MobileMonday London does mobile search

By Dr Paddy Byers


Last night’s Mobile Monday event addressed the topic of mobile search and was definitely one of the better events I’ve been to. We heard from the BBC, Yell, taptu, Mobile People and MCN. A few themes came out of the presentations and subsequent discussion and I’ve put down my thoughts on them below. Note also Paul Segala’s comments on the evening.

Discussion points

Disputed demand for mobile search. This is the issue as to the latent demand or opportunity for mobile search as compared with conventional web search, and is really just a specific instance of the general “demand for mobile web” question. Diametrically opposing views were expressed. We heard that currently mobile web searches are running at about 1% of web searches, and that possibly more than half of these searches are miscounted. Some cited this as evidence that mobile search is “broken”. For me, I thought it was a surprisingly high figure, given the state of the art – I use my mobile browser more than most, but I know I do at least 100 web searches for each mobile search. It reminded me of Michael Mace’s recent post where he urged us to set real-world expectations for takeup of mobile services.

My take-away was that the motivations for mobile searches are usually different – more often driven by an immediate need, less often idle curiosity, and almost never just recreational browsing – but the type of information being sought is just as varied.

The mobile user has an attention deficit problem. Maybe “persistence deficit” would be a better description. If a search takes too long (ie measured in terms of time or the number of clicks and scrolls) then the average user decides it’s not worth the effort. A threshold of 12 clicks+scrolls was cited; in an independent study more than 90% users are prepared to put in this amount of effort, but fewer than 30% are prepared to continue for more than 20.

My take on this is that it depends on the information being sought and the motivation. My guess is that the data reported came from much more casual situations – eg finding a new wallpaper to download – than finding the number of a taxi when stranded in Brixton at 1am. But generally I agree that if a search doesn’t get you what you want (or come very close) first time around, most will decide that they really can’t be bothered.

Some companies are motivated by this to build search services that are quite different from the traditional web search model, discussed below.

The small screen size obviously also prevents the web search experience transferring directly to mobile, with one describing the screen as a “pinhole” to the world.

Browser-based interface vs resident client. There was some discussion on this, just as there is for every other connected service. You do have to wonder what could be more naturally browser-oriented than search.

The proponents of the resident client argued that this gave a much more usable service (better interactivity and lower latency); in principle it could also get access to location information and use it to improve search context. However, my main take on this was the that resident client approaches arose not because their usability is so good, but because the browser experience is so bad. There seemed to be a consensus that there was a case for resident apps now but browser-based experiences will dominate eventually.

Location-based search is nothing like as good as you might think. The idea that you can use location to inform and prioritise searches is oversold. First, there are technical weaknesses – position acquisition takes too long, and is often inaccurate – and for many use cases the whole concept is flawed. Often I’ll want my searches to target the place I’m going to be next, not where I am now.

Business model. There were several issues discussed here. Independent vs content-owner search services (BBC as an example); subscription or pay-per-use services vs free-to-air advertising-funded services; plus the ever-thorny issue or operator-provided search services (with walled and unwalled search scope).

On the first issue, I think these both have a role. In particular, I expect a content owner’s own search facility to do a far better job of indexing and searching their site than I could do myself using an independent search engine. Furthermore, if I’ve gone to the effort of navigating to a specific site’s search feature, I’ll want it to get me valid results pretty much straight away. However, even on the conventional web, few sites’ own search facilities outperform Google so these need to improve to deliver any value.

On the second issue, there seemed a pretty strong consensus that the advertising model will dominate mobile search, as it does on the web today. There are some arguments for advertising being even more powerful on mobile – a mobile user performing a search will arguably have a greater buying intent because there are fewer casual window-shoppers. The point was made that a mobile search will often be made very close to a buying decision, so there is huge value for an advertiser being able to inject a specific message at that point.

Operator-controlled search services seemed to be the elephant in the room; these were hardly discussed. However, they clearly have significant power and arguably the operators have the most to lose but also the most to gain from effective unwalled mobile search.

User interface. This is another area where there were directly conflicting views expressed by different speakers. Some advocated the simplest possible interface with a text box for search terms; others looked to diverse interfaces (eg voice, sms) and forward to more interactive interfaces (eg the possibilities offered by an iPhone touchscreen).

Semantic search

Underlying what was presented there is a clear divergence in philosophy between the various search approaches. Nobody disagrees that the constraints of the mobile environment (limited tolerance of navigation, false results, limited content compatibility, etc) mean that services need to be improved relative to, say, the existing Google mobile web search offering. One obvious approach is to start from the full generality of a text search engine, but do a better job of pruning and sorting results.

The other approach regards any result that takes more than 12 clicks to reach as unreachable and therefore makes no attempt to reach those results. Instead, these approaches are based on the idea that the majority of mobile-accessible information on the web is already in structured repositories. Therefore, they create semantically-aware search services, allowing service providers and merchants to hook their sites up to the search service explicitly and programmatically. Based an understanding of the data, the service can reduce the experience to a 3-stage process: search-select-consume. I first mentioned this approach when MCN announced their solution at Mobile 2.0; yesterday MCN demonstrated their first commercial incarnation of the service. Taptu were unable to elaborate much on their solution but I’m speculating that it’s essentially the same approach.

So, how do these approaches compare? The semantic approach sounds a little like a walled garden – in that the service provider enables access to a specific whitelisted set of sites (ie those whose data is understood by the engine). In this context, is this closed system just a practical necessity given the constraints of the mobile, or could it become a commercial barrier like the protectionist walled garden of the traditional portal?

The other argument on the side of the semantic approach is the need to ensure the service is effective in order to build user confidence; it is better to provide a limited service very well than it is to provide wide-ranging functionality but to leave users frustrated. So, in this embryonic market, the semantic approach is arguably a good way to incubate subscriber confidence and build a mobile search brand.

Also, how good are the two approaches at delivering the desired experience for the different kinds of search? There is a category of searches that have an “deterministic” desired result – such locating a specific local service (taxi, flowers) or a specific business; if those businesses or products are known to the semantic search engine then it will perform extremely well. However, there is also a broad category of searches that are much more speculative and no significant subset of sites will stand a chance of supporting those requests effectively. The example from the BBC speaker was “when were yellow balls first introduced at Wimbledon?” (which was an actual instance of a query successfully resolved using the Ask mobile service).

It seems that somehow both approaches need to be used and optimised for the context of each kind of query.

The missing context

In the days before the internet we used a variety of sources of information. If you wanted the phone number of a known local restaurant you’d look in the phone book; if you wanted to discover a local restaurant you’d look in the Yellow Pages; if you wanted news you’d look in a newspaper; and for other information you might use specific reference sources, or go to the library. You were implicitly deciding the kind of search you were after by your choice of source.

When making an internet search, much of this context is gone; you put some search terms into Google, expecting it to do all of those kinds of searches and with perfect results, without ever having to tell it what kind of search you were doing. The semantic and traditional search approaches illustrate how this context, if available, would allow the engine to choose its scope and algorithm appropriately. Getting past this initial, implicit, communication between the user and service seems to be one of the barriers. Add to it the awkwardness of entering search terms, the limited viewability of results, the time taken to navigate hierarchical information, and it’s easy to see why it’s got some way to go.

Someone expressed the view that “users don’t want to say what kind of search they want before they start”. If so, there has to be some other way of establishing the context. Yell do it by having a brand proposition that replicates the yellow pages. (Except it doesn’t – it replicates the Yellow Pages plus the commercial section of the phone book. So if you’re going to Yell to discover, you can easily find that you’re frustrated because the first umpteen results in any category are the businesses you already knew about but didn’t want.) Another approach is one where each result returned also has a “show me more results like this one…” link, so the user has another way of adding context by indicating the suitability of a particular kind of result.

The bottom line

Mobile search can grow massively but still remain much smaller than conventional web search. Don’t expect it ever to get that big.

Mobile search still hasn’t settled on usage and business models. There are at least as many advances needed in usage and business models as there are in technology before it fulfils its real potential.

Over the coming years, it will continue to be a fertile ground for innovation, yet still a huge frustration for users who, for the most part, only had modest expectations to begin with. Eventually, powered by advertising money, someone will make a tidy sum from it. Will we all be better off?