The vast majority of data relating to an organization’s products resides within their Engineering data (Product Specifications), Financial data (Margins), and Quality data (Customer Complaints). Unfortunately, this data is generally maintained in separate systems from the Commercial performance metrics, which are typically stored in reporting systems used only by sales. Such data is of little value to anyone outside of sales and therefore never opened. The challenge is to pull this data together in a timely and consistent manner to aid the leadership teams of an organization to make different decisions.
Visibility starts with a defensible product record
Getting portfolio visibility on top of that as a dashboard is typically a definition problem in the first place. Until you agree on what a product is (including its variants), on the needed attributes for each product and how to track them, views on your product portfolio will not add much value.
Before even going into building out different views, you need to first agree upon the definition of what a ‘product’ is (different variants, attributes etc). A workable minimum set of attributes for a record of a product would be, lifecycle stage, business unit, historic revenue & historic gross margin (over a trailing time period), historic development spend, relevant & current regulatory/certification status, and the strategic objective that the product is trying to meet.
Test the record against a real question
A simple test question, If the executive team had recently asked a question such as “Which of the products launched in the last 3 years are not reaching their margin objectives?” then how long would it take to get this information from the underlying systems? More than a day? Then there is a huge analytical gap that cannot be solved quickly enough.
Prioritization models that survive contact with the business
A scoring framework to rank products to invest in typically has too many attributes and is too complex to maintain (some attributes are not discriminative and therefore do not make a difference between the scores of very different products). Also, typically each attribute is weighted and maps to one data field which is already maintained by someone for some other purpose. Therefore, it is better to use a small number of weighted attributes (typically between 2 and 5).
- Strategic fit, scored against named corporate objectives rather than a general sense of alignment
- Commercial value, using risk-adjusted contribution rather than headline revenue
- Resource intensity, measured in constrained capacity such as regulatory, clinical, or specialist engineering hours
- Technical and delivery risk, with a documented basis for each score
- Portfolio balance, covering how much the product duplicates or cannibalizes existing lines
Make sure every score references back to evidence such as a forecast that the person filling out the scores has access to (ie owned by someone). So that a score of 4 for example on commercial value for a particular product refers to a specific forecast for example owned by someone.
Rank against capacity, not against each other
Rating products against each other for funding within constrained capacity of initiative teams delivering the highest ranked products in particular adds value and surfaces the effort being wasted on delivering low ranked products with the same teams delivering higher ranked products.
Scenario planning with portfolio data
Swatting up different numbers for ‘optimistic’, ‘pessimistic’ and ‘most likely’ scenarios does not add value. Instead, more value is gained by testing different combinations of variables.
| Scenario type | Variable flexed | Decision it informs | Data required |
| Capacity constraint | Reduction in specialist development hours | Which programs pause or move to partners | Resource allocation by project and skill |
| Rationalization | Removal of lowest-contribution SKUs | Tail pruning and complexity cost | SKU-level margin and shared cost allocation |
| Regulatory change | New compliance requirement in a key market | Remediation sequencing and market exit | Certification status by product and geography |
| Demand shift | Category decline of a defined percentage | Reinvestment and pipeline rebalancing | Segment revenue history and pipeline stage data |
Keep the assumption set visible
So many assumptions are hidden away in a spreadsheet, in the bottommost or 2nd to last sheet, in an “Assumptions” tab. Make every assumption a named input, state who provided it and on what date, then re-run your scenarios as necessary rather than having to re-work your whole analysis.
Centralizing information without centralizing control
A single source of truth does not necessarily have to reside with a single team of analysts. A so-called ‘federated’ model with different departments owning different aspects of the content, all working within a common framework and standards, can be highly effective.
Teams running that federated model are usually better served by a strategic product portfolio management platform than by another spreadsheet, since it will help to ensure that the proper fields are used to score products, who updated what when, and whether the information is current or not. But it does not solve the higher level problems around prioritizing products and services.
Making the decision process itself auditable
There is more to measuring the value of analysis in support of decision making than simply whether or not it is present. One should try to determine whether or not the analysis that was used to make a given decision was correct and even more important, whether or not that decision could have been re-run as the underlying data changes.
- Record the decision, the date, and the portfolio data version it relied on
- State the condition that would trigger a review, such as a forecast variance threshold
- Schedule the review at the next portfolio cycle rather than leaving it to memory
- Compare the realized outcome against the original case and feed the gap back into scoring weights
There is so much more to enable companies to improve their portfolio processes from one year to the following year than running through the same cycle of processes and meetings again and again. Measuring past accuracy of the models one has run already is a good start.
