Measuring Long Journeys - Time Profile Model

This approach to measuring effectiveness of long and complex customer journeys is to combine digital tracking and marketing mix modelling. We start with a simple model, then gradually refine with more nuance, until it provides accurate predictions, suitable for planning and budgeting.

Initial Model

The following analysis is done for each customer segment:

1. We start with the journeys we can track

A subset of customers leave enough digital traces for us to see how long their customer journey takes. From these journeys we measure:

  • how many convert after 1 month, 2 months, etc. (or weeks for shorter journeys).
  • how this differs by customer segment.
  • excluding already-decided customers, such as brand-search visitors.
This gives us a response curve — the “shape” of how interest turns into conversions over time.

2. We use this duration pattern to model how campaigns generate results over time

A campaign does not deliver all its results immediately — they unfold over time according to the duration pattern measured in step 1.

We know when each campaign ran and how large its activity was, and we can group campaigns into categories (for example, by channel, message, or format). Within each category, campaigns behave in broadly similar ways, and the impact is proportional to their spend.

By combining the expected time-distributed contributions from all campaigns, we can estimate the underlying strength of each campaign category and, from that, the return each type of activity generates.

3. What we learn from the model

The initial model gives us a clear view of how campaigns convert over time and how different activity types contribute to short- and long-term results.

It moves us from last-click thinking to full-journey, time-aware attribution, using real behaviour, and without relying on tracking each individual user.

The results can be used to calculate ROMI, CAC, averaged over the model time period.

Validating and Enhancing the Model

It is natural to question the assumptions behind the initial model, particularly in businesses with long and complex customer journeys. Key assumptions are:

  • the trackable journeys we use are broadly representative of all journeys
  • the customer segmentation applied in the analysis is meaningful
  • the way campaigns are grouped into categories reflects real behavioural differences
  • campaign impact grows proportionally with spend
  • the effects of different campaigns can simply be added together, without amplifying or interfering with one another
These assumptions are reasonable starting points, but each can be tested, validated, or refined within the same modelling framework to improve accuracy and confidence.

Assumptions validation

Depending on the business and the data available, the assumptions behind the initial model can be validated in several ways:

  • Test campaigns can target an isolated audience, such as a specific geographic market, where behaviour can be tracked end-to-end.
  • Business knowledge can provide objective ways of defining customer segments or grouping campaigns.
  • Alternative model options can be tested to see which configuration produces higher-quality results.
  • Media knowledge and academic studies can guide more accurate methods of modelling campaign impact. For example, reach and engagement are often stronger proxies than spend and can be applied to owned and earned channels as well. Likewise, instead of assuming proportionality to spend, the model can incorporate diminishing returns.

A good model should reflect the real dynamic across the marketing funnel — delayed decision making, the limited audience available at each funnel stage and through each channel, user interactions across multiple channels, the decay of impact after campaigns end, and the contribution of overall brand value (awareness and reputation).

Model quality is assessed using two main criteria: how well it explains current results and how accurately it would have predicted past outcomes. The model is refined through multiple iterations and evolves over time as business conditions change.

Results and Insights

Once the model reaches satisfactory quality, it can be used not only for calculating ROMI and CAC, but also to forecast customer volumes under proposed marketing plans and to optimise plans to achieve desired outcomes. It can highlight which activities drive long-term growth, which channels are approaching diminishing returns, and how shifts in spend influence results over time.

In combination, these insights support more confident planning, budgeting, and optimisation for long and complex customer journeys.


© sys2biz Pty Limited, 2020 | Privacy Policy | LinkedIn | Facebook