Case Study of Workflow Optimization With Million-Dollar Effect Nobody Expected

It was my first major project in my career as a developer. Joining the team, I found a well-established real estate platform that had thrived for over a decade, with a massive user base and unrivaled dominance in a major US city. My primary assignment revolved around optimizing the premium listings placement workflow.

The platform relied on Google Ads contextual advertising service for publishing premium listings. This required duplicating a significant number of listings in two separate databases - the platform itself and the Google Ads system. Moreover, since Google Ads incurs costs, it led to a decline in the platform owner's revenue. Consequently, we were tasked with developing and integrating an in-house mechanism that would offer similar functionalities.

That situation left me deeply puzzled at the time. Why would one choose to publish their own listings on their own website using an external service, especially a paid one? It appeared illogical and inefficient. What's more, this mechanism persisted unchanged from the platform's very beginning! However, over time, the once-mysterious circumstances started to make sense to me.

This article aims to share these findings with you. Moreover, this project presents an excellent opportunity to explore the question of when the need for business processes optimization arises, how it unfolds in practice, and the often overlooked factors that influence the outcome.

Projects like this indeed involve hidden factors, not only for developers who are just starting out. Specifically, the client of this project, an experienced entrepreneur, was genuinely amazed by the results achieved. In the very first month after adopting the new advertising placement mechanism, the platform generated profits of over a million dollars, far exceeding the highest expectations.

Let me offer an explanation of the factors that, from my standpoint, contribute significantly to the success of this optimization project.

Success Factor #1 - Validated Concept

So, the developers have the task of building the in-house advertising placement mechanism. What do they need for this? First and foremost, they need to know what it should be like. Remarkably, we encountered no challenges in this regard.

The client clearly described the system that the required mechanism should support.

• Real estate sellers are offered multiple service packages with different price tiers.
• The most economical package involves the listing being ranked in search results according to its relevance to the buyer's search query.
• Additional charges enable listings to secure prominent positions in search results, feature distinctive design elements, appear alongside search results, and be highlighted in email newsletters notifying buyers about relevant new listings.

With this clear information in hand, the developers can efficiently plan the necessary improvements and accurately anticipate the timeline. However, this is not always the case. This system has already been meticulously perfected and proven its efficiency through years of practical implementation.

To provide a contrast, let's envision the same project during its early stages as a startup.

• Was it possible for the client to foresee the precise conditions that would appeal to real estate sellers?
• Or the specific techniques that would successfully promote their listings?
• Or the means to ensure profitability in this business?

The answer is no. These questions can only be answered through trial and error.

In cases where an entrepreneur lacks confidence in their concept, developers will implement improvements in small increments. Following each step, it will be necessary to test the application in real-world scenarios and be prepared for the possibility that some steps may be flawed.

That explains the paradoxical fact that despite having a dedicated team of custom development specialists, the client of this startup project opted to utilize the ready-made Google Ads mechanism. By doing so, they gained immediate access to a wide array of convenient experimentation tools for a considerably lower cost compared to developing such tools from scratch.

Success Factor #2 - Technical Feasibility

So, the developers have a clear vision of their goals, confidence in the concept, and even a preliminary plan. Can they proceed? Not quite yet. First, they need to ensure that the costs associated with implementing the improvements are not so high that this "optimization" becomes meaningless.

Our task was to develop a mechanism that could rival the performance of Google's algorithms, known for their intricacy.

1. Firstly, there is the contextual targeting algorithm, which involves displaying only relevant listings based on the homebuyer's search query.

• In the Google system, this algorithm utilizes advanced user intent analysis that proves highly useful for ad placement on platforms like Google's search page. For example, it empowers Google to identify the ads that would capture the interest of a user conducting a search query like "weekend with kids".
• However, our project did not require such complex analysis. Platform users utilize simple criteria, such as the number of rooms, living area, bathrooms, price range, and so on, when conducting searches. Hence, just filtering premium listings based on these criteria sufficed for our purposes.

2. Creating equivalent analytical reports to Google Ads, which provide sellers with data on listing impressions and clicks, also posed no expected challenges.

3. The most demanding task was to design an algorithm that could effectively choose between relevant listings of varying costs to present in each individual case. We anticipated the need for conducting a series of experiments, and that is exactly what happened.

• The initial concept of prioritizing expensive listings proved unsuccessful. It resulted in a scenario where, for instance, during the first 30 minutes of each hour, users were only presented with expensive listings, while cheaper listings were exclusively shown during the remaining 30 minutes. This cannot be considered an effective promotion strategy.
• Ultimately, we concluded that implementing an hourly impression limit, with a higher limit for expensive ads, would be the solution. This way, the algorithm can choose any ad for display as long as its limit for the current hour has not been reached.
• After implementing this algorithm, we needed some time to monitor the system's behavior, adjust hourly limits based on user query volume, and ensure that all premium listings received the agreed-upon number of impressions.

However, even accounting for the extra time needed for experiments and monitoring, our team was confident that developing this mechanism would unquestionably be more affordable for the client compared to paying for the Google Ads service.

So, How Does One Know if They Need Business Process Optimization?

As previously mentioned, the outcome of our work has surpassed all expectations, leading to an extraordinary surge in the platform's profitability. It is not unreasonable to assume that if such optimization had been implemented earlier, the client could have been enjoying such profits for many years. Hence, the question arises: why didn't it occur?

Isn't it obvious that when the current system becomes ineffective and starts causing problems, optimization becomes necessary? However, in reality, it's not that simple. In an imperfect world, every system brings its own set of problems, and accurately predicting the level of usefulness of a specific improvement is also a challenging task.

This project has uncovered an important realization: the client, without initially setting out to explore this matter, was unaware of the extent of their expenses in maintaining the current system and the ease with which they could replace the "complex Google Ads algorithms" with an in-house mechanism.

The key to achieving timely and successful optimization lies in systematically and deliberately examining such matters.

• Is the system adequately meeting its objectives?
• Are the current costs justified?
• What alternatives are available?
• What could be the costs and benefits of transitioning to an alternative system?

The answers to these questions won't come intuitively; they require purposeful collection and analysis of the relevant information. It's a distinct aspect of maintaining any application, demanding continuous attention.

The significance of this aspect becomes evident when we consider this project as an example. As we can see, it has the potential to yield or miss out on six-figure profits for many years.