Capability: Customer Strategy

Multi-touch attribution (MTA) vs marketing mix modelling (MMM)

Posted on: 24 July 2025
Multi-touch attribution (MTA) vs marketing mix modelling (MMM)

What is marketing mix modelling (MMM)?

Marketing mix modelling (MMM) is a statistical tool that helps organisations understand and quantify the impact of marketing activities on consumers’ behaviours, sales, return on investment (ROI) and more. It breaks down an organisation’s performance by channel, incorporating various types of data to evaluate effectiveness and determine which marketing activities are most heavily influencing the organisation’s business outcomes, which we explore further in our blog on marketing mix modelling.  

Based on a series of steps, MMM begins with data collection of marketing variables, followed by an analysis of the data collected to identify relationships or patterns and building a customised model to showcase actions and results. Finally, scenario testing can be conducted to gauge possible outcomes, leveraging the results to optimise marketing strategies and bolster decision-making. 

What is multi-touch attribution (MTA)?

Multi-touch attribution values each customer touchpoint leading to conversion, with its goal being to decipher the marketing channels or campaigns that should be credited with the conversion. The intention of this is to measure the effectiveness of each channel or touchpoint so that marketers are aware of where they should focus efforts and resources and allocate future spend in the most effective ways possible to enhance customer acquisition efforts.  

Through multi-touch attribution, a more comprehensive view into customer journeys can be gained, enabling organisations to create better strategies or optimise their ad spend in line with market shifts. The ability to see how each touchpoint impacts a sale is what allows organisations to dissect customer journeys and allocate budgets accordingly.

What are the differences between multi-touch attribution (MTA) vs marketing mix modelling (MMM)?

Aggregated versus disaggregated data

Aggregated data is statistical data used in MMM that is grouped into channels, regions or times to assess trends in terms of how channels contribute to sales. Disaggregated data, on the other hand, is behavioural data that is used in MTA to gain the most detailed insights possible at user or individual level.  

Organisations require aggregate information for visibility into external trends that may be affecting marketing efforts and conversions. In comparison, the precise level of detail available through disaggregated data is critical in MTA as it is required for assigning multiple touchpoints within a customer journey.

Objective and impact assessment

MTA uses trackable customer interactions to understand the importance of each touchpoint. As a result, one of the most substantial differences between these two is their objective. MTA focuses on the impact of specific, individual touch points and their sale or conversions impact, whereas MMM focuses on the overall impact of your marketing mix and how that combination influences sales or other outcomes.

Choosing the right approach for your company

MMM’s main goal is to help organisations deduce overall business outcomes and MTA helps organisations understand the contributions of individual touchpoints to conversions or actions. MMM includes both online and offline channels, whereas MTA only includes digital channels that track individual user behaviours. 

While MTA may not be easy to implement due to ever-changing customer journeys paired with uniting all touchpoints across various devices, channels and platforms, it does enable flexibility and offers a more granular understanding of what does and does not work within marketing initiatives. This flexibility and granularity equips organisations with insights that allow for informed, data-driven decision-making for digital marketing campaigns.

When to use multi-touch attribution modelling (MTA)

Multi-touch attribution has become a staple for organisations requiring tactical insights and are focused on short-term optimisation by measuring and quantifying the impact that their digital marketing campaigns are having. The visibility that multi-touch attribution modelling provides into the success of touchpoints across a customer’s journey is unparalleled. This insight is critical for organisations to consider amidst consumers’ increasing wariness of marketing messaging. Through this, the right audiences and their respective marketing preferences can be identified across channels, enabling customised messaging to be created and the right consumers on the right channels at the right times to be reached. 

Maximising ROI can also be made possible through multi-attribution modelling by engaging with consumers in fewer though more frequent and impactful marketing messages that ultimately shorten sales cycles. 

When to use marketing mix modelling (MMM)

Marketing mix modelling should be used when needing to understand the combined impact of advertising spending, promotions, pricing and distribution channels. It can be particularly impactful for organisations that are well-established and have a plethora of data over the course of many years to work with. From media activities to external variables including macroeconomic factors and competitors’ activities and internal variables like product distribution, product changes and price changes, countless categories can be monitored for organisations to analyse data and understand the relationship between sales and these elements. Its [immunity to the everchanging privacy landscape] is also a key advantage.

How to use both approaches together

Both MTA and marketing mix modelling MMM are key approaches in the realm of marketing analytics. When used together, MMM can offer macro-level views into marketing impact on revenue, while MTA can supply granular insights into the effectiveness of specific marketing channels. Organisations that understand when and how to use both approaches will find themselves transforming their marketing strategies and maximising their ROI.  

Combining these two approaches when building an attribution strategy is often recommended. However, MMM will ultimately be most effective for gaining long-term, strategic insights that can bolster planning and financial outcomes, whereas MTA is best suited for short-term, tactical insights that can enhance day-to-day optimisation, campaigns and decision-making. 

How CACI can help

CACI supports businesses in their delivery of optimised marketing efficiency by:  

  • Determining the value and performance of activity through evolved multi-touch & econometric modelling
  • Producing results to sustain & increase growth through targeted investment & improved marketing performance
  • Delivering improved accuracy, consistency and availability of marketing performance insights
  • Enhancing capability by evolving data, technology & process
  • Supporting the provision of ongoing strategic & delivery resource.

Find out more about the impact that digital attribution modelling can have on your business by contacting us today. 

Watch a session from our recently event on how to optimise marketing performance through Commercial Mix Modelling.

Sources:

Breaking down data silos in analytics

Posted on: 30 June 2025
Breaking down data silos in analytics

In today’s experience-driven world, customer journeys are anything but linear. They begin on mobile, pause on desktop, skip to in-store, loop through social media—and often, they leave behind a trail of fragmented data that brands struggle to unify. This in turn leads to frustration and a poor customer experience.

Data silos in marketing analytics often arise from disconnected systems, departmental boundaries, and legacy technologies that don’t communicate with each other. These silos prevent a unified view of the customer, making it difficult to deliver consistent, personalised experiences across channels.

At CACI, we see this disconnected data challenge as a pivotal opportunity. We support brands in breaking down data silos and rebuilding customer understanding with clarity, purpose, and precision.

CACI’s approach: connecting the disconnected

We enable brands to identify, intercept and connect fragmented users and journeys across their entire online and offline estates.

Here’s how:

  • Unified customer view: Our data solutions stitch together disparate identifiers and touchpoints to create a single, accurate view of each customer. One profile, all channels.
  • Journey analytics: We map real customer journeys—not just clicks or page views. Our insights reveal where users drop off, what triggers conversion, and where friction hides.
  • Omni-channel attribution: Our Commercial Mix Modelling solution allows brands to understand what truly influences customer decisions, from ATL and OOH sighting through to more targeted, 1-2-1 channels. We help brands attribute value accurately and make smarter decisions about where to place marketing spend, with a channel, segment and customer lens.
  • Real-time interception: Powered by advanced analytics, we empower brands to recognise a fragmented user in the moment and serve the right message, at the right time, in the right place.

Transforming data into decisions

Disconnected data isn’t just a technical problem—it’s a customer experience problem. When brands can’t “see” their customers clearly, the customer feels it. Poor targeting. Broken journeys. Irrelevant messaging.

At CACI, we see data as the fuel for transformative brand experiences. By helping brands unify, enrich, and activate their data, we unlock meaningful personalisation, higher ROI, and lasting loyalty.

Ready to connect the dots?

Ready to break down your own data silos and unlock the full potential of your analytics? Start by identifying where you stand. Use our Digital Analytics Self-Assessment Checklist to evaluate your current capabilities and uncover opportunities for growth.

Data harmonisation: speak the same language across teams and systems

Posted on:
Data harmonisation: speak the same language across teams and systems

Unifying your data story: speak the same language across teams and systems

In the world of digital analytics, there’s a silent struggle happening behind the dashboards: inconsistency. Different teams. Different tools. Different definitions. And the result? A fragmented data story that no one can confidently act on.

This is where data harmonisation comes in. It’s the process of aligning data from disparate sources into a consistent structure, using shared definitions, taxonomies, and metrics. By harmonising data, businesses can eliminate confusion, reduce duplication, and ensure that insights are accurate and actionable across the organisation.

At CACI, we work with brands to bring order, clarity, and alignment to their data foundations—because without consistent taxonomy and standardised metrics, even the most advanced analytics can mislead rather than inform.

The hidden cost of inconsistent taxonomy & metrics

When marketing, digital, product, and analytics teams define success differently, it creates confusion. What counts as a “conversion”? How is “engagement” measured? Is “bounce rate” the same across platforms?

Without a common language:

  • Reports contradict each other
  • Decisions are delayed
  • Cross-channel comparison becomes unreliable
  • Teams lose trust in the data

In an era where fast, data-led decisions drive competitive advantage, this kind of friction is a blocker to growth.

CACI’s approach: building a solid analytical foundation

Our approach simplifies  complexity and align on the fundamentals. Here’s how:

  • Taxonomy harmonisation: Through collaboration with stakeholders, we design and implement a unified taxonomy that reflects your brand’s goals, customer journey stages, and platform specifics.
  • Metric standardisation:  Consistency in metric definitions ensures everyone speaks the same data language. We bring consistency to how key metrics are defined, calculated, and reported—so that insights are trusted, comparisons are accurate, and decisions are aligned across the organisation.
  • Governance frameworks:  We help embed processes, guardrails, and tools that ensure ongoing data integrity, even as teams scale and evolve.
  • Enablement & training: A good taxonomy isn’t just a document—it’s a mindset. We deliver practical enablement to ensure adoption and understanding across your organisation.

Why it matters

When taxonomy and metrics are consistent, analytics becomes a true strategic asset. Brands can compare performance across campaigns, channels, and regions. They can move faster, with confidence. And perhaps most importantly—they can trust the story their data is telling.

Struggling with inconsistent metrics and siloed definitions? Let’s bring clarity to your analytics and confidence to your decisions. Use our Digital Analytics Self-Assessment Checklist to evaluate your current capabilities and uncover opportunities for growth.

What is Marketing Mix Modelling (MMM)?

Posted on: 9 April 2025
What is Marketing Mix Modelling (MMM)?

For any marketing activities to be successful, understanding consumers’ behaviours and whether a channel is oversaturated is essential. While data and analysis play undeniably important roles in this, marketing mix modelling (MMM) plays an even greater one, representing the merging point of data and analysis with the psychology of consumer understanding.  

Marketing mix modelling (MMM) is a statistical tool that enables an understanding of how each part of an organisation’s marketing activity impacts consumers’ behaviours, sales, return on investment (ROI) and more. Through MMM, an organisation’s performance can be broken down by channel and various types of data can be incorporated to evaluate the effectiveness of marketing activities and determine which are making the most substantial differences to the organisation’s overall performance. 

Benefits of marketing mix modelling (MMM)

  • Enables organisations to quantify and measure marketing channels effectively to assess which drive the most sales and return on investment 
  • Equips organisations with long-term insights that will bolster planning through effective forecasting and marketing campaign generation based on previous performance  
  • Helps organisations allocate budgets according to the best performing channels due to measuring growth based on investments
  • Instils confidence due to its statistical reliability and being privacy-safe, both of which are particularly important in a post-cookie world
  • Offers organisations a holistic view of the impacts that various factors will have on achieving specific KPIs, ensuring marketers can make more informed decisions based on how and when marketing activities will impact KPIs. 

How do marketing mix modelling (MMM) & commercial mix modelling (CMM) work?

Marketing mix modelling (MMM)

Marketing mix modelling (MMM) is used by organisations aiming to understand how marketing activities impact KPIs being measured. Its ability to measure the impact that certain pricing choices, promotional offers, product launches or advertising campaigns may have on sales makes it a game-changer for organisations. 

In MMM, the dependent variable used to assess the relationship between sales and marketing activities is usually:  

  • Sales volume: to assess the impact of different marketing activities on sales 
  • Revenue: to track the amount of money generated by sales 
  • Competitor analysis: to understand how your organisation’s marketing activities are affecting your position in the market. 

In contrast, the independent variables in MMM are the marketing activities or factors that might drive those results, such as: 

  • Advertising spend: the amount invested in promotion across various channels. 
  • Price: to explore the impact of price adjustments on sales 
  • Promotions: discounts, coupons, or offers that could increase sales 
  • Distribution: the potential impact of product availability across various locations on sales. 

Commercial mix modelling (CMM)

Commercial mix modelling (CMM) is an analytical approach that examines a variety of commercial factors that drive an organisation’s performance. It begins with collecting data from across the organisation on pricing, promotions, distribution channels, products and more, combining the resulting data into a cohesive dataset. The insights presented within the dataset help organisations gauge which factors contribute most to performance and where investments result in the highest returns. It also enables organisations to test various scenarios— price changes, promotional adjustments, changes within distribution channels— to assess the potential impact on performance. Through this, organisations can optimise their overall commercial mix to grow and become more profitable.  

How does commercial mix modelling (CMM) differ from marketing mix modelling (MMM)?

While both commercial mix modelling (CMM) and marketing mix modelling (MMM) are granular approaches that help organisations analyse the impact of marketing activities, their scope, methodology and applications differ.  

Scope

CMM offers a broader approach when it comes to evaluating the marketing activities that would impact an organisation’s performance, integrating various functions to optimise revenue and profitability. It encompasses external, non-marketing data sources such as weather, seasonality, competitor pricing, interest rates, etc.  

MMM, on the other hand, is more partial, purely marketing data that offers a more detailed and expansive result. As a statistical analysis method, it quantifies the impact that marketing activities— campaigns, paid advertisements, promotions, etc.— have on specific KPIs. Focusing more on media and investments rather than a wider marketing strategy, its granularity is what marks its stark contrast to CMM.  

Despite the broad scope of CMM, it is just as granular and technical as MMM. 

Methodology

CMM blends analytics, business intelligence and strategic insights, considering both internal and external factors that can affect an organisation’s growth. The approach entails: 

  • Scoping & data auditing: Understanding the KPIs and defining whether the model should target revenue, acquisitions, renewals or some combination form the scoping basis. Data auditing includes tech and journey mapping to determine the stages comprising the funnel for lead gen and closing, as well as the tools and tech used at each stage. 
  • Data collation & cleaning: This includes a data request to outline the full scope of what can be used in the model, with cleansing, organising and playback taken into consideration to check for completeness and broad accuracy. During this stage, data is also combined and reaggregated for ingestion into the model. 
  • Exploratory analysis & feature configuration: Plotting all the raw data to understand distribution and periodicity and exploring this raw data to identify gaps and anomalies is conducted during this stage. Correlation analysis helps find feature relationships and possible collinearity, feature types are configured for use in the model and decay is applied (AdStock) to channel features to simulate the memory effect of advertising. Diminishing returns to channel features simulate channel saturation and other transformations such as smoothing or feature combination.  
  • Pre-processing & feature engineering: Calendar and dummy variables can be included to represent milestones and seasonality, with each variable transforming across a range of parameters to find the most realistic behaviour. 
  • Commercial mix modelling (an iterative process with pre-processing & feature engineering): Once the model for the approach is scoped (e.g. logistic vs. linear, pooled, nested, hierarchical) and fit for processed features to optimise accuracy and generalising power, it is then checked against existing commercial knowledge and external priors and returned to feature processing to refine variables and tune parameters accordingly. All candidate variables are imported and tested from the pre-processing stage. Finally, the model is refined continually by adjusting variables to optimise statistical measures of accuracy. 
  • Optimisation & simulations: The present channel saturation is analysed, the optimal channel mix is delegated for specific budgets and results are presented from scenario simulations to understand which channels have headroom and which are oversaturated. A budget guide is provided for optimising revenue and the ability to plan for different scenarios: mitigating headwinds, capitalising on opportunity and planning contingencies. 
  • Next steps & recommendations: Recommendations are given based on budget optimisations and added value. 

MMM, in comparison, focuses on econometric modelling and regression analysis to determine the contributions made by various marketing channels on an organisation’s outcomes. Econometric modelling is a statistical, mathematical approach that quantifies the relationship between marketing activities and business outcomes, built with historical data. Regression analysis is a technique used within econometric modelling to measure the impact of independent variables (marketing activities) on dependent variables (sales or revenue). 

Application

Senior executives and C-suite employees may use CMM for longer-term strategic planning and decision-making, whereas MMM would be used by marketing teams to optimise spending and budget allocation towards campaigns or advertisements.  

The broader scope of CMM enables senior executives and C-suite employees to gain a complete picture of the various commercial drivers and their impact on marketing rather than isolated results. On the other hand, the granularity of MMM ensures marketing teams strategically plan and forecast how changes in spending across channels might impact sales and plan scenarios accordingly. 

How to build a marketing mix model

The first step in building a marketing mix model will be to collate and prepare your data. This will involve collecting historical data on sales and marketing spend across different channels and should go back far enough in time to effectively capture market conditions and seasonality fluctuations. 

Next, selecting the appropriate model to facilitate this will be crucial. Selecting the model can come from its robustness or flexibility, catering to your organisation’s unique needs. 

Building the model will come after this. This will include defining the relationship between marketing spend and sales or other KPIs and considering carryover effects, saturation or external factors. 

Furthermore, fitting the model will use your historical data to estimate the parameters of the MMM. Once the model is fit, the results can be analysed to precisely determine their contributions towards each marketing channel. 

Finally, the insights gleaned from these results can help you adjust marketing strategies accordingly, increase budgets within the highest-performing channels and reduce it in those underperforming. 

Examples of marketing mix modelling (MMM)

Organisations across a variety of industries can apply marketing mix modelling (MMM) to lead to improved outcomes. A few of such examples include:  

  • Consumer Packaged Goods (CPG): Gathering data on sales, advertisements, campaigns and pricing can help CPG organisations understand which channels—digital advertising, TV campaigns, etc.— drive the most overall return on investment. 
  • Retailers: From seasonal promotions to discounts and the influence of both in-store and online presence, retailers can leverage MMM to understand peak performance periods, digital sales and foot traffic to allocate budgets accordingly or reassess promotional calendars.  
  • Financial Services: Financial institutions can use MMM to evaluate their multi-channel advertising efforts and ensure they are reaching the appropriate audiences, encouraging sign-ups.  

Why businesses should choose CACI to carry out CMM 

CACI supports businesses in their delivery of optimised marketing efficiency by: 

  • Determining the value and performance of activity through evolved multi-touch and econometric modelling 
  • Producing results to sustain and increase growth through targeted investment and improved marketing performance 
  • Delivering improved accuracy, consistency and availability of marketing performance insights 
  • Enhancing capability by evolving data, technology and process 
  • Supporting the provision of ongoing strategic and delivery resource 
  • Helping businesses dig into bespoke segments and utilise in-house data products to unlock insights 
  • Offer businesses location-based insights into the effects that marketing has at various levels, from stores to regions.  

Find out more about the impact that marketing mix modelling can have on your business by contacting us today. 

Click here to read our short infographic to learn how CACI’s Commercial Mix Modelling can transform your business strategy.

SOURCES:  

Turning strategy into success: enabling the right customer experience to deliver on your growth targets

Posted on: 6 March 2025
Turning strategy into success: enabling the right customer experience to deliver on your growth targets

As businesses grow, so do the expectations of their customer experiences and those delivering it. Acquiring more customers, more products and, therefore, more data comes with increased complexity and an increasing demand on those in marketing, data and IT to enable growth.  

A successful growth agenda must consider not just the business goals, but the actionable steps to achieve them. These steps should include understanding the data, the insights it will deliver and the technical capabilities of how to scale. This can be a daunting challenge given the complexity and scope of the market.  

In this blog, we will explore how an enhanced customer experience can be delivered alongside business growth and the common issues businesses face, from operating at scale and delivering high-value experiences with limited technical resource to optimising technology for growth.   

The challenges of operating at scale

Businesses experiencing rapid growth on the journey from start-up to breaking into the mid-market and beyond find themselves in a complex landscape. With growing volumes of customer data, pressure to deliver an effective customer experience and legacy technology from early in the business’ trajectory, the challenge of maintaining that growth can be significant.  

These challenges can be broken out across the pillars of Data, Technology and People and Process. However, these are fundamentally reliant on each other, with each requiring attention to enable an impactful customer experience. 

Data is critical for understanding your customers and provides the foundation for your customer experience. Leveraging your customer understanding enables personalisation and brings you closer to a 1:1 relationship. The challenge is drawing insight from your customer data, considering things like segmentation and modelling to understand behaviour, then making that understanding actionable and available to be used for personalisation.  

This brings us to Technology. Underlying martech is the engine for customer experience, fuelled by data. Businesses often hold on to legacy systems and processes, which can become limiting factors when experiencing growth. Scaling existing technology can create bloat and operational inefficiencies as the aspirations are built on unsteady foundations.  

With Data and Technology in place, there is still the strategic element to consider. Mature, data-led businesses treat their customer experience as an iterative process. By monitoring campaign performance and understanding their customers, their communications are personalised across content, timing and channel, and they are constantly assessing how they can be more relevant and engaging for their customer. 

Enabling this iterative cycle requires a nimble customer experience, paired with the breaking down of silos between marketing, data and IT functions to enable your team to work efficiently and keep up with consumer demands.  

By putting customer understanding and activation along with the right tools in the hands of the marketer, they can identify and deliver high-value activities across customer acquisition, retention and win-back.   

Delivering high impact solutions with limited technical resources

The answer to the question: “What are the high-value activities of our brand?” is often hidden within the data. Focused insight gathering and customer understanding can reveal where in the process customers churn, what marketing is effective and what could be done better.  

Often the challenge here is again one of resource: finding someone with the right skills and resources to draw insight from data, present it clearly for marketers, data experts and C-suite, and then use this to inform the customer experience.  

This process is ongoing and iterative. A one-off solve will always be limited as customer needs, products and demographics evolve over time. Therefore, effective growth requires a dedicated measurement framework to ensure the customer experience is driving results.  

By taking a customer-first approach and prioritising your high-value activity, you can articulate what you need to deliver the experience effectively, whether that is more understanding of who your customers are, better capacity to automate and personalise, or more substantial reporting to continue iterating to grow closer to your customer.  

Optimising technology for business growth

High-value activities depend on the right technology. Legacy systems often hinder growth by limiting access to real-time data and advanced capabilities required for superior customer experiences. 

Growing businesses often outgrow legacy platforms, which lack the sophistication needed for modern demands. Upgrading to advanced marketing platforms allows companies to enable personalised, multi-channel journeys, automate tasks at scale and leverage AI-powered insights, helping marketers meet evolving customer expectations with greater efficiency. 

Selecting and implementing the right technology can be a challenge, however. The martech space is incredibly crowded with, at time of writing, over 14,000 products (per Scott Brinker’s State of the MarTech report), all with their capability, functionality and requirements. To select the right tool, businesses must consider what their aspirational customer experience is, how the tool integrates with the rest of their stack and how they are going to deliver value quickly after embarking on implementation.  

What should brands do and how can CACI help?

To successfully scale up your business without compromising your customer experience, CACI suggests considering your data, technology and operational processes ahead of making major changes. The key to growth is working backwards from your aspirational state to construct an actionable maturity roadmap. This ensures you are dedicating time and effort to the immediate priorities that will bring value back to the business while working towards your goal state. 

Our tried-and-tested approach of bringing together experts on Data, Technology, People and Processes has delivered results for complex brands like EasyJet and ASOS. CACI’s data-led, customer-centric approach focuses on enabling the customer experience by understanding the overall business vision and customer needs, considering market positioning and the steps a brand can take to sustainably and effectively deliver on their ambition. 

If you are looking to accelerate customer data or technology changes by connecting and activating your insight, please get in touch to discuss what strategies and solutions that our team of experts can help you deliver.

Related case studies

How brands can start delivering on the promise of ‘real-time’ communications

Posted on: 3 September 2024
How brands can start delivering on the promise of ‘real-time’ communications

Real-time CRM communications are becoming a universal expectation from customers regardless of industry or platform. In fact, 72% of customers expect ‘immediate service’ from the brands they interact with. This can appear daunting for a business that has never considered how to begin serving real-time communications, both in how to enable it, then in how to use it effectively. Before we discuss this, however, we must first dissect the meaning of ‘real-time’.

What are real-time communications and how do they work?

‘Real-time’ refers to the capturing and processing of data. To be able to register a customer behaviour or ‘event’ immediately is a powerful thing – to then use that event to trigger a communication gives customers a sense of responsiveness and targets them when their interest and propensity is likely to be highest.

For this to work, a tracked event must take place, which is then linked to a central customer or user profile. This event will trigger an automated process within your CEP or CRM platform to then deploy a communication to that customer. A couple of examples of this are:

Real-time communications triggered by customer behaviour

There are various customer events that brands will be looking to target through triggered communications. One of the most common examples of these is an abandon basket campaign – if a customer began a purchase journey but fell out of that journey without completing it, brands can target them immediately with a prompt to finish their order while the purchase intent is still there.

Real-time dynamic messaging based on customer attributes

Depending on certain customer attributes, customers can also be served dynamic content based on what we know about them prior to starting a new web or app session. If we imagine a customer that may have purchased from a brand once, but then not returned within a certain number of months, it is possible to welcome them with a communication that recognises that and potentially incentivises them to purchase again upon starting their new session (for example – ‘Welcome back, here’s 5% off!’).

In both examples, a brand would need to tie the app or web session to a particular customer, have an event set up to register ‘basket additions that did not result in a sale’ or ‘return visits over ‘x’ months’, and to pass this information through to a chosen CRM platform to trigger an automated communication. Both are powerful use cases that demonstrate the potential value that can be unlocked through the real-time capturing and processing of data.

How CACI can help

CACI’s experts have extensive experience in helping brands start their journeys into real-time communications, from the initial identification of the right use-cases (as above) to identifying the right enablers across data and technology to make it happen.

In doing so, enabling real-time communication can be an incremental process, where use cases are tested and evaluated for their value and built upon steadily, eventually leading to a fully connected data ecosystem with more complex real-time strategies.

To see how CACI can help you begin planning your real-time strategy, contact us today.

Related content:

Why should businesses utilise the latest LLMs and latest NLP techniques?

Posted on: 30 August 2024
Why should businesses utilise the latest LLMs and latest NLP techniques?

In our rapidly evolving world, leveraging cutting-edge technologies is no longer a luxury, but a necessity, and Natural Language Processing (NLP) stands out as one of the most transformative tools available. NLP focuses on the interaction between computers and human language, this is commonly seen in systems such as Large Language Models (LLMs), Interactive Voice Response systems (IVRs), and voice assistants. These technologies have the power to revolutionise a company’s service by making interactions more efficient and effective, whilst reducing costs, so why haven’t more companies harnessed them? 

Let’s consider customer service – an area where the technology has already made significant strides. Many businesses still have systems that heavily rely on human operators, requiring them to tackle customer calls with highly specific and complex issues. Implementing new NLP systems can lessen the reliance on these human operators, leading to decreased wait-times, improved efficiency, and 24/7 availability. However, these systems often come with significant costs and require substantial infrastructure changes. If not executed properly, they can lead to unintended consequences and ruin the customer experience. Therefore, before adding new systems, you must understand and quantify why customers are contacting you and identify where systems can enhance the customer journey and reduce cost.  

What AI tools are there for text analysis

Various AI approaches are available to address a wide range of problems. We can categorise them as follows: 

  • Generative LLMs: Examples include GPT-4 (ChatGPT), Gemini, and Claude. These are the models that excel at generating content e.g. summarising a customer call.
  • Non-generative LLMs such as BERT, RoBERTa and their various forms: These models are used extensively in applications that require deep understanding of context or meaning e.g. accurately classifying known topics for a customer call.
  • Traditional NLP techniques: This category encompasses rule-based systems, Word2Vec, and more. They work well with simpler tasks. E.g. detecting if a particular service is mentioned in a customer webchat.

What’s the difference between generative and non-generative LLMs?

Fundamentally, LLMs like GPT-4 and BERT are built from the same building blocks called transformers, so what makes them differ?

Typically, a transformer is comprised of both Encoder and Decoder parts, but it’s been found that models can be specialised through stacking either encoder or decoder blocks. GPT-4, a generative LLM, is often referred to as a decoder-only architecture. This allows the model to receive an input, then generate text that is contextually relevant to the input. Not only does it mimic human-like text, but these responses can also be seemingly creative.

BERT, on the other hand, is built using encoder-only architecture, so think of it as a specialist in both reading and interpreting human language, rather than generating it. Non-generative LLMs, when utilised effectively, offer considerable power without a lot of the overheads associated with the generative LLMs. While some infrastructure is necessary for their implementation, the costs are not prohibitively high, especially when employing distilled models. For instance, users can avoid making expensive high frequency API calls to generative LLMs or using extensive computational resources. Additionally, users have greater control over model customisation, allowing them to achieve optimal performance for domain-specific tasks. These advantages make non-generative LLMs an excellent choice for handling highly sensitive data within a secure, isolated system e.g. a client’s secure inhouse database and system.

The following table offers a high-level comparison of the different NLP tools:

Are traditional NLP techniques still relevant? 

Although LLMs are highly adaptable and have great performance across a wide-range of tasks, traditional NLP techniques remain relevant due to their task-specific tunability. These methods have been in use for decades and continue to play a crucial role in various niche applications. Traditional approaches often benefit from cost-effective compute resources and specificity, but they require more manual tuning to achieve optimal results, and typically only work well on low-complexity tasks. In general, these techniques are better-suited for curated, lower-performance internal systems, where they can carry-out dedicated automated tasks inside a pipeline.

Intent classification in action 

Back to our customer service example – using a combination of NLP techniques, generative, and non-generative LLMs, we can identify the intent of customers when speaking to customer service operators. 

In the first instance, we can apply quick traditional NLP methods to identify if this alone is suitable for our task. However, due to the complexity of customer interactions, it is unlikely that this will produce robust results. The next step would be to employ a generative LLM on a subset of calls to identify intent topics. While this may provide sufficient insights to enhance the customer journey, for truly informed business decisions, it is essential to gain a holistic understanding. Therefore, quantifying the number of calls related to each topic might be of interest.  

To quantify the number of calls it is best to use a non-generative LLM like BERT, as they will outperform their generative counter parts, are much cheaper and far easier to implement at scale. Previously we have had great results using these types of techniques and methodologies in a range of different projects.

How CACI can help

If you’re looking to enhance your business with cutting-edge NLP solutions, our in-house data science teams are here to help. Contact us today to start transforming your use of data and stay ahead in the ever-evolving landscape of AI and data science.

Related posts:

Connecting paid media to first-party data: a path to enhanced customer engagement

Posted on: 29 July 2024
Connecting paid media to first-party data: a path to enhanced customer engagement

In today’s competitive market, the cost of acquiring new customers is continuously increasing. Coupled with changing privacy regulations and consumers’ growing demands for personalised experiences, traditional acquisition strategies that are heavily reliant on third-party data are becoming less effective. Herein lies the crucial role of first-party data in acquiring the right customers and effectively retaining them.

Getting started: why alignment and collaboration matters

Realising the potential value of first-party data requires effective collaboration between Paid Media and CRM teams. When these teams operate in silos, valuable customer insights and behavioural data are not shared effectively, and brands risk a disjointed experience alongside increased ad spend.

Aligning these teams around cohesive goals and strategies ensures that the rich, actionable insights derived from first-party data are used to inform and optimise paid media campaigns. This cross-team collaboration can significantly enhance targeting accuracy, message relevance, and ultimately, customer acquisition and retention.

Unlocking the value of first-party data across CRM and social

When it comes to delivering impactful CRM campaigns, particularly on highly competitive social channels, first-party data is invaluable to delivering a relevant and cohesive customer experience. By implementing the following, brands can ensure the effective and impactful utilisation of their data:

  1. Comprehensive customer profiles: By integrating data from various touchpoints—such as website interactions, purchase history, and email engagement—brands can build rich and comprehensive customer profiles. These profiles enable precise segmentation and targeting, allowing for highly personalised ad content that resonates with specific audience segments to come to fruition.
  2. Behavioural targeting: First-party data can be used to understand customer behaviours and preferences. For instance, if a customer frequently browses certain product categories but hasn’t made a purchase, targeted ads featuring those products, along with special offers or discounts, can be highly effective in driving conversions.
  3. Dynamic and personalised content: Social media platforms offer advanced tools for dynamic ad content. Brands can use first-party data to create ads that dynamically change based on the viewer’s profile and past interactions. Therefore, creating personalised and distinctive comms at key customer moments not only increases engagement, but also adds a competitive advantage through an enhanced overall customer experience.
  4. Cross-channel consistency: Ensure that the customer experience is consistent across all channels. If a customer has already purchased a product, avoid retargeting them with the same product ads. Instead, use the opportunity to introduce complementary products or services, thereby adding value and enhancing the customer journey.
  5. Real-time optimisation: Leverage real-time data to continuously optimise campaigns. Monitor customer interactions and campaign performance closely, and use these insights to make timely adjustments to your targeting and messaging strategies.

How CACI can help

In the context of rising customer acquisition costs, the alignment and collaboration between paid media teams and CRM teams have never been more critical. This strategic integration not only enhances the customer experience, but also drives better business outcomes—improving acquisition efficiency, increasing customer loyalty, and ultimately, boosting the bottom line.

As the digital marketing landscape continues to evolve, brands that prioritise the seamless integration of their marketing efforts and harness the power of first-party data will be best positioned to succeed. The future of marketing lies in breaking down silos and fostering collaboration, ensuring that every customer interaction is informed, intentional, and impactful.

CACI’s team of experts have extensive experience in helping clients enhance customer engagement through a multitude of strategies and solutions. If you or your business are ready to explore how first-party data can lead to effective customer acquisition and retention, please get in touch to discuss how we can help you.

Related content:

How to estimate affluence with satellite imagery

Posted on: 19 July 2024
How to estimate affluence with satellite imagery

When looking to understand the geodemographics of a country, a segmentation can be an invaluable tool for describing the differences between neighbourhoods to drive decision making, guiding you towards the areas where your customers or lookalikes are. Indeed, CACI’s Acorn helps thousands of organisations better understand and target their audiences within the UK market. One of the key ways Acorn differentiates between top-level Categories is by affluence, which is a crucial factor for a segmentation in a business context. 

In the UK, there is a wealth of data we can draw upon to build geodemographic segmentations like Acorn, including a robust and detailed census, land registry, and most importantly, a well-defined, small-scale Postcode system. But in foreign markets, such detailed data often doesn’t exist, and where it does, it can be of poor quality, hard to verify and at a regional level. So, how can we build a reliable segmentation in these markets? 

Satellite imagery as a novel approach

In many countries, the nuances in affluence between neighbourhoods can be gleaned not from looking at tables of data, but by looking at them from above. Satellite imagery is incredibly useful when traditional data sources are lacking, but visual differences between affluence levels are clear. Take, for example, the below images of two areas in a desert country: 

In the image on the left, there are large buildings, geometrically defined roads, pools and greenery, which is expensive to maintain in a desert country. This area is likely to be of a generally higher affluence. In the image on the right, there are buildings of uneven height, densely packed together along uneven and jagged roads. This area is likely to be of a generally lower affluence compared to the image on the left. 

We can see by eye the differences between these areas, but we can’t feasibly label all the areas of a country manually. So, how to do we do this programmatically? 

Enter Convolutional Neural Networks (CNNs), a well-established deep learning technique that’s the bedrock of image analysis. Inspired by the visual cortex of an animal, they are trained to identify the patterns and shapes in an image and use this to predict the likely classification of objects or the image as a whole.  

For successful usage of a CNN, however, quality training data is vital. In classic examples of image recognition, such as the MNIST dataset of handwritten digits, most people would have no issue labelling the training images correctly to feed to the model. This is trickier for labelling the affluence of a small area, though, as you need deep local knowledge and the time to manually label thousands of images to achieve a model with usable accuracy. CACI has invested heavily in building a robust pipeline for this process, allowing us to achieve the scale required for accurate modelling. 

H3: The unifying geography

We now have a methodology for generating some information about affluence, but we still have another problem to tackle – what geography should a segmentation be built at? 

The natural response might be to consider administrative boundaries. This is the level at which most governmental social and economic data is released, so it makes sense to consider this as an option. However, the irregularity of the shape and size of administrative regions in many countries means that it can be hard to compare areas like-for-like, hampering effective decision making. 

H3 – a geospatial indexing system developed at Uber – splits the globe into a grid of tessellated hexagons at varying scales, from the largest scale at 110 hexagons to the smallest at ~570 trillion hexagons. It’s gained popularity thanks to its ease of use, speed and availability of algorithms and optimisations for working in its geography. 

H3 is a great alternative to Postcodes in areas where they can’t feasibly be used. It can be applied consistently across a country and at a low level of granularity, meaning that any segmentation applied at this level can clearly show the differences between areas in an accurate way. It’s also easy to aggregate up to other geographies, allowing you to integrate the data into other systems where data is not so granular. 

How CACI can help

Combining the power of H3 hexagon geography with the information gained from analysing satellite images, we can gain great insight into the relative wealth of areas in countries where existing data is simply not available. 

The ability to apply image analysis, however, means nothing without deep expertise in segmentation and location strategy. By combining your knowledge of your customers with our expertise in data science, insight and location, CACI can support you in your journey. 

Whether you’re an established international organisation or looking to move into a new market, contact us today to find out how we can help you take the next step in achieving your goals.

Related posts:

Get ahead with CACI: Unlock the power of AI and ML in your CRM

Posted on: 19 June 2024
Get ahead with CACI: Unlock the power of AI and ML in your CRM

Setting the stage:  

The field of Machine learning and AI has evolved rapidly in the last few years, especially in fields where large quantities of data and quick response times to queries are crucial. But given lots of these techniques and methods have been around for a much longer period, why has it taken so long for other industries outside of small start-ups and ambitious tech giants to leverage these methods in similar ways? 

CRM is an essential component of any company’s strategy. The ability to communicate with and understand customers is more important than ever due to the low barriers to entry in highly competitive global markets. Companies have only brief moments to convince customers that they are the right choice for shopping, spending time, or engaging. Optimising these initial and subsequent contacts is paramount to success. 

Beyond just expanding your customer base and attracting new clients, CRM is vital for any company’s retention strategy. The most advanced cutting-edge models in the world are utterly useless if we don’t know how to activate and capitalize on the value they represent. 

ML Foundation:  

In the CRM space our main goals are increasing consumer retention or spend, and we do this via figuring out the most effective ways to communicate with people. This can be broken down into when to speak to them, how to speak to them and why to speak to them.  

Recommendation engines lie at the core of many of these architectures, models that are designed to figure out what you want before you even know you want it. Broadly they work by looking at the kind of customer you are, then at customers like you, then finding things that they’ve bought recently that you haven’t.  

You can even simplify this down into just looking for customers who have an identical purchase history to you. Maybe a laptop you can buy on Amazon doesn’t come with a charger, so commonly when people buy this laptop their next purchase is a charger!! (You can often see this simple logic in the “People also bought” section of Amazon). But even these simple implementations are incredibly powerful in some ways, an educated guess is always going to be better than a random one. 

So how do these methods relate to CRM? Well, the general structure can be pulled away and applied to any subject. When we think about how to engage with a customer, we’re going to look for ways we engaged with similar customers and how these performed. The customer who likes Sabrina Carpenter will probably need to be spoken to in a different way to the Motorhead fan. 

This is simple stuff, right? Well exactly, but it’s a method to show that the underlying AI processes in these platforms aren’t really all that complicated – there’s a lot of room for improvement especially when implementing bespoke solutions with larger data sets.  

The next (generative) step:  

So, we already have ML methods that can tell us when and why to talk to people, great! But what’s the next step? 

All that’s left of our final stage is how to talk to them and what to say, stages which can and are currently being revolutionised by the advent of enterprise grade Generative AI. 

A current pipeline for devising CRM processes may involve creating template communications that are then populated with more specific information, for example customers in a certain segment defined by age and tenure are assigned one template and differing segments are shown another. 

This approach can be time consuming if it needs to be completed for each campaign, and may miss a level of personalisation that people will respond to, feeling as though each message is tailored to them rather than being an email blast they just happen to be caught up in. 

Skilled AI engineers armed with LLM’s can create a unique voice for each consumer, ensuring that quite literally all communication they will ever receive are exactly personalised to them and their engagement habits with your brand. 

Imagine attempting this even a few years ago, assigning a team of people to trawl through millions if not billions of rows of data to ensure that each customer got the perfect messaging for them would have been completely impossible. 

In practice this level of granularity in communications is probably unnecessary but it speaks to the potential these models have in this space – the sky truly is the limit. 

Even starting off small with these steps, giving a small part of a communication a generative component, allowing for large scale A/B testing and continuous model training, the effectiveness of these comms will improve over time. 

Freeing this time up from your CRM team will give them more time to tackle more involved problems that can’t be automated. 

How can we help you on this journey?

Don’t get left behind. Partner with CACI and our experienced in-house data science teams to integrate cutting-edge ML and AI into your CRM processes and experience unparalleled growth and customer satisfaction. Contact us today to learn how we can help you stay ahead of the curve.

Related posts: