Pharma forecasters have an important job to do. They have to create forecasts. And they need to be good ones.
But often the briefing information is something like:
“We need a 10 year forecast for our Brand ‘Brill’ which will be launched in 2 years’ time’. We need it by the end of next week. Your time starts now.’
(OK, I made the last bit up.)
The temptation is to ‘get stuck into the data’ and start building the model. The indication is chronic – so let’s agree it is going to be a patient-based prevalence model.
It’s so tempting…because there is so much data out there and you only have a short time to deliver.
However- the key to success is making sure that what you deliver is what the stakeholder needs to make their investment decision. You have to get inside the stakeholder’s mind – give them some context, insights – matching Forecast Insights’ Mantra, which is…
‘Forecasting – it’s more than just the numbers – much more’
So…hold back. You need to know more before even thinking of opening the spreadsheet.
In essence we need to discuss and agree with the stakeholders’ answers to the following four questions:
- What investment decision does this forecast support??
Forecasts are a vital aid to investment decision making. Whether it’s the weather forecast convincing you to take an umbrella or a Pharma forecast helping you decide where to back the M$700 in a Phase Three trial – there will be an investment decision behind a forecast. We need to ensure the outputs of the model support the decision being made. If, for example, the forecast is to help decide the capacity required for a new manufacturing facility – the forecast would need to deliver in, say, units of formulation or kg of bulk drug.
And there’s a little voice…’OK now for the spreadsheet…’
No…not yet. We need to know the following…
- How complicated do I need to make it?
Forecasts need to be fit for purpose.In the early phases of drug development, the forecast will need to give an indication of what the opportunity is and an indication of what the peak market share might be and the speed in which this share will be reached. A very precise answer will not reflect the uncertainly surrounding what the future will look like. Embrace uncertainty and realise the limitations it imposes on precision.
‘Now for Excel – right?’ Well…nearly….
- Apart from the numerical outputs what else do you want to see?
The assumptions driving the forecast. They need to be agreed with all the forecast users. One of the most heart-breaking messages a forecaster can hear when presenting a forecast is, ‘I never agreed to that assumption’. And it’s usually someone senior.
Agree the straightforward things – the granularity of the forecast and the forecast horizon, the set of forecasts needed, units of measure etc. Then move on to the more inherent assumptions: will be first into market; will the current TPP will be the basis of our conjoint; will our main competitor launch a year after us with a worse profile?
The structure logic of the model. Describe the model structure step by step.How is the total market to be assessed? How will we segment the patient population? How will our peak share be determined, and how quickly will that share be achieved?
‘Now! – aw…come on…’
- The last task is to produce a Forecast Definition Document for Brand ‘Brill’ (2021-2031).
The above structure and set of examples is not exhaustive but if we get answers to the three questions above, again it is tempting to start to build the model. However, I believe creating and delivering a forecast model is a project, not an exercise in the world of Excel and as such it needs a project definition and a sign off.
It needs to summarise the feedback from the stakeholders as described in steps one through three and run into no more than two pages. It represents ‘the deal’ between the forecast creator and the decision maker.
Once this is agreed – the data collection and model structure can nowbegin.
And with a much greater chance of success.
Alec Finney, MD – Forecast Insight