Old model: vertical and sequential, the assumption dictates the process.
The old CDD process looked something like this: I was in charge of the expert interview workstream for a project managed by a big 4. We had created a template so that our answers could feed directly into their deliverable. Halfway through the process, the Senior Manager asked to not change any columns, the analytics was already set up, and their slides would not work if we touched the formatting. The basic premise behind that ask is that the speed demanded by a CDD could only be achieved through massive teams, each with ownership of a workstream. That meant that once the process was established, the conclusion had already been reached and the deliverable guided the analysis. Recommendations were almost foretold.
The financial model analogy: all data sources simultaneously
The old models had a two-tiered structure. There was a dynamic data-driven core (usually based on the market sizing & financial models) that was live and linked to the PowerPoint. There was then a large static shell built around it including the competitive positioning, the customer findings, the qualitative synthesis, which was manually assembled and frozen. At best, the static shell was nudged to align with whatever the numbers now said. The transformation is that the static shell can be dynamic. The new way of operating is driven by the data you have available in a specific market but once you have the data, you can build the story dynamically rather than sequentially. The story and the data move together. Although we are not quite yet in this fully “all fronts at the same time” deliverable, the methodology below is built around that architecture.
Structuring a CDD in the age of AI
Each CDD process is bespoke, which is why purely platform-based CDD models are structurally a bad idea. For what follows I will show what we have developed for a fictional (yet based on experience) process including a market sizing, competitor analysis, and customer analysis, and on top of this expert interviews. Other processes could include a more precise churn analysis, pricing analysis, evaluation of the leadership team, PESTLE analysis, etc.
Step 1: Answering the real brief & remaining compliant
A CDD process is like a matryoshka doll of nested questions. You first have compliance, banks require a due diligence document and it needs to be done. That first step is the least commercially relevant, but is the one that justifies the expense. Then comes the hypothesis testing. This asks if the investment thesis holds, if the market is the right size, and if the growth potential is real. Traditional CDDs always cover these points with the limitations stated above, but to usually a good standard. There was however always a last question that drove the investment decision: the “real brief”. That one varies by deal and can be very soft (something that is required by a senior investor or the investment committee), or something that is more political (there are divergent points of view amongst partners and you need to find a possible landing zone). That last question, the “real brief”, is the one that can potentially cause a deal to fall through. Under the previous cost hierarchy, driven by large teams required by speed, that real brief was buried in the deliverable and never really tested and approached. Addressing it properly would have meant rebuilding workstreams that were already frozen. In the new model the cost hierarchy no longer maps to the question prioritisation.
The innovation in CDD starts at the point of briefing. You don’t have to limit the conversation to the amount of analyst time required to run a task
The innovation in CDD starts at the point of briefing. You don’t have to limit the conversation to the amount of analyst time required to run a task, you can attack each of these layers with the same budget because the deliverable is dynamic. Answering the real brief doesn’t require rebuilding from scratch, it requires rerunning with the right parameters and testing it in a scenario analysis. A project is now driven by the type of data that is available, not by the number of questions asked.
Step 2: Starting from the data, not the conclusion
In the old model, the questions determined the data you looked for. In the new model, the data determines the questions you can answer. In very closed markets, you still rely on traditional methodologies powered by human intelligence sources, informed reasoning and triangulation (e.g. you are studying a specific product, and make informed assumptions on margins, cost and volume to obtain market size and competitive positioning). But those cases where you have very little are not as common and are more the purview of frontier and niche markets.
The space where the new approaches are pushing the frontier are markets with some data or vast amounts of data.
The space where the new approaches are pushing the frontier are markets with some data or vast amounts of data. Previously a bit of data was the same as no data at all, with the current methodologies we can build disaggregated data into solid models. For instance, our team was recently able to collate the number of companies operating in a very niche sector by aggregating data from NOMIS in the UK. Then we were able to extrapolate spending for a product by company size from internal customer data and triangulate cost by size of company from primary (interviews) and secondary sources (online pricing). This data source gave us a multi-year assessment and we were able to build ETS estimation to assess market sizing in the future. Before 2025, building this from scratch would have been reserved for the largest and most expensive mandates. In the same way, projects with a lot of data can now be rethought and data can be presented in platform format and local dashboards based on HTML outputs to visualise multiple scenarios and work with the client to choose the best approach.
Step 3: Leveraging customer data at scale
In this new reality, you can also do primary research at scale in a way that was unaffordable for mid-market deals. You can now conduct expert interviews but also customer interviews with automated voice interfaces that can conduct NPS and Key Purchasing Criteria (KPC) analysis but also obtain better data than the one captured by questionnaires. Once you have the customer data you can link it to the type of customers that provide that opinion (not only relying on averages, but on other company characteristics like size or tenure) and understand which type of customers are more at risk from churning, or buying more. You can then mimic their answers by linking what they say to their characteristics, a technique known as synthetic data modelling. This can then create a better understanding of which types of clients are more likely to grow organically and where most of the risk is. With the right level of data sophistication, this can be fed into a dynamic market sizing analysis. All of this can be included into CDD processes with relatively little added cost when compared to what it would have cost a few years ago.
Conclusion:
We are moving from a siloed model where workstreams are broken into pyramid-shaped teams, to a model where more questions can be answered, where the analysis depends on the type of data available, not just financial models and customer surveys, and where multi-year extrapolations can be achieved using synthetic data and remote customer interviews. This will cover compliance and the real brief. The CDD will become something different that is still being invented with clients but that will allow for many more data sources than the ones used up to now. Although the tools allow for more precision now, the change is just a movement towards more good consulting. I was working with a large corporation who was looking to acquire “a billion dollar firm”. I managed a team covering several countries and we started coming up with a list of potential acquisitions ordered by a preference matrix. After a few iterations with the client it became clear that the objective was the inorganic acquisition of enterprise clients in a specific sector. Therefore, we were able to tweak the categories to help them build a client acquisition story. Today, with the same data lake, I could do both and tell them here is your billion dollar strategy, this is your inorganic client acquisition strategy. All essentially with a similar budget and potentially a much smaller team.
