TULImust 2023 konkursi finalistid
Töö pealkiri
Using advanced dynamic statistical modelling system to transform marketing to the leading department for achieving financial goals-Case of PHH Group
I KOHT
Töö lühikokkuvõte
Initiative and Analytical Alley's Dynio, an advanced dynamic modeling system, aimed to establish an analytics ecosystem for PHH Group. This system ensures the comparability, holistic planning of offline and online media and predictability of ROI across various media channels. The goal is to transform the marketing department into the primary leader for achieving business goals.
Mis oli töö eesmärk? Lisa palun konkreetsed numbrilised ja mõõdetavad eesmärgid.
E-commerce media strategy relies heavily on Google Analytics data, often resulting decisions based solely on direct ad sales and on incomplete sales data, that can be off from your real sales up to 20-30%.
Why is your Google Analytics off by up to 20-30% in your revenue and the difference in revenue differs a lot from day to day? This is due to disabled cookies, sampled data and other regulative restrictions that in time limit the opportunity for Google to follow all the sales that your company is making. Analytical Alley has controlled data for multiple companies and see that on daily and weekly levels the discrepancy is up to 50% in some cases. This margin of error is not sufficient to make any strategical and tactical decisions.
The key question: Why are we planning for one of the largest e-commerce companies in the Baltics with incomplete data and unaware of how offline media impacts sales and enhances digital channel efficiency?
The goal for this project was to move away from incomplete media view from Google Analytics and transform PHH Group’s marketing/media to leading department in reaching and leading business goals and achieve the set business goals for 2023. Following brands belong to PHH Group: Kaup24 and Hansapost EE, 220.lv LV, Pigu LT and Hobby Hall FI.
For that PHH Group and Initiative with the analytical help of Analytical Alley needed to prove the importance of marketing and media in revenue creation and increase each medium ROI’s and achieve following business goals with set total budget limit for the whole Baltics and Finland.
Question with multiple country media budget is following: how to divide the budget between countries, does it depend on the size of the country, total financial results and or other factors?
PHH Group set goal for all of the brands to reach on average 37% GMV (General Market Value)/Revenue growth in 2023.
Why is your Google Analytics off by up to 20-30% in your revenue and the difference in revenue differs a lot from day to day? This is due to disabled cookies, sampled data and other regulative restrictions that in time limit the opportunity for Google to follow all the sales that your company is making. Analytical Alley has controlled data for multiple companies and see that on daily and weekly levels the discrepancy is up to 50% in some cases. This margin of error is not sufficient to make any strategical and tactical decisions.
The key question: Why are we planning for one of the largest e-commerce companies in the Baltics with incomplete data and unaware of how offline media impacts sales and enhances digital channel efficiency?
The goal for this project was to move away from incomplete media view from Google Analytics and transform PHH Group’s marketing/media to leading department in reaching and leading business goals and achieve the set business goals for 2023. Following brands belong to PHH Group: Kaup24 and Hansapost EE, 220.lv LV, Pigu LT and Hobby Hall FI.
For that PHH Group and Initiative with the analytical help of Analytical Alley needed to prove the importance of marketing and media in revenue creation and increase each medium ROI’s and achieve following business goals with set total budget limit for the whole Baltics and Finland.
Question with multiple country media budget is following: how to divide the budget between countries, does it depend on the size of the country, total financial results and or other factors?
PHH Group set goal for all of the brands to reach on average 37% GMV (General Market Value)/Revenue growth in 2023.
Kirjelda lühidalt töö teostamise etappe.
Analytical Alley created Dynio, an analytical modeling system that uses each brands real sales and all other data for the market: macro data (inflation, wages, seasonality), weather, competitor actions in media and business, and PHH Group brand’s business and media activities on a daily level.
Designed for adaptability, it adjusts to changes in marketing KPIs or optimization goals. This gives the model opportunity to re-learn if marketing KPI’s change from example from profit to GMV or other way around or revenue goal changes to budget restriction goal. Therefore, the system was created from combinations of the most relevant mathematical methods to get the best of each approach and have the most accurate predictions of KPI for the future periods.
Analytical Alley's methodology involves multivariate linear modeling to identify channel and business impacts on GMV. Results feed into regARIMA and Random Forest models, ensuring 95% accuracy in predicting the next month's GMV based on planned media and business actions. Continuous updates capture changes in mediums, macroeconomics, competitor activity, and business aspects.
Models undergo constant monthly refinement and updates, accounting for changes in mediums, macroeconomics, competitor activity, and business aspects. Daily re-modeling gives insights to media planning, optimizing frequency and media ROI based on natural demand fluctuations.
Each month, Initiative crafts analytics-driven media plans completely based on the predictive models output and business goals to optimize media mix, timing and frequency for goal attainment.
Designed for adaptability, it adjusts to changes in marketing KPIs or optimization goals. This gives the model opportunity to re-learn if marketing KPI’s change from example from profit to GMV or other way around or revenue goal changes to budget restriction goal. Therefore, the system was created from combinations of the most relevant mathematical methods to get the best of each approach and have the most accurate predictions of KPI for the future periods.
Analytical Alley's methodology involves multivariate linear modeling to identify channel and business impacts on GMV. Results feed into regARIMA and Random Forest models, ensuring 95% accuracy in predicting the next month's GMV based on planned media and business actions. Continuous updates capture changes in mediums, macroeconomics, competitor activity, and business aspects.
Models undergo constant monthly refinement and updates, accounting for changes in mediums, macroeconomics, competitor activity, and business aspects. Daily re-modeling gives insights to media planning, optimizing frequency and media ROI based on natural demand fluctuations.
Each month, Initiative crafts analytics-driven media plans completely based on the predictive models output and business goals to optimize media mix, timing and frequency for goal attainment.
Millised olid töö tulemused? Kas seatud eesmärgid said täidetud?
From PHH Group models we found media contributes 45%-60% of total GMV, varying by brand. With proactive analytical media planning, PHH Group achieved revenue/GMV goals for 4 of 5 brands, elevating total media ROI by 22% compared to 2022 till September 2023.
The increase in medium specific ROI levels were achieved by daily and weekly optimized investment levels and natural demand based planning to leverage the effect for the moments, where customers were most naturally inclined to make the purchase.
One brand fell short due to intense competition with aggressive pricing and media strategies from main rivals. From the modelling system predictions we saw that media investments needed to reach set goals for that company turned out to be too large to be profitable and therefor budget limit optimization was chosen to still get higher media ROIs without reaching previously set business goals.
The system facilitates pre-testing of different planning methodologies, evaluating marketing changes' impact. For example, evaluate if it would be best to have 4 week campaigns with average investments level or 3 week campaigns with high investments levels and what would be the revenue gain or loss from these actions. This preemptive correction prevents actions that could diminish GMV.
The system empowers the CMO to demonstrate marketing efficiency to management, emphasizing the significant negative effects of reducing marketing investments on GMV. This prevents planned decreases in media investments, allowing for strategic testing without compromising GMV.
The increase in medium specific ROI levels were achieved by daily and weekly optimized investment levels and natural demand based planning to leverage the effect for the moments, where customers were most naturally inclined to make the purchase.
One brand fell short due to intense competition with aggressive pricing and media strategies from main rivals. From the modelling system predictions we saw that media investments needed to reach set goals for that company turned out to be too large to be profitable and therefor budget limit optimization was chosen to still get higher media ROIs without reaching previously set business goals.
The system facilitates pre-testing of different planning methodologies, evaluating marketing changes' impact. For example, evaluate if it would be best to have 4 week campaigns with average investments level or 3 week campaigns with high investments levels and what would be the revenue gain or loss from these actions. This preemptive correction prevents actions that could diminish GMV.
The system empowers the CMO to demonstrate marketing efficiency to management, emphasizing the significant negative effects of reducing marketing investments on GMV. This prevents planned decreases in media investments, allowing for strategic testing without compromising GMV.
Mis oli töös sellist, mis vääriks loomingulisuse ja uuenduslikkuse osas esiletõstmist?
For e-commerce the Google Analytics seems like the holy grail, but this is only due to the situation that it seems that there are no better solutions. Google Analytics does not only have very significant financial data results missing (due to loss of data from disables cookies, sampling and other regulative problems that are increasing in time), but also offers only short term view of performance analytics leaving out long term mass media effect, which in many cases increases the effectiveness of performance channel as they are the last touchpoints in media funnel and in many cases offline investments are more than 40% of the total investment and that is completely left out from Google Analytics.
Moving away from Google Analytics based performance marketing takes a lot of guts, but this approach has shown its benefits. Especially has it has shown to give the company more long-term brand effect, which keeps the sales up even when direct performance marketing investments are decreased.
This platform gives e-commerce companies the opportunity to move to holistic (online and offline) media channels ROI based planning to gain to full use the growing potential of the company and not get stuck in short term sales effect, where the struggle to achieve increasing business goals can only come from continuously increasing performance media investments.
Moving away from Google Analytics based performance marketing takes a lot of guts, but this approach has shown its benefits. Especially has it has shown to give the company more long-term brand effect, which keeps the sales up even when direct performance marketing investments are decreased.
This platform gives e-commerce companies the opportunity to move to holistic (online and offline) media channels ROI based planning to gain to full use the growing potential of the company and not get stuck in short term sales effect, where the struggle to achieve increasing business goals can only come from continuously increasing performance media investments.