Social media and search engines in Africa all come from elsewhere and look like fixed points in the landscape. Two stories this week unsettle that fixed picture. A Tunisian search engine called Ayn (see Social Media) is betting on local knowledge succeeding. The second story is about Rancard’s new social recommendations engine Rendevous that helps produce better response rates. Russell Southwood spoke to Rancard’s Kofi Dadzie about how Rendevous works.
I’ve known Ghanaian Kofie Dadzie almost as long as I’ve been working in Africa. He started with ERP software before making the move to run mobile advertising campaigns for a blue chip list of clients and successfully attracted external investment. These were largely SMS-based, VAS content which went one way: from the brand advertising to the phone user.
The decision to shift focus again started in 2009 when its Chair Patrick Awuah (the education entrepreneur who launched and runs the private education institution, Ashesi University) talked to the Board about obsolescence.
This was at a time, which Dadzie admits, was when the company was just getting comfortable after several years of struggling. The result was a new executive strategy which asked the question: what’s the next phase of growth in the mobile content distribution business?
The short answer was aggregation and discovery:”There was just too much spam. Everybody was broadcasting everything to everybody. We needed to improve relevance by placing things in front of people that were relevant but that they may not have thought of.”
The key to doing this was to find a way of going beyond what the current search recommendation engines do. For example, if you’ve already bought one kind of item, then you’re likely to buy it again. Also you might buy items that are similar to it.
The idea with Rendevous is that you can expand the area of social recommendation if you look at how people respond positively to things like peer recommendation:”From an engineering standpoint, the question was how to do it? You need to have a social graph mapping connections between people based on decisions of a retail oriented nature”.
“In an offline example, a wife walks into a shop and sees a short and thinks my husband would like this. In an online example, the wife may want to re-upholster some furniture. She goes on to What’s App and asks for a good person to do re-upholstering. How do we automate this?”
“We construct a large social graph – the Rendevous graph – which shows connected users and shared interests. We then have a piece of software to read the graph and make predictions using maths and algorithms. In effect, the recommending algorithm walks through the social graph and makes recommendations. The graph needs to be at a certain depth and scale to make accurate predictions. It’s currently relatively limited but it’s growing very fast.”
So far, so theoretical. How does this work in practice? In fact, Rancard, is already actively engaging with clients in Nigeria and Ghana to put Rendezvous, its social recommendation engine to good use. Some of the clients using its technology include large financial institutions, portals and FMCG majors through their advertising agencies.
In a scenario where a bank brand is running a campaign to increase app installs, with a Cost Per Install (CPI) key metric, minimizing the cost per acquisition (install) is an important goal to the brand, while ensuring that a relevant audience is harvested.
The Rendezvous Campaign Manager tool employs a multi-phase approach by delivering an ad campaign on “external media” to seed the target users, then deploys its viral programming to grow the graph of connected users who share interests related to products/services served in the campaign.
Ultimately, targeting of possible customers to reach (such as to install an app) is based on the trust lines between connected users in the Rendezvous graph and their shared interests. Thus, one friend with shared interests in jogging and exercising will not see recommendations from their other friends’ activities where they’re interests do not align – Rendezvous increases relevance with social proofing while reducing spam.
Rendezvous composes its social graph of connected users and shared interests over the lifetime of interacting users in various 3rd party apps and portals.
FirstBank, through its advertising agency, Yellow Brick Road, decided to leverage Rancard’s Rendezvous social recommendations engine http://www.rancard.com/index.php?id=76 to improve their HNI audience engagement and targeting with relevant banking products.
As a cloud-based social recommendation engine Rendevous served recommendations based on connected users and their shared interests, with trust (contextual) and familiarity as the way of identifying potential users. Through doing this, Dadzie said it boosted conversion rates over typical mobile ad campaigns by approximately four times while continuously minimizing cost per acquisition (CPA) over the lifetime of the campaign. The aim with FirstBank was to get them 35,000 app installs, 60% of which were achieved in the first 30 days.
The viral programming (VIP) stage of the app install campaign was designed to boost user-to- user item sharing from existing installs of the FirstBank Loop app, which yielded superior click- through rates (CTR) averaging 13% over the period compared to < 3% for mainstream ads served on digital platforms.
Furthermore, FirstBank now has a Rendezvous social sub-graph of 35,000+ connected users (with a total of > 3.5 million friends) that it can continue to engage with other financial products. Currently 80% of the users on the Rendevous social graph are in Nigeria and 20% in Ghana with around what Dadzie describes as 10 million “nodes”.
It also has a partnership with Integrat which handles advertising sales for a number of mobile operators including MTN, the aim being to drive higher advertising yields for brands and advertising agencies.