Know which channels and campaigns actually produce customers and allocate budgets more efficiently.
At a glance
Marketing attribution answers a simple question: what did a person do that made them a customer and how much did each marketing touchpoint contribute to that decision. The goal of every attribution programme is to get as close to that truth as the data allows, then use it to pick where the next pound of spend goes. Done well it cuts acquisition costs, redirects budget to channels that produce more customers and builds confidence in how marketing budgets are spent. Done badly, it can cause overconfidence and flatter underperforming campaigns.
Most mature programmes triangulate using multiple approaches to gain an accurate picture of performance. Multi-touch attribution gives the most granular and timely view. Media Mix Modelling sits at the other end of the spectrum, providing a broad view of incrementality. Incrementality testing provides a ground truth, validating and calibrating the other models.
Triangulation
MTA is granular. MMM identifies incrementality. Experiments validate the model.
Assigns proportional credit to every tracked touchpoint. Provides a timely and granular view of performance.
A regression model predicts how spend influences conversions. Focussed on top-down impact estimates rather than bottom-up tracking.
Actual experiments which are used to validate and calibrate MTA and MMM models.
Common pitfalls
Multiple ad platforms claim the same conversion. Made worse by broad reach campaigns and view-through windows.
Models get more complicated over time. Bugs are introduced. Trust is lost.
iOS attribution is complex and many marketing teams spend without understanding the nuances.
The common failure is placing too much trust in ad platform data, leading to double counting. When Meta, Google and TikTok each claim the same conversion, reported ROAS can end up two or three times the real revenue generated. Teams rebuild the media plan around those numbers and keep spending without seeing the results they expect.
The second is creating a black box that no one can understand. Many teams start with a simple query and then progressively build on more and more special cases, ending up with something that no one understands and is near impossible to debug. Metron Growth focus on architecture up front to make sure that attribution models can be extended without becoming unexplainable.
The third is underestimating the complexity of mobile attribution on iOS. IOS 14 completely broke the mobile attribution landscape and the major ad platforms have all developed their own solutions to mitigate this. Many companies are relying too much on an MMP or ad platform data without really understanding how it works. The Metron Growth team specialise in demystifying mobile attribution, helping your team understand the underlying data flows and develop a plan to create trusted attribution.
The strategic question
The question is not which tool to buy. It is which decisions your attribution needs to support, what level of complexity is proportionate to your data volumes / goals and how the output gets calibrated. Answer those three and the model, tooling and reporting architecture follow.
The teams that do attribution best focus relentlessly on identifying the questions it needs to answer, understand the nuances of their tracking and build on strong data foundations.
Garrett Scott
Head of Growth Marketing
“Cannot sing their praises enough. They brilliantly partnered with the team to execute a not at all easy project with almost no disruption, and a flawless rollout…can’t wait to see this continue to change the trajectory of Calendly.”
Multi-touch attribution assigns proportional credit for a conversion to every tracked touchpoint. MTA models rely on strong deterministic tracking and identity resolution to tie together all the different touch points a customer has with your brand. Once you have strong data foundations in place, it's easy to change between different models, whether that's moving from last touch to linear or incorporating more complex models like Markov or Shapely.
Deterministic attribution counts every measurable touchpoint from the raw event data, so the output is auditable. Probabilistic attribution uses statistics or machine learning to estimate credit. Probabilistic attribution is most often required on iOS where it's not possible to make a deterministic join between the ad's impression and the conversion. We prefer to start with deterministic approaches wherever available and then use probabilistic approaches to fill in data gaps.
Media mix modelling uses regression on aggregated spend and conversion data to estimate the incremental contribution of each channel to revenue. It can work well where you have offline channels which are hard to measure using deterministic attribution (e.g. television or out of home) or where you already have strong multi-touch attribution and want to improve your understanding of incrementality.
Apple's changes, from App Tracking Transparency through iOS 17 Privacy Manifests, removed most device-level tracking on iOS. Mobile attribution now relies on bringing together a wide range of different data sources: deterministic data from an MMP where IDFA matches are possible, probabilistic matching for other channels and SKAdNetwork data to fill in the gaps. At Metron Growth we build bespoke mobile attribution strategies for large mobile advertisers like Deliveroo, Picnic and Turo.
At Metron Growth, we create a single MTA or MMM model across all of your channels, which applies consistent logic to each channel and deduplicates conversions. This avoids the double-counting problem and gives you a source of truth that isn't controlled by the ad platforms themselves.
Incrementality testing can be used to calibrate your attribution for incrementality. We run geo tests with synthetic holdouts, branded-search pauses and platform conversion-lift studies to measure actual lift, then apply multipliers to attribution outputs to calibrate for incrementality. This can help you build real confidence in how you're spending your marketing budget and the quality of your attribution.
We review your entire measurement stack: event tracking, UTMs, taxonomy, identity resolution, sessionisation, channel classification, consent flows and MMP configuration. We then combine this with a clear understanding of your goals to make recommendations on how you can maximise the accuracy of your measurement.
Marketing attribution is never really done, rather it's an ongoing effort to continuously improve the accuracy and precision of your measurement. At Metron Growth, we focus on delivering a project which gets you to the next level whilst helping you develop in-house expertise to keep improving beyond the end of our project. Our projects typically take between 6 weeks and 6 months, depending on the complexity of your need.
Get started
30 minutes. No pitch deck.
Prefer email? contact@metrongrowth.com
