Build vs buy software? In a survey of 600 IT, technology, and business stakeholders, 93% of agreed delivering relevant information to digital visitors is crucial to their business — and that search is the primary tool to achieve this. Yet the survey also found many encounter challenges when it comes to building software for search.
In the era of generative AI (GenAI), where searchers expect answers instead of blue links, audience frustration runs rampant. Gartner predicts GenAI will eventually become a cornerstone of all AI software spending, reaching 35% of worldwide revenues by 2027. All GenAI use cases need a reliable information retrieval solution to unify, not isolate, enterprise efforts.
Gartner forecasts that by 2025, nearly a third of GenAI projects will be scrapped due to the difficulty of scaling from proof of concept (POC) to full production. This prediction may be nearing reality, with BCG finding that 66% of leaders are ambivalent or dissatisfied with their progress on AI and GenAI.
And 62% of those who expressed dissatisfaction with their organization’s progress on AI and GenAI also cited a shortage of talent and skills. Many teams face the question of building their own custom solution or purchasing software out-of-the-box when evaluating search options for digital projects. Building a solution in-house without dedicated expertise is an uphill battle, leading to increased costs, prolonged timelines, and the need for constant ongoing maintenance.
Still, there are many options available on the market, meaning enterprises aren’t limited to building or buying. Instead, enterprises can build, buy, or opt for a hybrid approach to better serve specific business needs. This blog outlines key considerations for choosing the best path for your next search project so you can make an informed decision.
Upfront Cost vs Total Cost of Ownership
The challenge with crafting any search experience is high client expectations that are set by the powerful standards created by Google and Amazon. These experiences instill preconceived notions of what a search platform should do “out of the box.” To the surprise of many, the expected “standard” is actually quite advanced, and many a traditional existing solution struggle to compete.
Features like dynamic faceting, federation, authenticatred search, exclusion logic, and more are often seen as the baseline, although from a development cost angle they’ll quickly surpass most enterprise budgets if built from scratch.
When these features are considered in pre built software, costs become much more manageable. While open-source options can support simple keyword search, taxonomy-based faceting, and other simple features, choosing a platform like Coveo enables you to do so much more from Day 1 — recommendations, personalization, and generative capabilities, to name just a few.
What Level of Ongoing Maintenance Are You Willing to Support?
Most companies assume open-source solutions save money while providing evolving code. However, open-source systems often require extensive fine-tuning to reach their full potential.
Today’s expectations around search are much more sophisticated than even just a few years ago. Rather than just matching keywords in documents, modern search engines decipher and compare lexical relationships to establish connections between the query or prompt a searcher has entered and the requisite answer or response.
Apache Solr and Elasticsearch, as two open-source examples, are impressive pieces of technology. But that means they require attention and dedication to their maintenance that might be unrealistic for today’s enterprise experiences.
One example is of getting content into your chosen search solution. With an open-source product like Apache Solr, that task is solely on you and your builders (assuming you have a team behind you).
By partnering with a search vendor like Coveo, you have flexibility over integration and connectors, that can help bring over 10,000 sources across more than 100 content types into your index.
Relevant Reading: Limitless Connectivity? Why Quality Matters Over Quantity in Enterprise Search Connectors
Scoping Your Search Experience
Understanding who you want to reach and how you want to communicate with them can answer several questions when evaluating whether to build or buy.
First, what kind of experience are you looking to offer — is this a company website? A support portal? A commerce store? Each touchpoint can provide information and answers to a variety of personas. When you know the experience, that next begs the question of what content — and that leads to where does that content live? This points to content repositories you’ll need to integrate or connect with to make that content accessible.
And, last but certainly not least, search is often evaluated in the context of a single digital platform — but we know that digital experiences are not based on a single channel. We interact with brands across channels and platforms, and it’s the sum of all digital touchpoints that defines a relevant experience. The expectations from one touchpoint can affect another, so how do you account for that?
By unifying digital journeys, so that users reap the benefits of not just relevant content but are served a much more relevant experience overall.
Crawling and Indexing
The biggest obstacles for the executives who are dissatisfied with their current GenAI deployments? Data-related issues have caused 55% to avoid certain use cases. Concerns included using sensitive data in models (58% had a high level of concern), data privacy issues (58%), and data security issues (57%). Organizations were much more worried about using sensitive data (e.g., customer or client data) than they were using their own proprietary data (e.g., sales, operational, financial).
Data forms the foundation of impactful digital experiences, yet it often lives across multiple disconnected systems and silos. When delivering the most relevant data, neither you nor your users should be worrying about which source the right data is hiding in. For generative AI application, seamless access to relevant data is essential; an LLM’s performance directly depends on the data it can retrieve, a principal commonly referred to as relevance augmented generation, or RAG.
Data can be made accessible in a variety of ways; in the search industry, there are three essential index types:
- Siloed
- Federated
- Unified
The first, siloed search, means searching within a particular system or repository, restricted to the search functions therein and only able to access the data stored in that repository (assuming you have the correct permissions to do so). To find what you’re looking for, you must search each system.
Federated search takes it one step further, where a search engine queries multiple repositories and brings back relevant/matching documents. While this brings search results into one user interface, it still puts the burden of understanding which document holds the answer on the user.
Modern search platforms offer unified search, which does much more than querying the indices of separate systems. Instead, it creates a unified index across content repositories and ranks the content for relevancy to not just the query entered but also information available from the user conducting the search, whether authenticated or not.
Unified indexing has an additional benefit: it enables the integration of powerful generative AI tools. Coveo’s new Passage Retrieval API for example, allows enterprises to leverage their unified index with a chosen LLM to enhance relevance in search outputs, in addition to a variety of other use cases.
Relevant Reading: Supercharge your RAG system with this Advanced Retrieval API
Understand What Kind of Search Technology You Need
Another scoping question you’ll want to answer is, how do my users search?
Commerce is often plagued with short queries of no more than two to three words, which can make it difficult to ascertain intent. For support questions, users might type in acronyms or a related name they refer to the product with, meaning you’ll need technology that can bridge the gap between their words and your content, so your users don’t end up with a zero results page.
When you understand who your audience is and what kind of experience you want to offer them, this points to the underlying technology needed to deliver on that experience.
Traditional search functions on keyword matching, also referred to as lexical search. While simple and fast, it typically cannot handle misspellings, synonyms, or polysemy (when a term has multiple meanings). And on top of that, it does not account for context or word meaning, which can lead to irrelevant results.
Today, searchers have come to expect the intuitive responses of semantic orvector search. This technology uses Natural Language Processing (NLP) to understand the meaning behind words and represents words as vectors in a high-dimensional space. The distance between words indicates their semantic similarity. Vector search can handle all the shortcomings of lexical search, but often requires a large amount of data to train a model — expensive and time-consuming if you built such a search platform yourself.
Both search techniques are equally valuable, which is why at Coveo we offer both as a combination that we call ‘hybrid search’ via our Passage Retrieval API and CRGA offerings. This means your search technique is applicable to the needs of the experience you want to deliver — it will not fall short when simple information retrieval is needed, but it can also extend to understand intent and surface answers, whether as exact snippets or generated results summarizing documents.
Modern Expectations like Personalization
Creating a consistent digital experience is one challenge; delivering a personalized one is another, especially when building your own search platform.
From where content is stored to how it is delivered (i.e., crawling and indexing, cloud availability), to what content should be accessible to whom (i.e., respecting security and permissions, complying with data governance regulations), and everything in between, offering personalization is a huge task when attempting to accomplish this alone.
The biggest benefit from choosing an off-the-shelf solution is the ongoing support and core competency you get from partnering with a software vendor. Coveo, for example, is rated #1 for Enterprise Search by Software Reviews, which surveys independent feedback from customers. What did they have to say about us? Look for yourself:
Analytics
A huge benefit of search is that it not only provides a great user experience, but also valuable business insights. In the same survey of technical stakeholders, 93% said that they see search analytics as valuable business intelligence.
An AI search platform like Coveo comes with analytics and a slew of reporting types out of the box. Plus, you can customize your own reports and even define your own search events, depending on what metrics are most important for your business.
Search analytics provide real-time data on user intent, preferences, and pain points by:
- Recording events: Events or actions (e.g., clicks, views, searches) are recorded by search platforms like Coveo. For example, our Usage Analytics (Coveo UA) tool captures and stores search data then makes it accessible via dashboards. If said event falls outside of the above described, you can also create custom events.
- Reporting trends: Dashboards like Coveo’s UA dashboard make it easy to identify search trends by helping you visualize data with charts and graphs.
- Customizing reports: Since every company is different, report customization is essential so you can focus on the data most relevant to your goals. You might, for example, want to focus on content gaps so you can create a content strategy focused on filling those gaps.
Search analytics are exceptionally good at uncovering hidden insights and trends in customer behavior because they allow you to track and analyze every action performed by each user.
Get the Search Platform Your Enterprise Needs
Open-source options have obvious benefits for those who choose to stick with them.
But for larger enterprises with technical debt like information silos and digital transformation goals to improve customer-facing, customer support, and employee search experiences would benefit from taking a measured and considered look at search for their enterprise tech strategy.
Want to get a closer look at the Coveo platform? Check out our on-demand demo hub, which highlights Coveo features like Automatic Relevance Tuning, Generative Answering, and much more.
Dig Deeper
Looking for more technical details for evaluating search platforms? We have an ebook for you: Buyer’s Guide to Enterprise Search Platforms. From indexing pipelines to intelligent automation, get under the hood so you feel confident in your choice of enterprise search solution.