The Challenge of Inactive Customers
Upcoming SlideShare
Loading in...5
×
 

Like this? Share it with your network

Share

The Challenge of Inactive Customers

le

  • 3,626 vues

CGAP has conducted quantitative research on the challenge of inactive customers in branchless banking. In culmination of this research, we have released this report that helps providers understand and ...

CGAP has conducted quantitative research on the challenge of inactive customers in branchless banking. In culmination of this research, we have released this report that helps providers understand and develop strategies to address low customer activity in their services.

Statistiques

Vues

Total des vues
3,626
Vues sur SlideShare
3,625
Vues externes
1

Actions

J'aime
2
Téléchargements
163
Commentaires
0

1 intégré 1

https://twitter.com 1

Accessibilité

Catégories

Détails de l'import

Uploaded via as Adobe PDF

Droits d'utilisation

© Tous droits réservés

Report content

Signalé comme inapproprié Signaler comme inapproprié
Signaler comme inapproprié

Indiquez la raison pour laquelle vous avez signalé cette présentation comme n'étant pas appropriée.

Annuler
  • Full Name Full Name Comment goes here.
    Êtes-vous sûr de vouloir
    Votre message apparaîtra ici
    Processing...
Poster un commentaire
Modifier votre commentaire

The Challenge of Inactive Customers Presentation Transcript

  • 1. The Challenge of Inactive Customers:Using Data Analytics to Understand andTackle Low Customer ActivityClaudia McKayToru MinoPaola de Baldomero Zazo February 2012
  • 2. While the number of branchless banking services has grown rapidly, thevast majority of registered customers are not actively transacting Most branchless banking providers struggle with high rates of inactive customers 140 120 120 100 80 While 120 branchless banking 60 implementations have 1 in 6 been launched since 1 in 11 2007 only 11 of those 40 21 have reached 200k 20 11 active users 0 Implementations launched after Implementations with >1 million "Confirmed" implementations 2007 registered with >200k active users In a CGAP survey, 64% of managers said less than 30% of their registered customers were active, and active rates of less than 10% are not uncommon 2 Source: CGAP and Coffey International, data as of Q1 2012.
  • 3. It is challenging to develop a viable business model with low activity ratesLow activity rates sharply increase the cost a If acquisition costs per active customer are thisprovider must invest to acquire each ACTIVE high, even in a best case revenue per customercustomer scenario it could take 10 years to breakeven CGAP estimates most services spend between USD 2-5 to register each customer. With 5% activity rate and a $5 per Assuming M-PESA Kenya revenue per customer as a “best customer acquisition cost, a service must invest $100 to case”, with an activity rate of just 5%, a service would take acquire each active customer 10 years to break even on the customer acquisition cost Acquisition Cost per Active Customer Time to Break-even per Active Customer$120 12$100 10 $80 8 years $60 6 $40 4 $20 2 $0 0 50% Active 15% Active 5% Active 50% Active 15% Active 5% ActiveWith current activity rates, most branchless banking businesses cannot generate enough revenueper customer to remain viable 3
  • 4. CGAP’s work on Inactive Customers• Moving a potential customer from awareness of a branchless banking service to regular use of the service requires different levers all to be working effectively and in an integrated manner: the agent network, product features, marketing, customer service, user experience and system/network. CGAP has developed a framework to map this process.• CGAP and others have produced publicly available materials on several of these critical issues such as agent networks and marketing.• One of the missing gaps in tackling this challenge is helping providers understand their customers and to identify the critical issues to resolve in their own services.• Ultimately, understanding customers is a complex undertaking and will rely heavily on a variety of tools such as focus groups and surveys. However, the first step for every provider should be to conduct data analysis on the veritable gold mine already at hand - its own database. CGAP has worked with four providers to understand how basic transaction level data can be used to shed light on customer activity. 4
  • 5. Who should use this deck Funders & Other Supporting Branchless Banking Providers OrganizationsThis deck is primarily targeted towards Funders and other supportingbranchless banking providers and aims to organizations in the financialhelp them use their data to better inclusion/branchless banking space mayunderstand their customers and the also benefit from this deck’s attempt toreasons for low activity levels in their identify the types of indicators and analysisservices. that shed light on customer usage and value. This may help funders standardize performance metrics across multiple organizations. 5
  • 6. How to use this deckThis deck does not provide answers on how to improve activity at any specific provider. Instead weidentify: (1) which data providers should collect about their customers and services, (2) what types ofanalysis providers can conduct based on the data collected, and (3) what kind of follow-up actionsproviders can take to further understand causes of customer inactivity in their own service. 1. Collect 2. Analyze 3. Act This deck helps providers This deck also identifies ways This deck does not provide understand what data they of analyzing data about specific answers to improve should be collecting and which branchless banking services activity as we found that the indicators they should be that we found gave the clearest influence of indicators on tracking. CGAP studied picture of how services are activity are not automatically hundreds of different variables being used and which factors transferrable across markets. and only included those are affecting activity rates. Instead we spotlight interesting variables that had statistically customer segments and significant impacts on activity behavior patterns that can be levels in this deck. followed up by providers through qualitative research and interaction with customers. 6
  • 7. Research introduction: quantitative analysis of data from 4 providersacross 3 regionsGeography: 2 South Asian 1 African 1 Southeast Asian Total customers: ~3 millionProvider types: 1 Bank-led 2 MNO-led 1 3rd Party-led • All services offer a mobile wallet focused on P2P transfers and bill paymentsSelection Criteria: • In the market at least 18 months so we could track trends over time • Perceive low customer activity as a challenge (though we did not deliberately seek out providers with abnormally low activity rates) 7
  • 8. Research introduction: methodology of analysis Data utilized Hypotheses guiding our Types of quantitative analysis analysis conducted Demographic segments Analysis of activity Customer influence activity rates trends registrations and demographics Regression analysis (probability of activity as Usage pattern segments dependent variable) help explain activityProviders Customer Cross analysis of usage transactions (at pattern and demographic least 6 months) segments Agents affect activity rate of customers they register Analysis of Money transfer patterns Agent registrations and Service usage in the 1st Analysis of agent demographics month affects ongoing demographics and activity registrations 8
  • 9. Outline of Deck1. General Activity Slides 11 – 15: Analysis of overall activity rate trends Trends2. Customer segments Slides 16 – 22: Interesting findings about customer segments and by usage patterns usage of services3. Factors Affecting Slides 23 - 29: Factors which were found to have a statistically Activity significant influence on activity rates Slides 30 – 34: Based on our analysis across these 4 example4. Takeaways & Action providers we can recommend certain indicators and analysis that Items providers should be tracking to help them understand and tackle the problem of low customer activity 9
  • 10. Main Findings on Customer Activity1. General Activity rates are low across all providers: average activity rate was only 8% and Activity 54% of all registered customers had never tried the service. We defined activity Trends as at least 1 transaction in the last 90 days. Each service has super-users responsible for a out-size proportion of total transactions. On average, 5% of users were responsible for 30% of total value2. Customer transacted. Providers should study these users to better understand the value proposition of their service. segments by usage patterns 48% of transfers happened locally within a municipality or province, indicating that P2P transfer patterns do not always follow the “send money home” pattern that was seen in M-PESA Kenya. The activity rate of customers registered by the best agents was over 40 times higher than those registered by the worst agents, so providers must learn from the best agents and intervene with the worst.3. Factors Affecting Activity Customer usage of a service in their first month after registration largely dictates their future activity, so providers should focus efforts on getting customers to not just transact in the first month, but to do specific types of transactions. 10
  • 11. Main Findings: General Activity Trends1. General Average activity rate across providers was just 8%, and activity rates have Activity Trends not increased much even as registrations accelerate2. Customer 80% of customers who are active in their first month after registration have segments by stopped transacting by their 4th month, so new subscriber growth masks a usage real drop in activity rates over time patterns3. Factors To more accurately track activity rate trends over time we recommend using Affecting Activity vintage analysis4. Takeaways & Action Items 1.trends 2.usage 3.activity 4.takeaway 11
  • 12. Activity rates are low across all 4 providers and have not grown along with customer registrationsOnly 8% of registered customers were active Cumulatively, activity rates have remainedacross the 4 providers we studied relatively flat while registrations have grown rapidly* 8% 1,000 50% Only 8% of registered 900 45% customers were active 800 40% [range: 1% to 18%] 700 35% 38% 600 30% 500 25% 38% of registered customers had once 400 20% been active but are now 300 15% dormant 200 10% [range: 12% to 68%] 100 5% 0 0% 54% 54% of registered customers had signed up but never transacted Active customers Registered customers Activity Rate [range: 25% to 87%] *trend numbers are cumulative across 3 providers Notes: • These are straight averages across all 4 due to large differences in numbers of registered customers across 1.trends 2.usage 3.activity 4.takeaway 12 providers
  • 13. The frequency of activity even among active customers is lowAt one provider 63% of all active customers averaged less than 2 transactions a month, thoughthere is a small cadre of highly active customers 63% of all active customers averaged less than 2 transactions a month Top 5% of customers averaged over 54 transactions per month 1.trends 2.usage 3.activity 4.takeaway 13
  • 14. Customer activity drops the longer they are with a service, so aggregateactivity rates mask a real drop in activity over time80% of customers who are active in their first Even if aggregate activity rates are stayingmonth after registration become dormant by flat, new registrations could be masking atheir 4th month real drop in activity rates % of customers transacting, If registrations 1st month – 4th month* stopped today…20% 900 25% Thousands18% 80016% 700 20% Overall activity14% 600 rate would fall 15%12% 50010% 400 10%8% 3006% 200 5%4% 1002% 0 0%0% 1 2 3 4 5 6 7 8 9 M1 M2 M3 M4 Registered customers Activity Rate *data aggregated from 3 providers 1.trends 2.usage 3.activity 4.takeaway 14
  • 15. Vintage analysis is a more accurate way to measure activity rate trendsas it reduces masking effects of new registrationsVintage analysis looks at whether customers Vintage analysis provides a more accurateare transacting 3 months AFTER registration, picture of activity rate over time, as it reduceseliminating masking effect of new customers noise from new registrations Overall activity rates allow new registrations to This spike in activity is mask drops in individual activity over time due to a registration 40% promotion Standard Activity rate trend line 35% 30% 25% Vintage analysis shows that the 1 month activity 20% spike did not actually Time lead to greater activity 15% rates 3 months later Vintage analysis looks at activity 3 months after 10% sign-up to give a more accurate trend view 5% 0% Activity rate Oct-10 Dec-10 Jun-10 Jul-10 Jan-11 Feb-11 Feb-10 Mar-10 Apr-10 Mar-11 Apr-11 Aug-10 Sep-10 Nov-10 May-10 May-11 Vintage trend line % transacted in 3rd month (vintage) % transacted in 1st month Time Note: data from one provider 1.trends 2.usage 3.activity 4.takeaway 15
  • 16. Main Findings: Customer segments by usage patterns Each service has super-users responsible for a out-size proportion of total1. General transactions. On average, 5% of users were responsible for 30% of total value Activity Trends transacted. Providers should study these users to better understand the value proposition of their service.2. Customer segments by 44% of all active users only performed one type of transaction on the service, despite a range of available transactions. This implies that providers should usage customize their marketing messages to different user segments. patterns3. Factors 48% of transfers happened locally within a municipality or province, indicating that Affecting P2P transfer patterns do not always follow the “send money home” pattern that Activity was seen in M-PESA Kenya.4. Takeaways & Action Items 1.trends 2.usage 3.activity 4.takeaway 16
  • 17. Studying “super-users” can help providers understand the real valueproposition of their serviceSuper-users included only the top 5% of customers by total value of transactions, but this 5%was responsible for over 30% of the total value transacted on the service* We defined “super-users” as the top 5% of customers by total value of their The average Avg value of transactions over the last 3 months monthly value of transactions for super $242.63 transactions for users 5%: super users was 6.5 times the average Avg value of Super-users value of transactions transactions for all $37.64 users across all users Super-users make Total value transacted on 95%: up only 5% of total service Everyone else customers but represent more than 30% of total 30% value of transactions on a service Super-users*average across 3 providersTakeaway: Super-users are valuable in and of themselves as they represent a major part of thevalue of a service. Studying them can also help providers understand what value propositiontheir service really offers to customers. 1.trends 2.usage 3.activity 4.takeaway 17
  • 18. We do not recommend studying super-users based on the NUMBER oftransactions as that can give misleading results about usage of a serviceAnalyzing transaction patterns by NUMBER of transactions over-inflates the importance ofairtime top-up transactions for branchless banking systems and understates the true value ofP2P transfers* 100% Other transactions 80% Airtime top-ups P2P transfers (sending) 60% 40% 5% 46% Super users defined by NUMBER of transactions are 20% often using the service for unintended activities: 24% For 2 providers, super-users defined by NUMBER of 0% 5% transactions were doing hundreds or thousands of % of total value of % of total number of airtime purchases a month, and turned out to be transactions transactions individuals acting as unauthorized resellers of airtime*aggregated data from 2 providersTakeaway: When analyzing super-users, a definition based on VALUE of transactions is likely togive more accurate results than a definition based on NUMBER of transactions 1.trends 2.usage 3.activity 4.takeaway 18
  • 19. Providers should understand why particular customer demographics are morelikely to be super-users and consider specifically targeting these segments Customer occupations seemed to affect the Gender also seemed to have an effect on likelihood of being a super-user at one whether a customer became a super-user at provider two providers Customers identifying as self-employed were far Female customers were slightly LESS likely to more likely to be super users than average, be super-users than males while students were far less likely than average 19.0% 16.0% 14.7% 11.2% % of super users % of super-users 10.6% % of all active users % of all active users 5.5% Self-employed Student Females 1.trends 2.usage 3.activity 4.takeaway 19
  • 20. Low and high value users utilize branchless banking services for verydifferent purposesLooking at two different providers, low and high value usage patterns are very different, butdifferences across markets means each provider must analyze their own customer transactions Debits Cash-out Top-up Bill Payments Transfers sent Balance Inquiries Left in account Credits 100% 5% 12% 90% 1% 80% 32.3% Cash-in 29% 50% 70% Customer 60% Account/ 19% 13.3% 50% 97.2% Wallet 40% Transfers 30% 29% received 50% 48.5% 20% 11% 10% 9% 0% Low High Low High Provider 1 Provider 2 Usage patterns for these 2 services differ widely, with Between low and high users, the biggest difference Provider 1 seeing significant bill payments usage while comes from the value of transfers sent Provider 2 seemed to be used as a store of value Note: “Low” refers to the bottom third of active users by value of transactions and “high” refers to the top 1.trends 2.usage 3.activity 4.takeaway 20 third, excluding “super-users”
  • 21. Many users only perform one type of transaction out of a broader offering, indicating a variety of segments and use cases for services While all providers offered a mix of services, Certain demographic factors were tied to a large proportion of customers only these single transaction-type segments, performed one type of transaction which has implications for marketing efforts Across 2 providers, almost 50% of users At one provider, those customers who only only performed one type of mobile money received money were 47% more likely to be transaction female than average 9% Savers† 8% 7% Top-up 6% Receivers‡ 5% 4% Senders‡ 3% 2% Variety of transctions 1% 0% Females as % of "receivers- Females as % of total only" customers†Defined as those who have not done any transactions other than deposits and withdrawals‡We did not exclude senders or receivers who had also used top-up 1.trends 2.usage 3.activity 4.takeaway 21
  • 22. P2P Transfer Patterns: Transfers do not always follow a standard “sendmoney home” pattern, with almost half of all transfers happening locallyAlmost half of the total value transferred was Looking at the number of transfers, the topsent within the same region when averaging sending cities are also receiving a majoracross 2 of the providers portion of the transfers 51.8% 48.2% 1 2 3 4 5 Top sending regions Transfers Inter-region Transfers Intra-region Transfers sent Transfers receivedTakeaway: Providers should not simply default to the “send money home” messaging andstrategy that worked for M-PESA Kenya as it may not fit in many markets 1.trends 2.usage 3.activity 4.takeaway 22
  • 23. Main Findings: Factors Affecting Activity Rates1. General Customer demographic differences are highly correlated with differences in Activity Trends activity rates, so providers should analyze demographic data to learn from these segments.2. Customer segments by The activity rate of customers registered by the best agents was over 40 usage times higher than those registered by the worst agents, so providers must patterns learn from the best agents and intervene with the worst. Customer usage of a service in their first month after registration largely 3. Factors dictates their future activity, so providers should focus efforts on gettingAffecting Activity customers to not just transact in the first month, but to do specific types of transactions.4. Takeaways & For MNOs with mobile money services, high value voice/SMS/data customers Action Items also tend to be higher value mobile money customers. 1.trends 2.usage 3.activity 4.takeaway 23
  • 24. Demographic Effects: Customer demographics can have significanteffects on activity, so providers should collect and act on this data (1)Gender: averaged across 3 providers, female Age: age of customers had statisticallycustomers were 41% more likely to be active significant effects on activity rates, but thethan males results were not consistent across providers Active Dormant Never Transacted Provider 1 Provider 2 At provider 1 younger 14.00% customers were more active 100% while at provider 2 older 90% 12.00% customers were more active 12.38% 80% Average activity rate 45% 43% 10.00% 70% 9.87% 60% 8.00% 50% 7.5% 6.00% 40% 5.5% 30% 4.00% 20% 2.00% 10% 15% 11% 0% 0.00% Female Male Youngest Quartile Oldest Quartile Females were more likely to be active, but also Because demographics affect activity differently slightly more likely to have never transacted across different market contexts, providers must collect and analyze their own data rather than rely on generalizations 1.trends 2.usage 3.activity 4.takeaway 24
  • 25. Demographic Effects: Customer demographics can have significanteffects on activity, so providers should collect and act on this data (2)Income level: At one provider, customers in Occupation: Activity rates for somethe lowest income bracket were 3 times as occupation segments was 3 times the overallactive as the wealthiest mean activity rate for the service* 90 day activity rate by Activity rate by stated occupation monthly income 35% 30% 16.9% 25% 20% Mean activity rate 15% 5.2% 10% 5% Lowest income Highest income 0%Only one provider collected data on income levels sowhile we encourage other providers to collect thisdata, we cannot make generalized claims on theimpact of income on activity *data from 1 provider 1.trends 2.usage 3.activity 4.takeaway 25
  • 26. Registration Agent Effects: Agents differ widely on the activity level ofcustomers they registerAcross 2 providers, the activity rate of customers registered by the best agents was over 40times higher than those registered by the worst agents Top 20% of agents Profile of customers All Agents by # of registrations registered by these agents Top 20% of Top 10% by agents by # of activity rate Activity Top 10% Customers registered registrations Rate by these agents make up 5.1% of total 39.9% customers Bottom 10% Activity Customers registered Rate by these agents make up 5.7% of total 0.9% customers Bottom 10% by activity rateTakeaway: Agent registration incentives should take into account not only the NUMBER ofcustomers registered but also the ongoing ACTIVITY of those customers. To do so providersmust monitor activity rates for each agents’ customers 1.trends 2.usage 3.activity 4.takeaway 26
  • 27. First Month Effects: Transaction behavior in the customer’s first monthsignificantly affects ongoing customer activity and valueThe number of transactions in a customer’s …but to get more VALUE from customersfirst month is highly correlated with their over time, the TYPE of transaction they areactivity in subsequent months… doing in their first month has a greater effect 35% 33% Likelihood of transacting in $2.00 Value of transactions 3rd month2 30% 25% 3rd month 20% $1.50 15% 12% 10% 4X $1.00 5% 3% 1% 0% 0 1-2 3-4 >=5 Number of transactions in first month* $0.50 Across 4 providers, customers who did 6+ transactions in their 1st month were 5.5 times $- more likely to transact in their 2nd month than Customers who only top- Customers who do other those who did only 1-2 transactions up first month transactions first month*data aggregated from 3 providers 2weighted avg across 2 providers 1.trends 2.usage 3.activity 4.takeaway 27
  • 28. First Month Effects: Customers rarely “graduate” over time to highervalue transactions making the first month even more importantAcross 2 providers, of the customers who did only cash-in + top-up first month, only 19% did atleast one other type of value transaction in their 3rd month, compared with 65% for those whotried a variety of transactions in their first month % of customers who do more than top-up in months 2-4 70% 60% 65% 64% 50% 51% 40% 30% 20% 20% 19% 19% 10% 0% 2nd month 3rd month 4th month Customers who only top-up first month Customers who do other transactions first monthTakeaway: Usage patterns in the first month largely determine future usage, which may meanproviders should focus more on getting new customers to try a variety of transactions 1.trends 2.usage 3.activity 4.takeaway 28
  • 29. Voice/SMS Usage Effects: How phone customers use voice and SMScould be a predictor of their activity and value in a mobile money serviceThe likelihood of a customer being an active Total value of customer mobile moneymobile money customer was correlated with transactions seemed to be even more closelytheir voice and especially SMS usage linked with voice/sms usage* At one provider, SMS usage in particular $6.00 Value of customer transactions(last 3 months) affected activity rates, with high SMS users 20% more likely to be active than low SMS users +21% +17% +20% $5.00 High SMS Usage $4.00 Low SMS Usage 0.00% 5.00% 10.00% 15.00% 20.00% $3.00 Activity Rate At another provider, consistently active mobile $2.00 money subscribers had a combined voice/sms ARPU (avg revenue per user) 20% higher than subscribers who had never transacted $1.00 Never transacted $- Consistently active Voice (minutes) Data (GPRS) SMS Low Usage High Usage $- $5.00 $10.00 $15.00 Voice/SMS combined ARPU *data from 1 provider only 1.trends 2.usage 3.activity 4.takeaway 29
  • 30. Takeaways and Action Items: Based on our analysis across these 4 example providers we can recommend1. General certain indicators and analysis that providers should be tracking to help them Activity Trends understand and tackle the problem of low customer activity In many cases providers are already collecting the requisite data and we are only suggesting ways of analyzing and acting on this data that may be new to some2. Customer providers segments by usage In other cases we are recommending providers collect new types of data that can help patterns them better understand, communicate with, and serve their customers3. Factors The following takeaways have 3 components: Collect, Analyze, and Act Affecting Activity Collect: Analyze: Act: Examples of the types of Types of data and Actions which can be analysis providers can specific metrics taken as a result of the conduct based on data4. Takeaways & providers should track data analysis collected Action Items 1.trends 2.usage 3.activity 4.takeaway 30
  • 31. Takeaways and Action Items: General activity trendsResponding to Activity TrendsCollect Analyze Act Vintage analysis: tracking not • Use vintage analysis to more only standard activity definitions accurately judge the Customer registration and but also how active customers are effectiveness of customer transaction data already collected by their 3rd month with the service recruitment efforts (including by most providers’ systems is (did customers do at least one new promotions or pricing) sufficient for this analysis transaction in their 3rd month) • Drops in vintage activity rates gives a clearer picture of activity are an early warning of problems trends over time with recruitment strategy For our analysis we defined 3 fairly • If most inactive customers are standard categories: dormant, this signals that • active [at least one transaction marketing & initial registrations in the last 90 days] are working, but the ongoing • dormant [transacted at least value proposition is broken. once but no transactions in the • If vice versa, then attention last 90 days] should be focused on better • never transacted customer recruitment 1.trends 2.usage 3.activity 4.takeaway 31
  • 32. Takeaways and Action Items: Customer Segments by Usage PatternsLearning from Super-usersCollect Analyze Act Compare demographics (age, • Set up interviews/focus groups gender, location, etc) against with a sample of super-users to inactive registered clients to understand why the find the Identify the top 5% of users by identify differences service so valuable VALUE of transactions using customer registration and • Adjust marketing to capture more transaction data Compare their usage pattern and people like them demographics against average users to identify characteristics • Recruit them as community which set super-users apart champions for your serviceUnderstanding the real value proposition for transfersCollect Analyze Act Conduct interviews/focus groups Identify top transfer corridors Try to find connections between to better understand what the top using existing customer and demographics and transfer purposes are for money transfers transaction databases behavior (ie, are students all on your service receiving long-distance transfers Additional demographics such as while small-business holders are Adjust marketing messages and/or customer occupation and sending and receiving many operations such as agent income could be very valuable transfers within their municipality?) recruitment based on reality of here transfer usage 1.trends 2.usage 3.activity 4.takeaway 32
  • 33. Takeaways and Action Items: Factors affecting activity (1 of 2)Monitoring effects of Agents on activityCollect Analyze Act Identify outlier agents who are • Study these agents to signing up many customers who understand why they have such Collect both the number of are INACTIVE or VERY ACTIVE low/high activity rates customers each agent registers in subsequent months and see AND the percentage of the • Take action with low performing what sets them apart from each agent’s customers who are agents if they do not improve other and from average agents active in later months and generously reward agents (using demographic data collected) with high active rates • Talk to these agents and understand what factors are Study those agents who are both increasing their activity (is it Collect any agent demographics signing up many customers who demographics or their possible such as type and size of are ACTIVE to see what sets them interactions with customers) business, location, age and apart from average agents (using experience of proprietor, etc • Take steps to spread any best demographic data collected) practices identified at these good agents to all other agents 1.trends 2.usage 3.activity 4.takeaway 33
  • 34. Takeaways and Action Items: Factors affecting activity (2 of 2)Understanding customer demographic effectsCollect Analyze Act Analyze activity rates across • Talk with members of these demographic factors and identify interesting groups to understand groups with higher/lower activity why the service works or does Collect customer demographic data such as gender, occupation, not work for their needs income level, age, etc • Determine which groups to target Identify interesting groups based on your service’s own data, not and revisit marketing and product based on patterns in other markets strategy where appropriateAdapting to the effects of customers’ first month transaction behaviorCollect Analyze Act Incentivize users to transact in first Customers’ first month month, focusing on transactions transaction behavior Using regression analysis identify that are most correlated with what number and type of higher usage later transactions in the first month are correlated with higher value future Customers’ transaction customers Incentivize agents to promote behavior in following months beneficial transaction behaviors in the first month for their customers 1.trends 2.usage 3.activity 4.takeaway 34
  • 35. Advancing financial access for the world’s poor www.cgap.org www.microfinancegateway.org