Balancing Short-Term Objectives and Long-Term Vision in Computer Vision-based Product Development

For those of us without the network and fundraising capability of Sam Altman the journey of R&D in Computer Vision and Edge AI is almost always a set of compromises. The quest for building the perfect set of CV Algos to augment a human’s daily tasks and automating repetitive tasks so that humans spend more time on decision-making is met with the conundrum of balancing the short-term objectives of keeping investors happy by creating MVPs for the so-called beta customers rather than building the originally envisioned solution.

This uncertainty creates unhappy engineers and frustrated sales as engineers are not happy with the MVPs constantly being released as well as the time wasted managing release cycles and versioning rather than creating the intended end-product and sales are frustrated as they are unable to satisfy the lofty end goals promised in the marketing materials leading to frustrated customers and ultimately customer churn which obviously frustrates the investors.

So how to manage this intricate dance of immediate deliverables that keeps our investors happy and allows us to move from one round of funding to another without creating a sustainable business against the backdrop of having enough time to build a product with High NPS and sets the firm on the path to an IPO?

That’s what I attempt to answer here based on my limited experience.

The Short-Term Goals Dilemma in AI & CV

Why is this an issue in Edge AI and CV? Simple; R&D and path-breaking discoveries take time.

MVPs can validate if the problem exists but most MVPs are not good enough for real-life use esp. in Computer Vision. My rule of thumb is Computer Vision solutions can be used to replace manual work only if the accuracy and speed of data capture are 2-3 times higher than that of a human being doing it manually and that the data captured is believable i.e there is proof that this data captured is actually accurate at a cost 30-50% lower than that of a human.

This cannot be satisfied in v1,v2, or even version 4 of an MVP. It needs to be a proper product

Currently, research in Computer Vision is of two types; creating new use cases through research and democratization of access for the current already created solutions.

Most B2B solutions are currently of second nature, taking the advancements of CV in the last decade and majorly used by the enterprise customers making it more accessible to Mid-market and SMB customers in the same space.

Thus, this emphasis on short-term goals, esp. from a managerial and management perspective casts a shadow over the building of a customer usable product.

But the issue remains, How do we release the product into the market without money running out?

The Delicate Dance of Management

So depending on the stage where the firm is the delicate dance to be played by he management will be different. Here are a few scenarios based on my experience:

Scenario 1: New Startup; 1st product; Experienced Founders ( Both Tech and Industry)

The assumption here is that One of the co-founders is well versed in the Industry or rather the use case for which the product is being built and there’s a co-founder who handles the tech side of things. In this scenario, the emphasis should be mainly on problem statement validation and finding the first 5 customers who are willing to co-develop with you as the solution matures.

Every Industry has a specific set of unassuming business owners who are willing to help others grow. For example, I have noticed most business owners who have a lifestyle business with good profits are always willing to lend a helping hand both in terms of data collection and setting up the device at their place along with paying some set amount for the solution.

The above will not only keep the investors happy but also will help your engineering team collect data and build the solution faster. The only downside is this methodology is primarily founder-led and does not lend to the creation of a sales engine.

Scenario 2: New Startup; 1st product; Experienced Tech Founders

90% of the time if the Founders have tech expertise then the natural tendency is to build a platform and provide the Algos as a service. Whilst, this is a good idea, there’s a reason why PaaS failed to take off whilst SaaS ruled the roost over the last two decades.

Customers want solutions that just work; Also, PaaS means to get the first 3-5 customers we will focus a lot on customization as that’s the prime value proposition of PaaS, and somewhere along the line lose site of the product roadmap. This means that over time, the product will not scale well and the ability to create copycat customers will decrease. But you will get your first few customers quickly, which can keep investors happy

But, that being said, this platform approach is a good idea esp. if you have a clear plan that over time you will drop the platform idea once you find that one use case that can be sold consistently in a repeatable fashion and in a given time frame based on the runway you have ( Based on my experience it takes 2-4 years to find the right use-case using this approach)

Also, here, we should be ruthless in cutting down the functionality of the PaaS solution as time goes on to make it more use-case specific otherwise not only do we lose site of the product but many times a use-case-specific product will replace you.

Scenario 3: New Startup; 2nd product

So you have a Successful 1st product in the marketplace. Mostly, in CV the easiest way to keep Everyone Inc. happy is to take a very specific use case with $100M-250M TAM create the end-product in 1-3 years, and gain 20-30% market share in the next 2-5 years.

Most importantly, the use-case chosen should be closest to revenue and should connect with your overall product plan i.e once the 1st product is deployed the second product should be an enhancement of the first by a magnitude of 2x.

This allows you to not only get your name out there quicker in the chosen Industry but also keep investors happy by showing progress on the topline and revenue metrics whilst also making sure the ultimate aim of building the Visionary product which has a multi-billion dollar TAM is still intact.

Some notes on choosing the use case:

• Ideally, it’s easier to penetrate the market if the first use-case chosen is closest to revenue i.e. the data captured is used to bill our client’s end customer. This allows for quick and easy adoption compared to operational workflow monitoring or data cleaning or better analytics kind of the solution
• Make sure, the first use case chosen has a quick installation time in hours rather than in weeks. This allows for quick adoption.
• The second use-case should most likey be a continuation of the first i.e if you are capturing data with the first the second product should not only capture more data in the same workflow but also provide recommendations based on data captured with the first and second product.
• This means that the third product can be a workflow monitoring solution that focuses on whether the recommendations made are actually being followed

The above process not only keeps the engineers engaged and customers happy it also works to the management’s advantage as they will have something new and substantial to report either on revenue or on product each quarter.

The real challenge here is that this is like assembling a complex puzzle, each piece or short-term goal forms the entire picture over time. Without a deep understanding of the industry and it’s regulations and compliance along with it’s trends it will be near impossible to build a multiple-product roadmap stretching out to years.

The Dual Challenge of CV Research: Problems and Accessible Solutions

In CV, it’s not just about finding solutions to known problems but also it’s about finding meaningful problems whilst also reducing the cost needed to access the solutions. This process is not linear; it’s an iterative loop where understanding deepens, perspectives shift and new questions emerge from each answer found.

Managing the above, whilst still managing the investors is what I believe is the new skill entrepreneurs have to learn and understand as they move from building SaaS-based solutions to research-based AI/CV-based solutions.