Show HN: My Minimum Viable Beta - Hashtagify Pro | Hacker News Reader

Show HN: My Minimum Viable Beta - Hashtagify Pro

pro.hashtagify.me

12 pointsdanmaz747 comments

I like the data that you're collecting. However, some of the visualizations are ineffective. Perhaps it sells in your target market, but things like bubble charts and gauges are horrible for showing data. You have pie charts but then proceed to display the numbers on the graph. Why do that? A table would have sufficed. Pie graphs don't demonstrate much other than that people can't figure out angles.

If you're making a metrics site you should look in to some of Edward Tufte's works. Things like sparklines (http://en.wikipedia.org/wiki/Edward_Tufte#Sparkline) allow you to display a large amount of information in a very compact area, and with high resolution. I look at "Average daily tweets" and see 22.5. Great. Was today a low day or high day? If you give me the history in something such as a sparkline, I can tell if this is increasing, decreasing, or simply an off day. Did Christmas influence the number of tweets? If I go back tomorrow I won't be able to see today's information in quite the same form, which may be what I wanted.

Sorry if this seems like a rant, but I see a lot of data visualizations at work. I drown in metrics, charts, and graphs. Most have information I want, which is why I look at them. However, I can't always get the information I need.

Just make sure your customer gets the information they need out of your graphs, not just what you want to show them.

Thank you for your suggestions.

The "Executive Summary", where you see gauges and pie charts, is where you should get at a quick glance an Idea about how your campaign is going: If the trend is good or bad, and you your tracked influencers contributed to that trend.

To signal a positive, negative or neutral trend, I didn't find a better alternative to gauges. Please notice that the actual numbers the gauges show aren't easy to interpret - they're related to the steepness of the linear regression of the daily reach. So what I'm actually communicating could be expressed as "a little positive" "strongly positive" "strongly negative", but a gauge look good to me for that. Reading the number isn't that significant per se.

I also remember a heated discussion here on HN about pie charts, but I'm still convinced that they're good when you need to show the relative impact of a few alternatives.

Including sparklines in the first page could be a good idea though, I'll have to see that with my target market - for now I only had one reference customer to talk to, so I reckon that that isn't a good statistical sample :)

More detailed data is in fact shown in subsequent tabs.

Last thing: The bubble graph isn't good if you need to really compare the dimension shown with the bubble radius, but the other two dimensions here are much more important and for those I think this graph works really well. What alternative would you suggest to show three dimensions in a single chart?

Thanks!

I had found the data in your other tabs. Sparklines aren't necessarily reserved to summaries; I didn't mean to give that impression.

As an alternative to gauges, you could try a single horizontal bar, with background coloring for what you have as red, yellow, and green (picture a sparkline but as a bar). It shows the same information in less space and lets you show other related data to the side.

But as I said, if it sells disregard everything I've said. There's a discrepancy between the people who buy this stuff and the people who use it, and obviously you're trying to pitch to the guy with the money. People like pie graphs for instance. You probably shouldn't ditch the pie, even though a 4x2 table would show the same data.

Clustered bar graphs are generally an alternative to bubble charts, but it depends what you're representing whether that works or if you should split the data into multiple graphs.

I have to ask what that graph is really showing. Do I read it from the center of the dot? 'Nodejscommunity' is more specialized and more influencing than 'ryah' by that interpretation. Or should 'nodejscommunity' be more influencing because it has more followers?

Another question to ponder is whether the executive summary is a summary or dashboard. At my company the definition of a dashboard is something that doesn't scroll. So for whatever resolution you design for the data should fit on one page without scrolling. That would allow for someone to quickly open the page and make inferences about your data. A summary of course contains more information, but still at a high level.

Found an example for a dashboard: http://media.juiceanalytics.com/images/BonaVista_dashboard.p... It even includes the horizontal bar I'm referring to.

As this is in beta, if it sells or not is something I still don't know - so first I'll see how it goes with these kinds of graphs, but hearing more ideas is very useful to be able to do variations and see if and how they matter. For now my main concern is bugfixing and preparing to be able to sell, but then I'll work again on the refinement of the product itself.

In my summary I made the intentional choice of showing only the essential info. I hope to find out soon enough if that is exactly the information that my target is most interested in - my prospective customer likes it. But that's just one.

As for it being a dashboard or not, with content that is all above the fold... I could make the page liquid by adapting the size of the gauges and pie charts to your resolution, or just reduce their default sizes some to have everything above the fold with most resolutions. I'll consider that.

For the influence graph: Yes, the position on the axes is that of the center of the dot, so ryah in the diagram is less influential than nodejscommunity even having almost 10 times its followers. How is that? I don't calculate influence only based on your followers, but also on how much you tweet (in general and on the specific subject), how much you're retweeted or cited... that's why it's interesting to see more dimensions in that graph. The dimensions aren't independant from one another, and from that graph you can get a very good idea not just about how influential a user is, but also why/how.

But you made me realize that I need to explain that graph even better. I'll try to address this in my next iteration ;)

EDIT PS: I'll also add that the horizontal bars you linked are a good alternative especially if you have a target to reach. With negative/positive numbers and no target, I should try them to see how they work.

Some months ago I created hashtagify.me - just because it looked like a very interesting idea that would allow me to experiment with node.js and redis and to get some attention.

While working on other ideas, I was contacted by a marketer that needed a specific kind of analysis for Twitter campaigns. Hashtagify Pro is my solution to his - and I hope enough others' - problem.

This is still a closed beta, but I'd like to get all the feedback I can get - be generous with your comments, ideas and criticism!

EDIT: If you're interested in the service, I'm looking for beta testers, so just ask - here or writing to daniele.mazzini(at)gmail.com - and I'll send you an unlimited beta account.

looks like a great product. I think the demo interface you have is great and fairly easy to use.

I'm not your target customer, but I think the demo / tutorial might be a little daunting with all of the numbers around everywhere.

I also think there might need to be a little bit more explanation about what those individual dashboards mean or what type of action you can take against it.

best of luck with this.

Thanks! I too had second thoughts about the number of "instructions", but I also thought that those who are really (ie practically) interested will want in-depth explanations. Let's see what others think ;)

For a more "strategical" explanation I was also thinking about writing a short (free) ebook. I don't know if a contextual help would be a good place to explain that kind of things...

Reply on news.ycombinator.com