What PandaDoc learned from trying to build an internal search engine

PandaDoc is the all-in-one tool to create, collaborate on, and eSign documents on a mission to help innovative organizations scale and win by eliminating workflow inefficiencies and bureaucracy.

We sat down with Maryia Yermakovich, a technical writer at PandaDoc in the R&D department, to deep dive into how crucial it is to have documentation easily accessible and searchable for engineers at all times. We dive deep into a range of topics, including: 

1. How PandaDoc organizes engineering documentation
2. Advice for managing knowledge
3. Why momentum loss is the enemy of efficiency
4. Lessons from trying to build an internal search engine

“My advice for combating drag and promoting knowledge management is to remove as much decision-making and blockers from the equation as possible.”

How PandaDoc Organizes Key Documentation

Maryia's department at PandaDoc is roughly 300 people. She focuses on internal engineering documentation.

At such a large scale, it's hard for her to police content; if she had to press the publish button on all of her department's output, she'd be a massive bottleneck.

Instead, she aims to bring excellence to her department's product through structure and guidance.

PandaDoc's Information Tooling Stack

Maryia jokes that just listing all of PandaDoc's tools already sounds overwhelming, but she assures us that other companies use just as many tools.

PandaDoc has a publicly available version of its documentation for both customers and external developers. Still, where their external documentation is beginner-friendly and straightforward, their "under the hood” information systems embrace inherent complexity. For example, tool overview is a part of the engineering boot camp for new PandaDoc employees.

PandaDoc's engineers use GitLab for storing and managing code, Atlassian stack (Confluence and Jira) for product management, and Google suite for mail, spreadsheets, and presentations. Slack is for group chats and instant communication, Miro is a virtual ideation whiteboard, and Coda is the primary documentation station for product and engineering.

“The amount of tools we use is intimidating, but it's also prevalent in this industry.”

Knowledge Management at PandaDoc 

Knowledge management at PandaDoc is decentralized. It relies on its engineers and product teams to do most of the heavy lifting regarding info searching. 

Enforcing how people search for documentation at PandaDoc's scale would be both unpopular and impossible, but a certain amount of uniformity is necessary to maintain order. 

Maryia employs these strategies to help the process:

• She "builds paths where people already walk" by noticing where employees visit the most, what they search for the most, and what they ask for in Slack.
• She leverages content analytics to highlight the most prominent content pieces. 
• She influences how the department presents knowledge at the highest level. 

‍Documentation Hurdles Across a Global Workforce

Recently, a considerable portion of PandaDoc's employees relocated to several countries. During the first wave of relocation, HR created an all-inclusive document to help workers transition.

It was readily available on Coda with lots of helpful visa information for those moving. Unfortunately, later waves of relocation were smaller, so the useful document wasn't updated on Coda as frequently, and it became harder to find.

This phenomenon is a peculiarity of Coda, where if something remains untouched for a week, it sinks to the bottom of the feed even if it's still valuable. 

“On Coda, you must run to stay in the same place.”

{{demo}}

Why Momentum Loss is the Enemy of Efficiency

Maryia wants to enforce positive work patterns in her department. 

She noticed that engineers with the power to search for information alone are unstoppable. In her words, they become "fantastic methodological 10x engineers."

It takes a while to return to the original task after an interruption, and every question waiting to be answered is an interruption. As a result, it’s easy to fall into unproductive patterns where if you can't find anything – and sometimes don’t know what to search for – you take a break, the tasks drag on, and the team loses efficiency.

Maryia's perfect information access tool would be:

1. Highly Customizable – She realizes that people are reluctant to accept change; they cling to old tools and treat the new ones as “yet another tool.” She would let people have their preferred functionality.
2. Full of Choices – Engineers are builders who like to be included and play with tools. The ideal tool would allow them to remove the buttons they don’t need or choose helpful add-ons that fit their purposes.
3. Respond to a Real Need – A tool must prove extremely useful, and not just a “nice to have”.

How to Accelerate Engineering Momentum

Maryia advises enforcing productive knowledge management patterns to eliminate the decision-making process from as many repetitive tasks as possible.

If a technical writer wants their engineers to write or read, engineers must know exactly where they should be writing and if what they're writing is relevant. 

“Engineers lose focus when they wait for answers.”

Lessons From Trying to Build an Internal Search Engine

Many enterprise-stage companies like Invisible often try to build an internal Dashworks-esque solution to improve the ease of documentation discovery. Unfortunately, while some teams may initially succeed, the systems quickly fall apart due to the sheer complexity of the task.

Many companies hire an in-house development team to keep the engine running. If a company decides to build a search engine in-house, they’re forced to build each connector individually. Using Dashworks, 43+ connectors are included out of the box.

Diving deeper, it took our engineers at Dashworks an average of 100 hours to build a new connector – approximately $5,000 spent per connector built.

PandaDoc experimented with building an internal knowledge engine prototype but found that APIs in their tools changed so frequently that their search function kept malfunctioning.

A search engine solution can’t keep crashing because when employees stop believing in a tool, they stop using it, which ruins their momentum.

Engineers also won't run individual searches through five different apps, as that's a waste of time; ultimately, they post questions in Slack for team members to answer.

Cognitive Complexity and Searching for Documentation

Globally, documentation is widely available on the user side but frequently overlooked on the developer side. Internal documentation is trying to describe a thing as you build it, so it’s never finished and is always a work in progress. Gaps in documentation translate into a fuzzy discontent that's hard to measure.

In today’s world, cognitive complexity stays a hot topic. When addressing cognitive overload with knowledge management, you must tread carefully, pick the right tools, and, most importantly, listen to feedback from your colleagues.

Want to keep reading on managing complexity in the engineering org? Explore PandaDoc Bamboo Magazine.

{{demo}}