Invaluable advice from Hostelworld on upgrading clients to a web based management application from their standalone product

I just met Gillian Lennox – who since 2006 has been the product manager for Backpack Online which is Hostelworld’s hostel management application. Fascinating hour – she has a wealth of experience and advice relating to moving businesses from standalone software to a web based application – and it was not an easy ride!

In 2006 when she joined they had 280 Hostels using their PMS (Property management software) which was an standalone application like Memento 5 in delphi. (They had 10,000 other hostels who just uploaded free beds)
This PMS used to sync every 4 hours with Hostelworld.com.
They wanted to move to a web based application for 2 reasons.
1. If the hostel used a web based application – they could get access to all their free slots in real time
2. Support etc would be much easier.
(BTW they did get a huge increase in bookings once they had moved over)
From 2006 once the new software was ready all new clients got the web based software but they met massive resistance from existing clients who were very afraid of going web based only. Many of them did not trust the internet etc.
In face as of today – they still have 17 hostels on the old application who are still refusing to come over. Thats out of 1450 hostels in total who use backpack online. (There are 26,000 hostels on hostelworld.com but most still just upload free capacity). In the end it took them until 2008 to get most of the old hostels upgraded.
The 3 major issues she said they faced were:
1. Internet – people were afraid to go web based
2. People were afraid of double bookings (even thought they are web based – its based in php so there is a 60 second refresh where things can be double booked
3. Data – clients were concerned that the company data would now be out of their hands especially as there was data about bookings coming from other sources like Hostelworld competitors like Expedia.
1. For the first issue, they had to have a minimum standard of broadband – 1MB. And for the big sites they recommended they have a cheap backup broadband from a different supplier (e.g. our 3G dongle idea)
They had to spend a long time coaxing customers to try it out and see that it was better. With new clients it was much easier as they had no other expectations. Also the online software initially was not as feature rich as the standalone version – took 18 months to catch up.
Every morning they send the hostels an text based email with all their bookings for next 14 days in case the hostel can’t get online.
2. With Overbooking – she explained that overbookings have always happened in any reservation taking business – its a fact of life in a restaurant or in a salon or in a hotel. However in the past the business would accept the responsibility of this themselves but now with software they have someone to pin the blame on. For them it is a rare occurrence but it does happen and they explain up front during training that it may occur and mistakes do happen so not to get frustrated when it does – just deal with it as normal (They did try to build features in to the system that reserved beds just in case there was a double booking but it became too complicated and error prone so they ditched it.)
3. Regarding Data Protection – the main issue here was in Germany who have very strict privacy laws. As they were hosting people’s data they had to get a written contract from each hostel in germany. Also every person internally who works on backpack online has to sign a special contract/NDA regarding all the data that is hosted.
Gillian also summed up 3 pieces of advice:
1. Moving online has to be less hassle than what they are currently doing – you must re-assure them of this.
2. Continue to offer the standalone based software if you have to but make it much more expensive (this is something they couldn’t do because backpack online is free as they make a lot from all the bookings)
3. Be really clear and upfront with what can go wrong and how to avoid/deal with it – this way customers expectations will fall into line.

Results from our Wisdom of Crowds experiment

Results (152 Votes)

Python 30%

Java 25%

Other 17%

Ruby 9%

PHP 7%

Perl 7%

.NET 3%

MONO 1%

Delphi 1%

MySQL was the clear winner in the database vote with 51% followed by PostgreSQL with 21%. Click here for results

Analysis

This was the second vote we had but there was an interesting anomaly in both votes. In the first one, Ruby was the clear winner while Python was the clear winner in the second. These are similar languages so i was interested to know why one would outdo the other in different votes. A possible explanation is that most of the votes on the first one came from Reddit users where Ruby is a popular discussion topic (54 articles on reddit are about Ruby directly opposed to 37 about Python) while most of the votes in the second one came from Google staff where Python is king.

Result

We have decided to go with Java. It was a close second in the vote but it won out in our minds for two reasons. 1. POS applications need to be lightning fast and with Java we are more confident of the application being as fast as possible. 2. We have Java expertise and we think this expertise gives Java the weight it needs to shade it. I’m sure Python would have been an excellent choice too and you never know the next next version of Memento may be in Python.

I need your help with an experiment Part 2

Yesterday evening I ran a “Wisdom of Crowds” experiment relating to the technical direction of my company, Phorest. The concept for the experiment came from an idea we had, that if the opinion of a crowd is smarter than the smartest people in it – what would happen if the strategic direction of our company was decided by the crowd? Surely with enough information they would do a better job than we could? Or at the very least the crowd would be represented at meetings as a person – albeit the “smartest” person in the room. We thought to start with the current dilemma/debate facing the company – what technology to use in the next version of our core product – Memento (Touchscreen based, management software with POS and Appointment scheduling for hair and beauty salons)

The results of the vote were inconclusive (although Ruby was very popular) – but this is what we learnt:

  1. Most people voted “Other” – so the choices were not clear enough or there weren’t enough choices
  2. Over 500 people read the blog but only 110 voted (as of 3.30 GMT today)
  3. 98% of people came from Reddit/Google Reader and not Digg (It was posted on both under programming)
  4. From the posted comments it would seem i did not give enough background to enable people to make a decision.

So i am going to try again.

The next version of our core software will be web based. There are two issues to take into account. Firstly, it will be a touchscreen GUI and needs to be instantly responsive. Therefore we will be building a very thin client (that rarely gets updated) so that we only send requests back to the server when completing an operation. Secondly because it is a POS system, it needs to keep working in a limited capacity if the Internet goes down, To deal with this, we will be able to queue certain requests in the thin client (e.g. a sales transaction) until the PC is back online.  The team has good experience in Java, Delphi and Ruby. So what language and database should we use?

Free Vote Caster from Bravenet.com

I need your help with an experiment.

I have just finished reading The Wisdom of Crowds by James Surowiecki. The basic premise is that the collective judgement of a crowd is on average smarter than the smartest person in that crowd. As an experiment I’d like to try this out with some decisions that our company has to make in the next few months and report back in this blog with what happens.

I believe that Reddit, and Digg users would represent the four elements required to form a wise crowd:

Diversity of opinion
Each person should have private information even if it’s just an eccentric interpretation of the known facts.
Independence
People’s opinions aren’t determined by the opinions of those around them.
Decentralization
People are able to specialize and draw on local knowledge.
Aggregation
Some mechanism exists for turning private judgments into a collective decision – (we will use the voting app below.)

Our company, Phorest, is a software developer in the Hair and Beauty space. We are about to embark on the development of the next version of our core product Memento. Memento is currently a standalone Appointment and POS application built in Delphi 7. We have spent a lot of time researching about whether to go web based and if so – what technology to use. Our only concern with going web based until now is what happens in a busy salon if the internet connection goes down. POS applications cannot afford to go down – even for a few minutes. However there does seem to be some possible solutions to continue working offline in a limited capacity with current web technologies.

We have only three basic requirements for the solution:

  1. Touch Screen GUI (Big buttons)
  2. Robust
  3. Evolutive (Easy to evolve and add new features)

The idea behind wisdom of crowds is that you should vote even if you have very limited knowledge. Please click below to vote. (A lot of people are voting other – it would be great if you could post a comment here with what your “other” is. Bravenet voting doesn’t let me colelct that unfortunately)

Free Vote Caster from Bravenet.com