New Product Development
How quantitative research can help determine feature set and pricing
Anyone running a product or service based business is constantly involved in improving the product to meet the evolving needs of the customer segment they’re serving or looking to serve. What I’ve tried to do here is to detail the different quantitative research methods that I have used in the past for New Product Development across Hardware and Software, and really explore how useful each of these are when thinking about product development across the different product types → Consumer hardware products, Consumer services and Business to Business SaaS products.
The Actors
A Product Manager, Product Marketing Manager or Brand Manager in one of the following types of businesses
- Consumer hardware products: These could range from your standard FMCG goods such as a bottle of Coke, a packet of crisps to tech centric goods such as Smart phones and Headphones.
- Consumer service products: Facebook, Instagram, Snapchat are the first examples of consumer services that come to mind.
- Business to Business SaaS products: Some examples of products that fit this definition are Hubspot, Adobe Creative cloud or Slack
What’s the Objective?
To be precise, we are looking to identify the optimal set of features for a product that will result in maximum revenue
What are the different Business models, their dynamics and what research methods can we employ?
1. Consumer hardware products: Hardware products have long product development lifecycles and it is impossible to iterate once development begins without pushing back timelines, which means features are mostly set in stone. Additional features also add to the bill of materials (BOM) and thereby impact the price. Do you want to force the customer to pay more because you added a feature that you thought was cool ? It is therefore very important to ascertain the exact set of features that the product will offer at the very beginning of the product development process.
A more rigorous research based approach is especially useful when creating a new category. Some of the questions I’ve grappled with in the past when we were working on the 1st consumer grade 360 camera (VR camera) were a) How much would someone belonging to the target segment pay for it b) What % of people can make do with only “still photo” feature and what price would they be willing to pay? c) What % absolutely need “Still” and “Video” feature and at what price point?
In a mature category, while a lot can be achieved by just studying the competition, one can still use the methods described here to gauge the impact of a new feature to the product set. For e.g. you could investigate how much someone would pay for a 4K television that comes with the best quality audio speakers from Bowers & Wilkins or Bang & Olufsen.
2. Consumer service products: For the most part these service products are free for consumers to use, and therefore some of the more advanced quantitative methods that we typically employ to assess feature vs price tradeoffs don’t apply here . But that doesn’t reduce the importance of research prior to committing resources to building a feature. Let’s take a guess at what methods they may be employing
(i) Observe how customers use their current feature set. I’m sure facebook had enough data to tell them how frequently many users were direct messaging each other when they decided to work on the Messenger app. Also, employing mobile heatmaps is another great way to understand how users interact with your product. More details on heatmaps here. Both these methods generate enough data to qualify as quantitative research.
(ii) run a standard survey of their user base to assess latent needs
(iii) run qualitative studies (usability testing) on their core users (out of scope for what we’re discussing here)
3. Business to Business SaaS products: Similar to Consumer hardware products, they have a need to ascertain the price for a set of features. A common pricing structure for SaaS products is a tiered one such as the Hubspot example below.
Or it could be in the form of a bundle, a great example of which is adobe creative cloud. Both of these method can use advanced quantitative analyses to settle on the features/product, standalone price and the bundled price.
For larger enterprise accounts, these research methods may be useless. For e.g. If you are in the business of mobility software, have a handful of large prospects worldwide, and provide a solution that is central to the operations of your enterprise customer, then you should be focused on taking to your customer to understand what features need to be built, and be ready for hard nosed negotiations to get to a value based price that makes sense for both parties.
The Methods:
Standard Survey:
This is the primary method to test out the hypothesis of what the problem/need is for a given customer segment. By a standard survey I mean a survey run using a simple tool such as survey monkey and without too much data crunching. A standard survey can easily fetch you the following sort of information
- Demographic information → Age, Income,
- Psychographic information → Motivations, Interests, Hobbies, Activities etc.
- Experience → Beginner, Enthusiast, professional etc.
- Problems that require solution/User needs
- Competitor products being used
- Price
Some simple ways to ascertain price for a new product or feature in a standard survey is
a) Ask the participant how much he/she will pay for it
b) The Gabor Granger pricing technique
You are planning on developing a pair of headphones with noise cancellation. How do you assess what the interest level will be from your target segment?
You could use the stated approach where starting from a low price, sequentially present to the participant the headphone in question at varying price points and and ask if he/she would buy at each price point.
With the output, plotting % of participant that will purchase vs Price will give you the price sensitivity curve.
These methods do work if you are the only act in town (unlikely) or have a large and loyal following a la Apple (unlikely) and are only looking at launching a single product.
Choice based Conjoint Analysis
In the real world there is competition that you need to contend with. In the open market where customers have many alternatives, your headset is never going to garner 40% share at $50 as the previous simple analysis tells you. Also, while you add a feature, you may want to downgrade some other features to maintain the price point. You may also want to measure the impact of Brand on how your product compares with the competitors for like for like features. All these use cases call for the Conjoint analysis.
What the conjoint analysis really does is to deconstruct your product into its component features a.k.a attributes and assigns a share impact per attribute for that customer segment. In other words, how important is a feature with respect to the other features for that customer segment. This will give you an understanding of where the customer is willing to compromise and where he isn’t to hit a specific price point that he/she has in mind. And that is valuable because that’s really how we all make buying decisions.
How does it work:
You will need to define a set of features that you want test and Brand can be one of them. Each feature will have either specifications (2GB, 4GB etc. ) or level (Good, Better, Best) or the Brand name (Apple, Google etc.). The different price points to be tested will also be input ($249, $299, $349, $399). The test then requires creating a number of different combinations of the features at various price points to constitute products, and making the tester indicate a preference of product from a limited set of presented options. Here is a screenshot of any generic tool used to create the test. The tool will present multiple such screens with different combinations and the tester will need to make a decision on each screen.
Here is an example of how we can set up a test to optimize the pricing for Adobe creative cloud (SaaS offering) where each product can be an attribute with levels -> In Package and Not in Package. See Fig 6 below
Share impact of Attributes:
From the preferences that the tester exhibits on all the feature combinations presented, the formula/algorithm extracts how influential each feature is in the buying decision process. See example output of the headphones test below
The above output allows me to draw some key conclusions about the conjoint analysis
(i) Helps with understanding price sensitivity. In the above example, the customer segment can be inferred to be price sensitive as price is the most influential attribute by a long way(ii) Helps with feature decisions for hardware products. From the above example, I would therefore plan on releasing a product that is competitive with existing players in terms of price, and what will help is doing away with low impact features such as “fast charging”, “Material quality”, “Call quality” and “Pause Play sensing”. I would then take the money that I just saved on my BOM and a) put it towards creating differentiation on the higher impact features or b) pass on the savings to the customer
(iii) Helps with feature decisions for B2B SaaS products. As a SaaS PM, while I don’t really care about BOM, do I want my team to spend time on developing a feature that has low share impact (customer unwilling to pay)?
Preference Share:
The other output of the DCM analysis is the “Preference share” which is defined as “the estimate of people from within the customer segment, based on the tested sample, that would purchase the product if it were available”.
The preference share is the value that is generated from the construction of a product with its specific feature sets given the Share Impact of each feature. The preference share represents the share within the comparison set of products, and therefore should not be read as the market share. But it could also represent the Market share if the comparison set contains all the market players and their products.
Let’s take an example. Let’s say you as Product Manager at Brand-3, you are tasked with assessing the potential for a new headphones product (Product C -> Audio quality: Good; Type: In ear loose, Battery life: 40 hours of listening, Noise cancellation: Multi-mode, Fast charging: Yes, Fit & Comfort: Std, Taking calls: Casual) in this category at a price of $349. You would be required to plug in the feature details of your proposed product and also any or all competitor products you’d like to benchmark against (comparison set). You would also want to try other realistic combinations at their feasible price points and pick the product with the feature set that delivers maximum value where maximum value = Pref share x Price.
Here is a sample output
What can we infer from this?
(i) As PM of Brand-3, I find that the preference share for the proposed product is 20%. Without drilling into the details as to why, we can see that Brand 1 has a 1.75 X times the preference share of my product
(ii) I can use this preference share information to forecast my sales. Let’s say that Brand-1 sells 3M units annually of Product A in N. America and has a presence in 300 stores. Our plan sees us at 30 stores in Year 1. Leaving e-tail aside for the sake of simplicity, I come up with the following number
Forecast (Product C) = (3M)*(30)/(1.75)*(300) = 171K units (Year 1)
Subsequently, consider all the realistic product offerings that you could offer at the same time to the same segment and observe the results. Fig 9 below represents a scenario where Brand 3 releases two products C & C0 where C0 offers a subset of features of C at a lower price . The net effect is that revenue has increased over the scenario where Brand 3 has only a single product in market(5%*349+20%*299> 20%*349). This example is as pertinent to a consumer hardware product as it is to a software offering (think Adobe creative cloud at $50 and just photoshop at $20)
Optimizing product portfolio and visualizing the resultant total preference shares which are time dependent is best done using a Turf analysis. This is a subject for a separate blog. In the meantime, more here.
Tools and Approach
I have been using Conjointly to run pretty much everything we’ve spoken about here. Its an online tool for running a wide variety of analyses such a Conjoint, Gabor-Granger, Van Westendrop, Max Diff to name a few.
The approach I’ve generally taken
(i) Getting a survey audience that closely matches your target segment is still a challenge, and therefore (a) Use online tools such as Conjointly to run tests on your customer pool (b) Use an external research agency to run tests on the wider customer segment
(ii) Always makes sense to run as many different types of analyses and compare and contrast the results
(iii) You should always run separate tests with your customers and the other with the wider public belonging to the target segment, and compare the results. This will allow you to estimate uptake from your customer base and from the wider public
Conclusion
Here is a summary of the standard set of quantitative research tools at our disposal and where they can be applied
In conclusion, these advanced techniques help
(i) To ascertain in a more robust fashion how much interest there is in a feature. No better way to do this than to ascertain exactly how much its worth to them ($)
(ii) Maximize revenue through a) Optimizing price for a product b) Optimizing the product portfolio
(iii) Make more accurate sales forecasts
Useful Resources
- https://conjointly.com/
- Monetizing Innovation by Madhavan Ramanujam & Georg Tacke
- https://uxcam.com/blog/top-heatmap-analysis-tools/
