Things CMOs Should KnowPeter Krieg, President and CEOWednesday, Thursday, June 27, 2013
  30. 30. Thank you!Contact:Irina PessinManaging Partner, Data2Decisions USIrina.Pessin@d2dlimited.com+1 347 406 0247

×