The Rocket Model: Match

This is part two of our series, adapted from the best-selling book Platform Strategy. The previous article introduced the first stage of the rocket model for platform businesses: Attract


Finding the right options in a world of abundance 

After participants on both sides have been independently drawn to the platform and onboarded during the Attract stage, the Match stage of the rocket model brings them together. In order for both sides to interact, they need to be introduced first. This is unique to platform businesses since traditional firms sell directly and do not need a matching function.


For Airbnb, that involves presenting guests with the right properties in the right locations available at the right time; for Upwork, a global talent platform, it’s about matching companies with available freelancers who have the right skillset for the specific assignment.


The quality of the matching is critical to the success of the platform. In a world of abundance, the ability to filter and present customers with the right choices creates significant value. The matching function can also play a key role in helping the platform maximise positive experiences and network effects. For example, Amazon Marketplace prioritises product search results with the best price but the search algorithm also de-prioritises merchants with poor feedback. This ensures that consumers are less likely to buy from merchants that provide bad customer experiences, and this incentivises good merchants to join the platform which in turns attract consumers. Such self reinforcing loop illustrates how good matching helps enhance the health of the entire ecosystem.


Many ways to match

There are different ways to match participants. Matching can be done through a search function with selected parameters for product marketplaces, such as eBay or Amazon. For service platforms, a specific graphic interface tailored to the focus of the platform (e.g. geo-localised maps for Airbnb) is often used. Sometimes, the matching function uses the information that both sides of the market have provided to the platform, such as pictures and profile information on a dating site. It’s often useful to think about what kind of matching interface would provide the best user experience for participants at this stage. This will allow you to project yourself and see if a search box, a map, pictures or drop down menus are the most suitable way of matching your participants.


The complexity of the matching function depends on a number of factors. Horizontal platforms, with a wide range of products and services, usually require stronger search or matching functions than vertical ones that are more specialised. In some cases, however, the matching function is almost implied rather than explicit. This is often the case with payment networks (i.e. merchants are not actively matched with cardholders, although ‘Amex accepted here’ signs may help). The matching function can also be a mix of search and self-selection.


Ideally, the matching function is configured to provide the optimal level of choice required for a successful transaction. Some economists assume that maximising choices is always a good thing, yet we find that this is not always the case and can in fact result in reduced transactions. Finding the right balance between too many choices, which would confuse buyers, and not enough options, which would drive buyers away, is not an easy task. A curated, prioritised, relevant and timely selection maximising the likelihood of transactions occurring is the nirvana of platforms. This optimal matching may also be different from one user to another and depend on context (mobile vs. desktop) and therefore ideally needs to be personalised.


Criteria for effective matching

For matching to be effective for participants, results must be:

  1. relevant, i.e. meet participant needs, 
  2. filtered, i.e. present the right amount of information or depth, and
  3. timely.


The matching function should be optimised to return enough results for the user to find a relevant match but not so many as to overwhelm the user with too much information.

We can examine these criteria further in order to assess how to achieve effective matching.

Relevant results are those with the highest matching quality and which will lead to transactions. Search results should be ranked for relevance against one or a combination of key criteria such as:

  • product characteristics
  • price
  • availability
  • location
  • trust rankings


Results that have been filtered offer optimal depth without sacrificing choice. The selection need not be exhaustive but should be tailored, and may also be context dependent, e.g.results accessed on mobile may prioritise mobile-optimised options. Search filters are driven both by user-selected options and search engine algorithms. 


The time it takes for participants to be matched (or for a search to return results) can influence the perceived value of the available options. We recommend optimising for perceived value rather than just speed and performance. This can be counterintuitive since it doesn’t always mean the platform should provide instantaneous results. In many B2B contexts, asynchronous matching results can be correlated with perception of value. This is because participants associate the waiting with ‘work’ from the platform. 


Designing search that yields optimal results

In order to provide relevant, filtered and timely results, platforms need to develop efficient search functions. Robust search functionality is one of the strongest tools that platforms can use to maximise the effectiveness of their matching. There are three main aspects to search: 

  1. Recall – which participants or options are included in the search
  2. Sort – how results are ranked (search algorithm) 
  3. Display – how results are presented (format/user interface/user experience) 


These aspects can all be leveraged so that your platform’s matching returns results that are relevant, filtered and timely, which we already know are critical for platform success.


With recall, a minimum relevance threshold can help to narrow the search results. This could be pre-filtered results to include options above a set of minimum threshold criteria, e.g. keyword(s), exclusion criteria and/or to remove results below the threshold, such as those with a low score or red-flagged. For example, sellers with one star feedback score could be excluded from the recall.


If search algorithms can be very simple at the start, they ideally need to be regularly updated based on A/B testing and on evolving relevance criteria. Longer term, AI powered algorithms allow platforms to keep high matching efficiency. Here, it’s important to communicate to participants the principles behind the algorithm without going into the (ever-changing) details.


The display of search results should be focused on clarity and consistency. Matching and search results need to be perceived by participants as consistent, even if displayed in different formats (list, map) or delivered across different mediums like desktop, email, or mobile. Experience suggests that the best results are achieved when the distinction between sponsored and organic results is clear, for example, on eBay or Google. We also advise platforms to maintain multichannel consistency (desktop, mobile, email, voice) and enable alerts when relevant.


Preliminary questions to define matching criteria

Matching is a platform-specific activity, and key questions need to be addressed at the platform design stage. When beginning to craft a platform strategy, we use the following prompts to help define the matching/filter criteria:


  1. How will the matching be done?
  2. What are the key matching criteria?
  3. Will the matching rely on structured and formatted content or free-flow content? Will the function be automated or manually done by the platform owner (in the early days)?
  4. To what extent will people self-select? For example, are people expected to apply filters themselves to access relevant content?


Matching should ultimately be designed to grow the ecosystem in a balanced way and be self improving. In order for the value contribution of the platform to be maximised, matching should take into account the needs of the ecosystem as a whole. If matching is too biased in favour of one side, the platform becomes more ‘linear’ and risks losing participants, which in turn may undermine positive network effects.



  • The Match stage covers the processes by which platform participants will be introduced to each other, which is critical to the success of the platform.
  • The goal of matching is to present participants with the optimal level of choice that will lead to a successful transaction.
  • Results should be relevant, filtered and timely in order for the matching to be effective.
  • Efficient search uses recall, sort and display to ensure that results meet the criteria for effective matching.
  • Matching should be designed to grow the ecosystem – this will encourage positive network effects.


Next, we’ll look at the Connect stage of the rocket model and how this function expands on successful matching and enables transactions.

This article has been adapted from Platform Strategy: How to unlock the power of communities and networks to grow your business. Order your copy here.

Share this page

Subscribe to our newsletter to keep up-to-date with the latest industry insights.


3 comments on “The Rocket Model: Match

Comments are closed.