table1.cc Blog

Reproducible and Efficient Creation of ‘Table 1’ Summary Tables for Research Papers

Lessons from the First 12 Days of table1.cc

In this post we review the progress and development of our project that we began on December 22, 2019. At that date, we set out with the goal to significantly simplify and improve the process of creating Table 1 for clinical researchers and epidemiologists - more easily and efficiently through automation. Along this way we have learned several lessons we would like to share for people interested in our journey. We also believe this may be interesting in anyone else thinking to start their own website or project.

1. Google is important - critically important for your success

Unsurprisingly, as a new website or service, you require your target audience (researchers who need to create ‘Table 1’ summary statistics and value reproducibility and simplicity in our case), to know about the service you are offering. Nowadays, almost everybody finds things on the internet because of two things:

  1. Recommendations from within their social network
  2. The Google search engine

These two are what you must focus on. On a technical side, the Google SEO (Search Engine Optimization) Guide is a must-read. It’s a concise and crucial reference that helps you simplify effective scraping of your website via googlebot, Google's web scraper.

Figure 1. After a slow start, Google search traffic is growing consistently after adopting the Google recommended best practices. Source: Google Search Console

Figure 1. After a slow start, Google search traffic is growing consistently after adopting the Google recommended best practices. Source: Google Search Console

We have repeatedly updated our blog with 2-3 weekly posts, and plan to continue to do so in the future. As a result, despite starting out with zero popularity, we have seen a great increase in Google, initially manifested by an increase in googlebot visits to our sites. This has since directly improved our Google ranking, daily impressions (page views), and visitor numbers have started to increase (Fig. 1).

2. Twitter is good at amplification of popular content, but less effective than Google in starting to get content popular, especially when you are just starting out

Some of our other ways to create an audience for our automatic Table 1 generation solution have included the use of the Twitter network. This is a relevant player, but mostly so for established services and websites. Thus far, Twitter has been underwhelming for our project: We started out with zero followers and at the current time, 2 weeks and 19 tweets later, we have a total of 2 followers, both being us, the founders of the website.

Figure 2. One of our very modestly successful tweets. Source: table1.cc Twitter account

Figure 2. One of our very modestly successful tweets. Source: table1.cc Twitter account

In somewhat related news, feel free to pay us a visit, and check out our Twitter account ;)

Despite some high profile likes (Thanks, Dr Warrington!) Twitter has not been overly effective thus far in driving traffic. In summary, based on our experiences, Google rates quality content as defined by the Google Search algorithm, whereas Twitter emphasizes viral content when selecting what to propagate. (Fig. 3)

Figure 3. Difficult design choices and weighting of quality and popularity signals between Google and Twitter. Source: personal observation

Figure 3. Difficult design choices and weighting of quality and popularity signals between Google and Twitter. Source: personal observation

3. Search for the right metrics to track

During the first days we counted website traffic as a main measure of success. It was greatly rewarding seeing the access logs from our server increase in time. It took a few days to realize that a large portion of visitors were browsing our site, and reading the blog - yet only a very small fraction were actually using it - which only improved after we modified and simplified our site structure to help route visitors to the relevant landing page.

We are now tracking daily submissions of so-called POST requests (these are requests sent by the web browser when a visitor submits the data set) as our current key measure of utilization, which helps us more accurately model actual usage.

Figure 4. Our initial metrics were overestimating actual usage. We are now tracking POST requests as a proxy of service utilization.

Figure 4. Our initial metrics were overestimating actual usage. We are now tracking POST requests as a proxy of service utilization.

4.Engaging with your target audience beats all else, both on a fact as well as on an emotional level

Even though table1.cc is a web based solution, we look to solve real world challenges for potential research users.

By a great margin, the most rewarding aspect of working on this service has been to have the opportunity to engage with the potential target audience. It is such a rewarding experience to see live someone else use the product you are building — even while it is still in its infancy — and open the dialogue on how to develop this into an even better product in the future.

Thank you all very much for your suggestions and supportive feedback, it’s truly inspiring and we will work to implement it ASAP!

We greatly appreciate everyone’s feedback and would like to especially thank our focus group volunteers:

Bottom line

Critical success factor is the direct engagement with potential users.

And many thanks to you for your interest as well. Reach out to us on Twitter @table1cc with any comments — plus you can become our 3rd follower! ;) — as well as Feedback.

And if you haven’t already, check out table1.cc, the website that helps you generate your Table 1 summary statistics in no time.