2025年3月23日日曜日

(9) Pts Journey based Forecast Model

Insight based Brand Plan

In last article, we quantified pts journey like this.



When asked, "What is the relationship between quantified pts journey and the Forecast model?" I answer, "That's a stupid question, but quantified pts journey is (almost) the Forecast model itself." Since they are the same thing, it would be natural for the numbers in the Brand Plan, such as the number of patients and prescription rates, to be the same as the Forecast model, but we often see cases where the numbers in the pts journey and the forecast model do not match, and plans with inconsistent numbers are criticized in reviews.

If the brand manager creating the Pts journey is the one creating and tweaking the forecast model, the numbers rarely match, but there are cases where this is not the case (forecaster or insight manager is creating the forecast, etc.), and in those cases, if the numbers don't match, I take it as a sign that collaboration isn't going well and this is dangerous.

In my case, I often create and update the quantification of Pts journey and the creation of forecast models in parallel. What we will do is to first simply convert the slide shown above into Excel.


This is basically just a PowerPoint slide converted into Excel, so we need to make it clear where the input cells are so that we can make a forecast model (i.e. run a simulation).


I like to make the input cells yellow like this. Next, make sure to enter the Data source into the model’s EXCEL. If you don't do this, you'll have trouble later when you ask "What was the basis for this assumption?" or when the person in charge changes. In the Excel image above, the data has not yet been expanded horizontally by years and months and therefore is not yet a forecast model, so let’s first expand it horizontally by years.

It's starting to look a lot like a forecast model. Because this is a rough model, we do not use categories such as "patients who visit the hospital every year," "patients who have visited the hospital but not every year," and "patients who have no history of visiting the hospital," but rather allow us to perform a simulation in which the rate of outpatients based on severity is gradually increased. To track patient surveys every year and calculate the "outpatient visit rate by severity," we ask patients to describe their symptoms and then we can determine the severity based on those symptoms and then calculate the outpatient visit rate, etc.

If in the future the product share of product X among patients who visit the hospital every year becomes 60%, and targeting patients with hay fever who do not visit the hospital every year or patients who have never visited the hospital becomes a high priority strategy, then the model will be changed accordingly. On the other hand, the reason we are not doing this now is because for a while after launch, the group of patients who visit the hospital every year for hay fever + patients who visit the hospital but not every year = in terms of the pts journey, patients who are prescribed antihistamines are our high priority target customers. In this way, it is quite rare to change the forecast model to match the strategy, but I would encourage you to give it a try. If your brand strategy or KPIs change, you might want to change your forecast model as well. I think it's a good idea to keep this in mind. Simulations can be easily performed by linking the pts journey and Forecast models. It is also a good idea to link it to KPIs. If the number of new patients acquired each month is the KPI, then that number should be the input item (shaded in yellow). If market share is the KPI, then that should be the input item (shaded in yellow).

Let's go back to our Forecast model for Product X. The moments and opportunities we can influence that were identified in the Pts Journey, including outpatient rates, and drug prices that change year by year,
  • % of visit HCPs
  • Number of visits/hay fever season
  • product X rx%
  • Patient acceptance of product X
  • Drug prices
The annual model with this assumption is now complete. About a year and a half before launch, it's time to come up with a production plan, so we actually create a similar model by month. It would be possible to break down this yearly model into monthly units, but since actual sales and KPIs are also updated monthly, it is recommended that the Forecast model also be based on a monthly unit from the start. It's easy to tally up the data by year.

Although we have written about the requirements for a forecast model here, it can easily become a "precise but difficult to use model" that no one on the brand team can handle, that is difficult to simulate, and that is merely the self-satisfaction of the forecast craftsman. Since it is impossible for a Forecast model to boast 100% prediction accuracy, we recommend that you validate the Forecast model using actual monthly sales and use a simple Forecast model that is easy to simulate and easy to handle, while allowing for deviations of around 5% up or down.

I got question on LinkedIn, "How do you use prescription intentions obtained from quantitative survey results in the Forecast model?" Assuming that the market research questionnaire is well designed and bias is avoided as much as possible,

Market research prescription intentions= Rx intention when drs fully understand Product characteristics ,when key messages were received ≒Peak share

In this case, doctor prescription intention for product X is 40%, so we set the peak share of product X at 40% and incorporate it into the forecast model, and we do the same for the yearly model above. The reason for this is that the survey design covered all doctors who may prescribe antihistamines (i.e. 100% market coverage), and the prescription intention of doctors after understanding the product characteristics, key messages, and pros/cons presented in the survey can be interpreted as "the prescription intention of doctors who prescribe antihistamines after understanding the product characteristics, key messages, and pros/cons of Product X" even after the actual launch. However, depending on the actual market coverage, account opening, etc., if the coverage is 80%, for example, we will discount by that amount.

In the past, I have been involved in new products and indication expansions, and have looked into the prescribing intention and actual peak share based on a quantitative survey prior to their launch. When the prescribing intention in the survey was 50%, within a ±20% range, the actual peak share was in the range of 40-60% in many cases, so I think this interpretation is fine in practice.

Another thing that makes you wonder about the forecast is the uptake curve leading up to the peak share. Many factors come into play, such as product characteristics (how revolutionary it is), the competitive situation in the market, and the order in which it was released in that market, but it is common to benchmark the uptake of products in similar situations, and I agree with this. However, in reality, it is often difficult to find a similar product or market, so in such cases, the team will often agree on a basic strategy of keeping the dosage low in the first year, due to a two-week prescription limit, and then taking about three years to reach its peak. I also think it would be a waste of a huge amount of energy to put into this.

I have also asked doctors and pharmacists in quantitative surveys, "When will it be adopted? Will you start using it?", but I haven't had much success. So, although it's a difficult decision, I think it would be realistic to launch it based on benchmarks and team agreement, see the uptake at release, and then make adjustments as appropriate.

0 件のコメント:

コメントを投稿