Chat with us, powered by LiveChat
scientific research in social decision making. what's in a word? indicators of task completion

What’s in a Word? Indicators of task completion

Previously on this blog series we investigated a few factors that contributed to the odds of a decision being closed. This time we take a step towards more concrete an entity and focus on tasks. Decisions remain meaningless ink on the paper (or in the digital era, bits in a software tool) unless they are implemented. Some decisions are so small in scope they do not require further tools to help putting them into action. However, some strategic decisions can be huge in scope in which case it becomes necessary to properly organize the execution phase using for example Kanban tools. This is where tasks come in Fingertip.

Tasks, as their name implies, are concrete assignments for a person to check off their to-do list. Fingertip allows the user to create tasks under a decision entity and provides handy project management tools for keeping tabs on the decision execution phase. However, much like decisions, not all tasks are finished. Some tasks are never even started. What a nightmare for a project manager!

This blog post revolves around how the verbal task description might relate to the task completion rate. We employ some tricks from text analytics in our quest for figuring out which tasks have a tendency to be forgotten. The data we use comes from Fingertip using only tasks created in 2018 and earlier.

Title Length

We begin by simply looking at how the amount of characters in task title relate to the odds of task being completed. One can hypothesize that tasks described using fewer letters are simpler in nature thus easier to accomplish. If you need to write a small essay for task title, you might not be exactly sure on what it is that needs accomplishing in the first place. And indeed, the median for completed tasks is lower than in the other tasks and the distribution is more left-skewed than most. The mean character count in completed tasks is 35,9 while in all other statuses it is over 40. Anova test proves the difference is statistically significant. Clearly, tasks using fewer characters in title are completed at higher rate than more tasks with more prosaic titles!

Task status in relation to task’s title length

Prevalent words

Next, lets look at the words that appear in completed tasks vs the words that are prominent in the remainder. For this purpose we build a comparison word cloud with words more often appearing in completed tasks in orange and the words used often in tasks that are still on-going in teal. Quickly glancing at the word cloud already gives a quick idea on which tasks have a higher than average tendency to remain uncompleted. Words in teal have a lot of technology and software development related terms (development, documentation, production, change), while the ones in orange relate more to sales and marketing (e-mail, call, trial, meeting, website).

I can not attest whether these are universal themes or just Fingertip-related. As a data scientist I am more accustomed to the IT side of organizations and could theorize that agile methodology might render many tasks redundant when the development moves in quick succession, while in sales you call your customer, do or die.

wordcloud for words appearing often in open and closed tasks
Words frequently found in completed tasks (yes) and incomplete tasks (no)

Frequent topics

Finally, I will run a topic model algorithm, which can be described as a clustering method applied to text data. It is a statistical method which attempts to discover latent topics from a corpus of text data (say, sport news, politics news or entertainment news) on the assumption that each topic generates certain words at different rates (sports news for example often employ vocabulary that could as well be used describing military conflicts but you tend not to use them for celebrity gossip news). After topics have been developed from a data set, each document can be assigned to one or more topics based on the words used.

My topic model algorithm discovered three distinct topics that are best described by following words:

Topic 1Topic 2Topic 3

The first topic seems to relate to administering the Fingertip software, with ”user” the most distinctive word. HR-related tasks likely fall under this umbrella. Second topic is clearly sales, with email, trial and call as exemplary words. Third topic relates to software development, with ”test”, ”add”, ”tab” and two key terms used in the software.

Cross-tabulating the found topics against task closure finds that tasks falling under topic 2 are completed at twice the rate compared to the other two topics. Topic 3, software development that is, has the worst record, with topic 1 lingering few percentage points ahead.


Now these results naturally are not generalizable over all organizations at all times. First and foremost, they describe Fingertip as an organization and the different people and cultures working at different tasks. This information is crucial for us as an organization and I dare say it is so to other organizations as well. Fingertip is a great tool for finding out how your tasks fare. Tasks as a Fingertip object have traits beyond their titles, such as priority, that can also yield nuggets of gold when mined!

Read the previous installment on our scientific blog series:

Why good decisions get implemented

Too many cooks spoils the decision?

Lasse Winter is the leading Data Scientist at Fingertip with a background in social sciences and a specific interest in text analytics. He is passionate about gaining exciting insights from data. During his free time Lasse loves sports and reading, with a specific passion for football.

Fingertip is an online decision-focused business management solution designed to substantially improve efficiency, effectiveness, and empowerment in large complex organizations in the digital age.

Fingertip Licenses


Please fill in the license order form below to receive your licenses.