This client was looking for hierarchical price optimization strategy – combining the heterogeneity of individual countries, and aggregation at the region level at the same time. This would help them define and achieve revenue goals at both – country and region level. A simulator would allow them to test out different scenarios. The solution needed to be scoped for country level elasticities and simulation for aggregated region level sales goal.
Self service simulation for different pricing tactics
Product elasticity, cross product elasticity, and product lifecycle allowed us to estimate the impact of changing pricing levers with greater certainty
~2% (avg.) margin improvement across Top 5 brands
12%-17% improvement in forecast accuracy across Top 5 brands
6% reduction in inventory over 6-month observation period
The study enabled our client to constantly monitor business targets and review the pricing strategy based on quarterly performance
Pricing strategy & Optimization
A pricing strategy is a method to discover the best price for a product. But finding the right price is not as simple as the definition sounds. The client was using competitive pricing strategy – focusing on the existing market price of competitors. They were not considering their own COGS, as their assumption was that the market is saturated, and it is better to focus on the price of competitor products.
They needed a better approach to pricing – one that would maximize their profits and revenue. The client liked our approach of state space modeling to measure the stochastic components along with the use of multiplicative models to compute elasticities. To top it all, a self-service web-based pricing simulator would enable them to test out different pricing tactics
Improved revenue through strategy recommendations
Summary insights were fed into a dashboard that allowed simulation of pricing changes on contracts, recontracts and margin.
Final results were delivered through price elasticity dashboard. User inputs can be:
a. Target revenue
b. Desired margin
c. % changes in # of stores, among other factors
We overlaid business rules and constraints, competitive benchmark on top of elasticity based optimization.
High level business benefits include optimal markdown scenario, improved revenue & better liquidity. The client now creates benchmarks every quarter.
The multi-brand retail chain was facing stiff competition and reduced footfalls. All of their stores were located in shopping centres. They were looking to improve their visibility in the shopping centres and wanted to study the impact of marketing in generating footfalls. They were also investing heavily in BTL promotions including billboards within the shopping centre and promoting their brand within these centres. For ATL, they wanted to deep dive into Print effectiveness.
Right mix for driving footfalls
We used footfalls as a function of above the line (ATL), below the line (BTL), through the line (TTL), competitor activities and offers to estimate their impact on footfalls.
Marketing activities are driving ~20% of all walk-ins
Print drives ~10% of total footfalls driven by marketing
Shopping centre advertisements generate lowest RoI
Marketing activities were able to generate online activities, accounting for ~5% of all online traffic through Halo effect
Decline in footfalls. Stagnant acquisition rates.
The client – a multi-brand retail chain was facing the challenge to generate higher footfalls in its stores. They were investing heavily in multiple marketing channels including ATL, BTL, TTL activities. They wanted to understand the impact of these activities in generating footfalls. A major part of their investment was on Print ads – and desired a deep dive into print ad effectiveness – by size & position of the ads.
The client was observing a very disturbing trend – their acquisition rates were not growing despite marketing and promotional activities. There was added pressure from competitors, and they also wanted to measure the impact these competitors were having on their acquisition rates. We proposed a marketing effectiveness study to measure the impact of marketing drivers, a print deep dive analysis.
Modeling footfalls different from modeling sales in a MMM
The
multi-brand retail client was not looking at sales, but footfalls as
a KPI for marketing activities.
Actual
sales is a function of category mix and parameters that affect buying
decision Footfalls is considered a better measure of marketing impact
as it measures preference. We designed the study to:
a.
Measure the impact of marketing activities on footfalls
b. Deep dive into print effectiveness by ad attributes – size and position of ads
c.
Build a web based simulation tool to test out different spend
optimization and what-if scenarios
The
retailer has been able to re-organize their marketing mix and improve
footfalls by 7%.
The soap brand had a good visibility of how their marketing efforts were impacting short term sales. They wanted a study to understand the impact these marketing activities were having on their brand attributes – awareness, differentiation, relevance and esteem. Once the impact on brand metrics was available, they wanted to know the volume of long term sales being driven through these brand attributes and through marketing.
Marketing drives short term sales & branding
Marketing impact sales in two ways – directly, boosting short term sales and indirectly – impacting brand attributes that impact sales.
Sales driven by branding (long-term sales) is ~40% of total sales
Awareness is the biggest driver of long term sales with ~ 50% contribution
Short-term & Long-term have a combined impact of ~60% on total sales
The long term impact of marketing on sales is greater than the short term impact of marketing on sales.
Brand attributes drive significant sales volume
Impact of marketing activities on sales is difficult to model. Marketing may drive new users to try out a product – who may stick with the brand for a long period. Additionally, marketing tends to create a brand persona with a specific positioning, emotions, values and meanings. These factors have a far greater impact on sales compared with short term effects – that tend to have a decay effect and last only for a few days (or weeks in some cases).
It is therefore important to know the impact of marketing on short term sales and on brand attributes. However, this requires a two step process – the base sales needs to be estimated after accounting for short term impact of marketing through a standard market mix modeling analysis. The remaining base then needs to be decomposed into long term impact of marketing and base sales.
Impact of marketing on sales is long lasting
Marketing has dual role – to boost sales in the short term and branding attributes like positioning, reach, differentiation and awareness. Branding attributes have a significantly higher contribution to total sales volume than marketing driven sales in the short term.
However,
impact of brand attributes on sales can only be achieved through a
two-step process:
a.
Impact of marketing on short term needs to be estimated through a
market mix study
b.
The remaining base needs to be decomposed into long term base and
sales driven by marketing activities in long term
c.
The long term contributions then need to be mapped to brand
attributes like differentiation, awareness etc.
The ecommerce client was going up the Alexa ranking and figured consistently in the first page of significant number of search results. However, they were not seeing similar conversions rates across all their products. They wanted to understand the impact of multiple attributes – promotions, the product title, image size and the number of images were having on conversion rates – and if they could come up with best practices.
What influences visitors’ actions?
The client was observing different conversion rates across their product pages and wanted to understand the attribute features that drive conversions better
Price discounts have 12% higher conversions than other
promotions Products with 3 or 4 images, of half page size each have 7% more conversions
Listings with 30+ 5 star reviews has 8% higher conversion rates
The analysis helped the client make small, incremental changes to improve the shopping experience and increase the overall conversion rate by about 1.4%
Great listings. But the right way to list?
The ecommerce platform was spending consistently on digital marketing and they were getting more visits by the day. Some of the product listings were having a very healthy conversion rates of about 2% and higher. While a few other products were not even getting a conversion of 0.2% despite having similar visitor counts. They wanted to dig deeper into the problem.
They had already used A/B Testing and few optimization techniques, like single click check out on all pages, consistent product description and page structure across all products. What they needed was deeper – an understanding of how the promotions on products impacted sales differently, the role of product images, product reviews, number of reviews, overall product rating, the title (and searchability) of the listing and similar other product attributes.
Analysing the difference between a sale and a bounce
The
conversion rate analysis helped the ecommerce client bridge the gap
between what the customers are looking for, what information they
need to make a purchase decision, the product details (including
images), on-site language, content and customer reviews.
It
also helped them identify the reasons why certain product listings,
even with similar visitor counts had a very low conversion rate
compared to the ones that were popular with the customers.
We
used mixed modeling technique, with page attributes as features to
measure the impact of each attribute on conversion.
Bounce
rate metric shares a lot of similarities with conversion rate but
tell completely different attributes of the platform.
The CPG brand was already using market mix modeling (MMM) studies to estimate the sales being generated by different marketing channels. They also had a good understanding of Return on Investment (RoI) and the contribution of marketing To total sales. They were looking for a solution that would allow them to make use of the MMM results to test spend scenarios and try out multiple spend combinations across channels to find the best spend plan.
Generate Optimization results in less than 5 minute
By deploying Optimo – the web based SaaS platform enabled our client to generate results for different scenarios without customized results from their existing vendor
Generate results in less than 5 minutes
Web based, in-tool collaboration through team account
Quick, cost effective deployment
The tool transformed the way our client was using market mix modeling results, with proactive marketing spend planning at the tip of their fingers
Proactive use of MMM to plan future spends
Our client was performing market mix studies and was able to understand the effectiveness of their marketing spend only through retrospective studies. They used to get Return on Investment (RoI) estimates and response curve results. They used this to understand the sales being generated through the marketing efforts, monthly estimates of sales driven by each marketing channel and the investment saturation levels – but all as a part of retrospective study
What they needed was something more – a spend optimization tool that could help them test different spend scenarios, compare one against the other and then finalize their media plan. With our SaaS based Optimo – they got all that, and with a turn around time that they had never imagined.
Optimo is more than a spend optimization tool
Optimization
scenarios are generally custom generated.
With
Optimo, there is no waiting time. Once the MMM results are uploaded
on the tool, the client can generate following scenarios in no time:
a.
Optimized spend distribution for a fixed spend amount
b.
Optimized spend (amount and) distribution for a defined KPI e.g.
marketing sales goal of USD 80 Mn.
c.
Spend scenarios e.g. Sales impact of reducing TV spend by 50% and
increasing Digital spend by 20%
Now, there’s no waiting for the strategy and planning team. They spend their time more effectively analysing the pros and cons of any spend scenario they think of!
The specialty retailer was running multiple Promotional campaigns across stores. Some of them were well designed campaigns and a few others were in response to competitor activities. They were however not sure of which promotions were working for them and the product groups for which the promotion worked better. The client also wanted to know if running a promotion on one product group had an impact on other product group sales.
Right promotions to boost sales
Promotions took up significant percentage of the marketing budget. Their effectiveness was unclear due to multiplicity of promotions and competitor activities.
Volume decomposition by campaign and calendar
Reallocation of promotional spends using RoI estimates
Promo planning tool to simulate and build annual promo plan
The study enabled the client to simulate potential events and generate different promotions optimization scenarios.
Gap between promotions and sales
Measuring the impact of promotions involved is a difficult task due to multiple factors impacting promotional sales. The first challenges is the different types of promotions limited only by the imagination of the marketing team. Flat discounts, Buy One Get One (BOGO) promotions, cross sell & upsell offers, coupons, bundling etc. are all some common examples of promotions that businesses use.
Time (season) of promotion, the duration for which a promotion is active, the perceived value of promotions and the frequency of promotions all have a bearing on promotion effectiveness. Competitor promotional activities and their impact on sales as well promotions on one product having a halo effect on the other. We configured the study into post event analytics, pre-event prediction and pre-event optimization and state space model for sales decomposition.
Promotion planning tool
Different
promotions have different effect on sales. The primary objectives of
the study were:
a.
The promotions that are most & least effective
b.
What promotion category is most & least effective (Feature,
Display, Price Reduction etc.)
c.
Does a product group on promotion impact sales of other product
groups
d.
Impact of competitor activities on promo effectiveness
We
provided the client with a promotion planning tool to run different
scenarios and create an annual promo calendar. Promotion return on
investment (RoI) models enabled us to identify reallocation of spend
to achieve higher promo margins.
The client is a Top 5 global breakfast brand having significant marketing budget. They have a significant marketing budget and use market mix modeling to estimate the impact of marketing on sales. They however observed that cross channel effects were not being reflected in their current MMM studies and multi-channel effect was not being captured effectively. The client also wanted to know the directional relationship of channel interactions.
Market Mix for a Multi-channel world
Structural Equation Modeling (SEM) based market mix study measures the direct and indirect impact of marketing on sales, as well as latent variables impacting sales.
4% increase in marketing sales with 2% reduced spend through mix reallocation
A total indirect impact of 4.4% on sales through cross channel interactions
Aggregated indirect impact of 1.2% on TV from Weekly Ad, Radio, & Search
The estimation enabled the client to understand cross-channel interactions better and plan well for multi-channel campaigns with the recommended allocation of spend mix.
Measurement challenges in multi-channel interactions
Traditional regression based market mix modeling approaches pose certain challenges to measurement validity. These MMM studies are not designed to account for cross-channel interactions. Any such study tends to underestimate the impact of digital ads (which is generating eyeballs like never before). And the nature of regression models does not allow us to account for latent factors.
It is hard to justify investments in channels that do not drive sales directly. The traditional marketing impact measurement methodologies do not capture the impact of digital channels well. Even if interaction variables can be introduced explicitly, the direction of channel interactions is not known. We used Structural Equation Modeling (SEM) that estimates both direct and indirect impact of measured as well as latent variables on response.
A more accurate estimation of channel impact
The
growth of multi-channel marketing means a touchpoint may not generate
sales in isolation, but create a ripple effect that eventually leads
to sales. It is important to estimate these cross-channel
interactions to understand the true picture of how different channels
work in conjunction.
Structural
Equation Modeling (SEM) based approach can measure the direct and
indirect impact of marketing on sales. It also accounts for the
latent variables that are not measured explicitly.
While
regression can measure interaction effects by explicit input of
channel interactions, it is still not able to measure the direction
of the relationship i.e. it is not possible to know if TV impacts
Radio or Radio impacts TV or a bi-directional relationship exists.
SEM estimates measure the magnitude as well as the direction of
impact.
Digital Marketing Analytics can mean monitoring some basic digital metrics like visitor, page views, impressions, bounce rate etc. and using it to refine the campaign strategy – to boost these numbers. With the popularity of Search Engine Optimization & Search Engine Marketing, and agencies over-selling these, businesses should not assume that is all analytics has to offer for digital marketing.
The full potential of digital marketing analytics can be realized only if businesses are looking at the power of analytics – to predict what a customer/visitor is going to do next – and be well placed to serve them. The analytics maturity model would help businesses judge where they are in their analytics adoption journey, and whether they are data ready to move to the next level:
Fig: Analytics Maturity Model
From
a process perspective, each of the stages can be mapped to the
following phases:
a.
Basic: No/limited
understanding of the key KPIs and no clue of what to measure
b.
Developing: Fairly
developed reporting process with clear knowledge of what is happening
c.
Defined: Understanding
of what is happening and efforts are being made to understand why
something is happening through data-driven evidence
d.
Advanced: Clear view of
what is going to happen with predictive analytics solutions
integrated into the decision making process
e.
Leading: Developing
tools & innovating solutions based on advanced analytics, machine
learning and AI driven real-time decision systems
Here,
we will talk about Predictive Analytics and how businesses could
leverage predictive analytics for digital marketing and allow
themselves the competitive edge.
a.
Better conversions with Response Modeling
With
the data and content deluge, it is easy for the targeted segment to
scroll past the social media and web page content, not open emails,
ignore promotions and offers being served through banner ads, paid
marketing and so on. Predictive analytics comes to the rescue with
propensity models or supervised learning algorithms in machine
learning to predict the customers who are most receptive to a reach
out, and have higher chances of intended response. This approach can
provide you with two options:
1. Reaching out to the most receptive target audience with a fixed budget – therefore maximizing conversions
2.
Incur minimum cost to reach a defined conversion goal
b.
Campaign effectiveness and marketing mix optimization
One
of the primary goals of digital marketing campaigns (or marketing
campaign for that matter) is to generate leads and boost sales. But
how do you measure the effectiveness of digital marketing campaigns?
Digital attribution combined with market mix modeling holds the key
to understand the impact of digital marketing on sales. The
combination works because market mix modeling (MMM) estimates the
impact of digital on sales, that can then be input into the
attribution study to deep dive on what aspects of a digital
advertising (e.g. website type, time of day, festival vs. always on)
are most and least effective. Going further:
1.
An optimization model can help businesses come up with the best
marketing mix for the planned campaign (with inputs on intended spend
or defined KPI e.g. Sales goal)
2.
Optimal number of digital touchpoints and the best channel to reach
out to a lead
c.
Lead Scoring
One
of the most important applications of predictive analytics in digital
marketing is predictive lead scoring. Lead scoring simply means
assigning a score to prospects based on assessment of interactions
and transactions a prospect does with the sales rep. Sales rep and
marketers most often rely on their judgment and gut to score leads.
A
more scientific way to lead scoring is to deploy a predictive lead
scoring algorithm. Past conversion data, along with demographics,
geography, interaction and behavioral data coupled with past success
tags allows us to develop a predictive lead scoring algorithm.
d.
Cross Sell/Upsell
Marketing is all about generating sales opportunities. With digital marketing, it is easier to reach to a wider audience given the low cost of personalized communications – through SMS and emails. Add the fact that it is ~ 8x cheaper (average, across businesses) to sell to an existing customer than to acquire a new customer. Given the low cost of email and SMS, it is easy to fall in the trap of mass mailing and mass messaging trap. However, in the longer run, this may be counter productive – as intended audience may just block the SMSs or mark the email as spam. This is the worst situation to be in – not being able to reach customers when you genuinely need to communicate with them – not only for marketing.
Cross
Sell & Upsell solutions, coupled with the right medium, frequency
and time to reach have been found to generate up to 3x repeat
customers.
e.
Customer Segmentation, Cost of Customer Acquisition (CAC) &
Lifetime Value (LTV)
Though
strictly not falling under digital marketing analytics, this section
focuses on some customer analytics solutions that can be leveraged
for digital marketing analytics. Customer segmentation allows
businesses identify the key segments e.g. higher revenue generating,
more receptive to promotions. Among other things, it reduces the
customer acquisition cost (CAC). Once businesses have a clear measure
of CAC, predictive analytics can be used to forecast the revenue a
customer is going to generate in his association with the business
(known as Lifetime Value or LTV or CLTV). With the right CLTV
estimates, businesses can have an understanding of the segments (and
customers) who are going to be more revenue generating, and try to
boost LTV for other customers.
Digital
Marketing Analytics and specially, predictive analytics within
digital marketing is underused. We have a strong belief in the
growth of predictive analytics in digital marketing. If you have
questions on how predictive analytics, or advanced analytics, machine
learning and AI can help you with the business problems you are
looking at, drop us a line.
Brick and mortar retail stores are under immense pressure and the writing is clear on the wall – reinvent or perish. Experts predict the demise of brick and mortar stores and the rise of online retail. But, is it so black and white?
Shoppers these days are looking for seamless buying experience – and the indicators point towards the evolution of omnichannel retail experience that compliments offline and online buying – allowing both to co-exist. For brick and mortar retailers, it is important to understand the reasons for growth of online retail – and how they can leverage advanced analytics and machine learning to fill the gaps – and make use of data-driven insights to drive shoppers to their stores.
If we look at market share data across markets, online retail is just about 15-20% of total retail market. That means, even though ecommerce is booming, offline retail still holds the bulk of retail market share. Given the consistent, healthy growth of ecommerce – it is important to look at it as a long term, existential threat.
And once retailers acknowledge the threat of ecommerce – they need to find solutions to the benefits ecommerce offers – and find ways to counter them. We present some of these here, mainly from the perspective of data, and how it can be leveraged using advance analytics and machine learning to generate insights – and provide a rich shopping experience.
Data
advantages to ecommerce platforms:
a.
Access to customer browsing data
b.
Ease of personalization using purchase, browsing & product
similarity
c.
Access to shopper email and phone number
d.
Dynamic offers & promotions
e.
Drive customers to online store with targeted content & SEO
f. Ability to scale globally – allowing access to shoppers anywhere in the world
There are of course other benefits for ecommerce – lower warehousing cost, no (minimum) rentals for the online store compared to brick and mortar store. However, we focus our discussion on shoppers and leveraging shopper data to counter the growth of ecommerce.
1. Shopper Behavior: With the advancement of Machine Learning & AI, brick and mortar stores can make use of store traffic and heatmaps to understand the sections where shoppers spend more time, and trial and purchase behavior. This can provide retailers with insights that even ecommerce platforms don’t have – the items that a shopper considered, and the ones she tried before making a purchase decision
2.
Personalized Loyalty Programs:
Retailers need to come up with up with the right loyalty programs,
and leverage CRM data to make personalized offers. The technology and
analytics ecosystem allows retailers with the ability to develop
personalized loyalty programs. There is not need for retailers to
stick with a retention program that was designed in the 1980s as a
marketing strategy for retention. Shoppers have moved on, and
retailers should too!
3.
Shopper Contractibility:
An attractive loyalty program will encourage shoppers to share their
email and phone number. Retailers need to identify the right time,
right offer and the right place to make use of the personal
information that a shopper has shared, and not make it a nuisance for
the shopper. Location based targeting solutions should be combined
with the CRM data to identify the right place to contact shoppers
4. Dynamic Offers & Promotions: With the growth in technology, there is no reason for retailers to stick to paper based static price & discount tags. Digital tags could be leveraged to make use of shoppers’ behavior and interactions with an item – to generate dynamic, customized offers
5.
Targeted Content: The
importance of a strong, data-driven marketing team comes into play.
The marketing team should make use of ATL & BTL strategies to
reach out to potential customers. The importance of market
effectiveness, campaign deep dive and marketing channel deep dive
will allow retailers to know what works and use the analyses results
to finetune the marketing plan
6. Scalability: Brick and mortar retailers are not well placed to counter this advantage of ecommerce platforms. As a strategy, brick and mortar stores should focus on their strengths – drive hyperlocal and ensure deliveries in few hours
7. Social Platforms: Brick and mortar retailers need to invest in social insights (and not just listening) – to understand what shoppers are talking about within their segment, what products are generating social buzz and if these platforms can be leveraged to build a community of fans.
8. Own website & mobile apps: Even if a brick and mortar retailer has no plan to go online, having a website and a mobile app will allow them to engage with customers, know what products shoppers are searching and browsing for, leverage SEO & targeted content, create personalized loyalty programs – the opportunities are endless
Retailers should look at their strengths, weakness and opportunities vis-a-vis ecommerce platforms, and make use of immense competitive advantage that data-driven advance analytics, machine learning and AI have to offer. Rather than being bullied by the ecommerce platforms – it is time for brick and mortar retailers to fight back – and regain the lost ground.
Reach
out to us if you would want to discuss any of the above, or any other
solution you have in mind. Remember, it is not the size of the dog in
the fight, but the size of the fight in the dog.
The regional soap brand was having issues managing demand. Sometimes, they accumulated inventory impacting warehousing costs. At other times, they would face stock outs. They wanted to use better method to improve forecast accuracy leading to better planning and allowing better decisions. There objective was to improve forecast accuracy, assess results with a holdout sample and to identify the factors that were impacting demand.
Demand planning in a dynamic environment
The client utilized the state of the art machine learning model, combining the concepts of econometrics to identify factors impacting demand and improve forecast accuracy.
A highly accurate forecasting model with a MAPE of <8%
Accounted for internal factors as well as external factors impacting demand
Higher efficiency with reduced stock outs and warehousing costs
The model used historical sales data, product attributes, price, sales beat frequency, marketing activities and competitor activities to accurately forecast the product demand.
Variety of data sources, internal & external factors
The soap brand was facing inefficiencies in the sales process due to poor accuracy of the forecasting model. The existing forecast numbers were based on time series based model and used past sales data to generate forecast numbers. The existing model also wasn’t able to incorporate various internal and external factors into the forecasts and generate explainable and credible results. In some cases, the error was more than 40%.
We identified the need for a machine learning driven forecasting model that would improve accuracy owing to the non-linear approach. We also understood the need to utilize various internal factors such as sales beat, marketing activities, price, brand strength and external factors such as competitor pricing, number of substitute products and marketing and promotional activities.
Unmatched accuracy results in higher efficiencies
Machine
learning based forecasting models generally (not always) outperform
traditional forecasting models.
With
our approach of using not just sales data, but also data for internal
and external factors, we were able to:
a.
Reduce forecasting error by 80%
b. Simulate the impact of internal factors – price, beat frequency and marketing activities on demand
c.
Simulate the impact of competitor activities
We
used the forecasting model to predict the sales
numbers
at regional and district level. The use of sales beat and marketing
data enabled the client to plan these activities with better impact.