Yes, but since the data precedes implementation of our tags, historical online data will not be available at the user-level. Site visits, conversion events and online marketing data will be requested at the aggregate level (e.g. total daily site visits, total conversions by event, etc.).
Convertro will continue to implement site and marketing tags to collect user-level online data as we do today for MTA. In addition, UMAP requires the collection of historical time-series data at the aggregate (aka “market”) level. Typically, we will request three years of historical data at weekly granularity for the initial model build. Then, feeds will need to be set up to collect aggregate-level data on an ongoing basis for data refreshes, ideally at either a weekly or monthly cadence.
Successful adoption of UMAP will require our clients to embrace change in organizational behaviors and mindsets. We know that adoption of new tools and changing old ways of thinking often fail due to employee resistance and lack of management support, especially for enterprise level organizations. Successful transformation requires commitment and buy-in, not just within our client’s Marketing team, but often across different functional teams, including Finance, the C-suite, etc. As part of our UMAP solution, we advise our clients in change management best practices to help them successfully transition into becoming more data-driven marketing organization.
In addition to Weather, Location, Purchase Cadence, and other external factors currently factored into MTA, we can bring in other external factors, such as Competitive Effects, Economic Factors, Product Ratings/Reviews, Brand Tracking Studies, and others.
In addition to all offline and online media channels, we can measure the marketing impact of other marketing sources, such as Experiential Events (e.g. trade shows, live sporting events), Public Relations, Face-to-Face Sales Associate Interactions, Trade Promotions (e.g. in-store displays), Price Promotions (e.g. dealer incentives for automotive), Inventory.
Convertro utilizes a new form of the Micro-BLP type heterogeneous consumer logit model (Berry, Levinsohn, Pakes, 2004) in a system of equations form. This general consumer-level market demand specification is applicable to a wide variety of market response models due to its construction from a consumer-level model of choice. This allows Convertro to specify a model that reflects the nature of the market ecosystem in which a brand’s customers engage, and captures the way consumers respond to a brand’s marketing activities and engage with the brand and its products.
We realize this is a lot of data science talk. We happily set up deep-dives and model discussions with your teams to be sure all questions are answered in a digestible manner.
UMAP is an internally consistent and generalizable model of consumer behavior from bottom to top, and back down. The framework establishes a method to incorporate multiple sources of data, recording multiple consumer behaviors, as well as records of market stimuli at multiple levels of aggregation with overlapping information. Competitor models unknowingly skew and blurr resolution with fused, inconsistent sets of models estimated in various stages.
The consistency of the UMAP model ensures that the results upon which actions are evaluated and executed are both structurally concordant and glean the maximal amount of information from each and every data source, exacting the complete, measured response of consumers and markets to the full spectrum of market stimuli.
Whether consumers convert on online or at retail location or via call center, and whether they are motivated to respond due to a TV spot, a price promotion, or a digital marketing tactic, or a combination of these things, our unified model simultaneously extracts all the available information from the data and harmoniously links the bottom and the top.
Much like the robust and scalable stack that makes Convertro the best in data ingestion, model training cadence, and intuitive dashboard delivery, our unified science scales to any data situation and is retrofittable to new sources or data at ever-finer levels of detail. This ensures that our solution is widely applicable and can evolve over time as a client’s business changes.
Convertro’s unified modeling approach can leverage aggregate mass-media data at the same time as shopper-specific digital media exposure, as well as environmental factors. This means that we can have multiple data sources simultaneously measuring some of the same things (e.g. an individual conversion and the total conversions for, say, a product), informing the single model of consumer choice with more precision than a marketing mix model, which on its own only looks at average consumer response, and with more breadth and coverage than an multi-touch attribution model, which on its own does not account for total market demand.
This is a departure from the way we typically think about an econometric model, with each stream of data being a separate dependent variable; rather, multiple streams of data can be fed to inform key dependent variables. In other words, we have a model for consumer choice that uses two different streams of data, total demand and individual demand, to estimate a unique relationship between marketing and demand (e.g. sales and/or quotes). This means that even where we almost never observe all mass-media, we can still estimate a consumer choice model for conversions without necessarily observing mass-media at the consumer level.
Marketing Mix Models were introduced in the ‘80’s and give executive marketers a top-down view over the factors impacting their business, along with strategic recommendations on how to spend marketing dollars. Multitouch attribution has only been possible in the last 7 years, as more user-level data became available. Multitouch attribution gives mid-level marketers actionable insights about how marketing channels and tactics are performing.
Both of these methodologies have both pros and cons.
MMM includes both marketing and non-marketing factors which impact sales/revenue. MMM typically helps users do annual or semi-annual planning across all marketing activity. Executives need this holistic view.
MTA builds all its recommendations and insights from the user-level up. This makes MTA great for accuracy and actionability down to the lowest levels of advertising activity. Channel managers and marketing managers need this granularity.
MMM uses aggregated data, so insights at more granular levels are impossible. MMM typically requires heavy manual data wrangling and can’t refresh in real-time.
MTA can’t account for many non-marketing factors, because the data doesn’t exist at the user-level. Without these additional factors, MTA results can misinterpret marketing performance.
As advertisers become more sophisticated in their marketing activities, marketing analytics needs have grown and evolved as well. This means more advertisers want a single, unified analytic tool which provides both strategic and tactical decision support. MMM and MTA cannot be stapled together to form one unified view. This is why we have created the Unified Marketing Activation Platform (UMAP).
Our goal is to help marketers make smarter, data-driven decisions to make their marketing dollars work harder and faster. Why? Increased marketing efficiency (ROI) with speed and at scale will improve overall business performance and build a competitive advantage.
As a marketer, you face a lot of challenges that inhibit you from making better, more timely decisions, both strategic and tactical ones, leading to what’s commonly referred to as “analysis paralysis.”
No true single view of the customer
Too many gaps in customer data
Lack of full visibility into customer’s complex, multi-channel, multi-device path to purchase
Too many disparate tools and platforms
Too many data and research streams to manage
Lack of single source of truth
Lack of truly actionable insights
Insights are not prescriptive enough for real-world execution
Difficulty measuring/demonstrating insights-driven results
Insights not at user-level granularity for effective and efficient planning and course correction
Highly fragmented media
Organizational impediments- Ultimately, you need the right people, tools and processes in place to realize real improvements
Siloed budgets across marketing channels and teams make optimizations slow and challenging to execute
Lack of sponsorship and commitment from senior leadership makes change management difficult
Marketing measurement and analysis is an continuous process that requires commitment to testing and learning over time – can’t boil the ocean; analytics proficiency doesn’t happen overnight
Lack of confidence in current methodologies
Lack of trust or buy-in obstructs adoption of any new analytics solution
Knowledge that something is inaccurate (e.g. last click), yet difficult to migrate to a different approach due to concerns about $ and resources
Macro level, aggregate level approach is inadequate for surgical optimizations (e.g. traditional MMM just doesn’t cut it anymore)
Inability to quickly identify and correct issues
Slowness in implementation
Lack of integration creates workflow friction and issues that slow a marketer’s ability to act
Inability to “fine tune” or course correction
The time it takes to generate attribution results depends on the time it takes to capture an adequate volume of observations to yield statistically significant results. On average, it takes Convertro's algorithm an approximate 1,000 conversions and at least 100 marketing exposures to deliver accurate results. The algorithm is able to build attribution weights for all the sources but for sources with low volumes the confidence will be comparatively lower than other sources. Our attribution weightings are applied only for the level of granularity where we have enough confidence in the weighting. For example, if there are very few observations of a keyword in the data set (with therefore low confidence in the weighting), the attribution weighting applied will be that of the “subchannel” level - in this case, the search engine.
It depends on the type of data!
To collect online data, we use our native tracking tags as experience has shown that other methods require a great deal of time, effort and cost on both ends and almost never yield satisfactory coverage or results since they depend on data provided by third parties or compiled by humans (!!), without being able to verify the quality of the information. We have a simple universal server-side tag that works with most tag managers. Collecting offline marketing data is a bit trickier. For TV and radio, we will with your offline advertising agency to collect post-log reports on a weekly basis, transmitted to a secure FTP. Typical parameters include flight and cost, date/time stamp, network, program, creative length, time of spot, GRP, etc.
Convertro is also able to collect other type of offline data, such as in-store sales, phone orders or catalog feeds. Our most popular proprietary solution involves placing a view pixel within a confirmation email. This makes it possible for our customers to tie these users to prior online activity without sharing private user information with us. For some customers, we are able to match almost 100% of offline sales. Other customers that have different conversion data can feed them into our system and match it to online activity by partnering with LiveRamp. These matches usually have a success rate between 30%-50%. Phone orders are tracked by utilizing a smart combination of our in-house approach, the inputting of special codes, or by third party vendors such as Mongoose and ResponseTap.v
Yes! Convertro integrates with dozens of third-party vendors and solutions within the advertising technology ecosystem, including the industry’s most popular DSPs, DMPs, email marketing management platforms, ad servers, affiliate exchanges, bid tools, and BI tools. These integrations enable us to extract spend, exposure and campaign data from these systems, and in many cases pass our attribution data back for surfacing in the system (e.g. Marin, Kenshoo, Searchforce, etc.). This means you can put your marketing optimization on autopilot.
Standard spend integrations include:
- Google Adwords
- Bing Ad Center
- Bing PLA
- Yahoo RightAds
- Google Display / Content Network
- YouTube Ads (any type)
- Dedicated Media
- Conversant (Commission Junction, Dotomi, MediaPlex, ValueClick)
- Commission Junction Partner Tracking
- Impact Radius
- Google Shopping / PLA
- Amazon Seller Central
Note that all spend integrations pull data daily via API (or scheduled report, if API is not available). This means that not only you have the most updated ROI calculation immediately, but also that you don’t need to waste time and effort in data wrangling and file transfers!
Implementation timing is driven by resources allocated from your team and the amount of data ingestion. Some implementations can take as little as a few weeks.
After 30 days of statistically significant data collection is when Convertro will begin making systematic holistic and granular cross-channel recommendations directly in the dashboard.
Multicollinearity between predictors affects all algorithmic attribution models, including sophisticated ones with thousands of predictors as Convertro’s multi-touch algorithmic model.
To make sure we don’t fall in the collinearity trap, we use a modeling approach called Regularization.
Regularization can in the simplest way be thought of as a generalization of Occam’s razor reasoning to model fitting. That is, we’re balancing the effort of getting the best fit of our model to your data with model complexity (i.e. – we penalize for more “complex” models, preferring the simplest explanation of the data). We gain stability to outliers/errors, better generalization (handling overfitting) and handling of the collinearity problem, at the cost of a small reduction in fit to data. The specific approach we are using is called elastic-net regularization of logistic regression.
We have a powerful proprietary tagging system that allows us to follow the user path to conversion across multiple channels, and we employ sophisticated techniques to identify users who have switched devices, deleted cookies, or otherwise broken the path that led to their conversion across different devices. We call this feature “cross-device matching”. Convertro is able to track a wide number of events attributable to the same user: clicks, display, social media shares across all digital channels, as well as in-app events, in-store transactions, phone and direct mail, and even TV and radio. For example, we can tell you that a user watched your TV ad that ran during Spongebob, and a week after that googled your brand from a smartphone, and in 3 days made an online purchase, as shown in the path below:
The only way too look at match rates scientifically, in our experience, is look at accuracy rates as well, which is dependent on the technique used.
Convertro has developed its own proprietary cross-device solution that leverages our pool of unique Convertro user IDs observed online across 160+ brands over the past 4.5 years. Specifically, in particular for brands that rely on user logins, we are able to identify different cookies and devices that belong to the same user with 97% matching accuracy. This means we can match almost 100% of devices for some brands. Other brands that by nature do not have great ways to identify customers (e.g. CPG, store-heavy retail) need to settle for 30% or even less.
Before developing our own solution, we evaluated several 3rd party solutions, but none of them yielded the level of accuracy of our technology. Many vendors, for example, use a statistical method to match devices whose accuracy is too low to be useful for user-level attribution purposes. Considering that we track more than 5.4 billions unique device IDs, if we used statistical inference we could probably match 1 billion devices, but with an accuracy rate of less than 50% – you would be better off tossing a coin! Instead, using our tracking-based, binary ID-based reassociation method, we match a lower number of devices overall, but the accuracy is almost perfect and the number is still significant.
Yes! Convertro’s scenario planner (a feature directly accessible in the dashboard) helps you understand how changing your marketing buys will impact your conversions, revenue, and ROI. While Convertro’s spend recommendation tool can provide information about your optimal marketing mix, sometimes you might need to forecast what would happen if you were to use a specific marketing mix, and compare these results with your current mix and with what the spend recommendation tool suggests. This is where the scenario planner comes in. It allows you to input your proposed spend for specific marketing channels and see what the predicted impact of changing spend in this way would be on revenue, conversions, and ROI.
The scenario planner uses elasticities to compute your optimized and expected revenue, conversions, and ROI. Unlike a linear model, which assumes that every additional dollar of marketing spend produces a proportional increase in revenue, elasticities capture the saturation effect you see as return from increased investment on a particular marketing channel begins to diminish. Elasticities are calculated using historical spend and conversion data.
Our planning tool also takes into consideration business logic: it won’t recommend increasing spend on channels where investment is by definition capped (e.g. saturation on PPC branded keywords) and enables the user to limit the recommended spend on specific channels to incorporate more personalized business logic (e.g. “I can’t spend more on display, don’t recommend that to me”).
Watch the video to see it in action:
Short answer: it’s the science of determining what media are driving purchases.
Long answer (from our white paper, “The Definitive Guide to Marketing Attribution“)
Before someone purchases a product or service, they are exposed to numerous marketing “touchpoints". These touchpoints cover a wide range of interactions, from seeing a television commercial to conducting online price comparisons on a comparison shopping engine (CSE) site. Attribution is the science of assigning credit or allocating dollars from a sale to the marketing touchpoints that a customer was exposed to prior to their purchase. When a paid product or service isn’t directly involved, so called “conversion events” such as a signup or registration for a website can be used instead of a sale, and credit for such a conversion can be assigned to marketing touchpoints in the same way.
In proper attribution modeling, that credit is assigned proportionally to each touchpoint according to its influence on the customer’s purchase or conversion decision. The goal of attribution is to determine which touchpoints are producing a positive result, and, by using the cost of each touchpoint, an advanced attribution system can then show which touchpoints are profitable. This allows unprecedented optimization of marketing expenditures.
Early attribution models fell well short of the goal of understanding each touchpoint. For example, the “Last Click” or “Last Touch” model assigns all credit to the last touchpoint and ignored all earlier activity. The First Click model does the opposite, giving all credit to the first touchpoint and ignoring all others. Despite the extreme inaccuracies in these models, they are still used today by many advertisers to determine the value of touchpoints. For example, many affiliate publishers are still paid by advertisers today based on a Last Click model.
In advanced attribution modeling, relationships between various touchpoints are well understood and modeled accordingly. For example, customers that search for your product online after seeing an ad on broadcast TV will behave much differently than customers that come upon your product after doing blind internet searches for a type of product that they know they need. Understanding these interactions across marketing channels (broadcast TV to search) and across devices (tv to a smartphone, tablet, or PC) is the key to any attribution system. Failure to account for these types of interactions will make any attribution system inaccurate and flawed.
It is only recently, with technological advancements that allow user behavior to be followed anonymously and in a truly privacy compliant way, that advanced attribution systems have become possible. For the first time, we are able to observe all touchpoints leading to a conversion, and make highly accurate predictions of what particular marketing expendi-tures and interactions will produce what results. We know this is true based on the results obtained by advertisers using these advanced, and scientific, attribution systems. This has enabled a revolution in marketing measurement that many companies are already taking advantage of to dramatically improve their results.
At Convertro, we use our own algorithmic attribution model, informed by accurate data that we collect first-hand, to power our marketing optimization software.
Of course we work with agencies! We have agencies as customers, such as 3Q digital, IMI, Camelot and Dumont Project to name a few. These agencies use Convertro to both differentiate their own offering as well as improve the performance of the advertising for their clients. We have advertisers who have us work with their agencies to execute our recommendations such as Starcom Mediavest, Omnicom and Havas. We are brought into advertisers by agencies like Zenith, Vivaki and Mediacom. We understand the challenges of agencies and have crafted solutions to address their specific needs including specialized pricing, co-branded dashboards and sales collateral. We also have dedicated support folks who specialize in agency relationships. If you are an agency and are not working with Convertro, request a demo so we can set up a time to share our product with you and help you close and keep business.