The primary success criterion is whether the individual donor identifying with our profile feels that the portfolio of projects covered by the service is more effective than the result of each donor acting independently. We broke this requirement into individual indicators:
- Is the overall outcome more satisfying than an ad-hoc, advertising-driven, individual portfolio?
- Did the discovery method produce a better selection of work than passive individual efforts or a basic search via search engine or existing directory?
- Are donors willing to give more money than before?
Secondary success criteria were:
- To deliver a portfolio of grants driven by our donors’ quantitative preferences
- To deliver a number of grants not realistically attainable by an individual donor
- To deliver a portfolio of grants covering all categories in our scope
- To include projects that would normally find it difficult to reach our donors individually
- To support charities of different sizes, and to support small charities with repeated grants
- To include a geographical spread of projects both in the UK and worldwide
Here we provide a summary of the service outcomes, before sharing details of our experiences in each of the solution areas.
All of the success criteria listed above were met to our satisfaction, with the following achievements:
- £25,971 raised and distributed in line with expression data
- 78 grants made, ranging from £100 to £1000
- 5 regular and 2 occasional donors, all personally known to the charity’s board, all with registered expression data
- 9 years of accepting donations (2010-2018)
- 10 categories of charitable cause
- Geographical coverage of England, Scotland and Wales,1Projects were selected so that every location in Britain has at least 1-2 funded projects based within 50 miles. and causes abroad covering 6 continents
62p was the average monthly donation per grant, or £5.52 if every grant had recurred annually.2Donation per month per donor per grant, with £25,971 over 108 months, 5 donors, and 78 grants (or 8.7 annually-recurring grants). This shows that the portfolio’s breadth could not have been achieved without bringing the donors together. These figures depend primarily on the number of donors, average donation and average grant size, but in general will decrease as more donors become involved – it would take just 2,500 median donors to reduce the larger figure to 1p (though we would expect to increase the average grant size in practice). We note that our group gave at a rate about double the national median, though this has little significance for our conclusions.
The service ran at low financial cost covered by the directors. The main overheads at this small scale were accountancy, postage, and web hosting costs.
Our categories involved varying degrees of nuance in interpretation. Some broader categories warranted detailed internal discussion, which we outline here.
For health projects, there are not just different types of ailment but very different types of response and long-term strategies, and making some choice between these is unavoidable. Taking responses to cancer as a familiar example, we find projects at every point of the scale from prevention to cure to palliative care to family and bereavement counselling, and providing everything from research to direct action at each point. For our scope we wanted to strike a balance between direct action, positive long-term strategy, and representing areas that might be undiscovered or forgotten by individual donors.
Education was a hard category to interpret widely, especially after we took the decision to treat school-building projects abroad under the Development category. While in developing countries there are many projects to build and run schools relying on charitable assistance, domestically the need is harder to interpret, with basic education provided to all and funded solely by government. We found promising projects for vulnerable and underprivileged groups and rehabilitating offenders, but beyond this it was problematic to find appropriate activities because of the contrast with educational work in developing countries. Education scored highly in donor preferences, and we do not think this is an accurate reflection of domestic needs outside of general schooling, so for this category in particular defining the scope earlier could have avoided confusing donors.
The addiction and abuse categories we presumed to carry strong primary interpretations of drug and alcohol addiction and domestic abuse. We felt that this dominance was not problematic at this time, but did not intend to neglect other types of addiction or abuse.
An extra complication in the above was that child abuse was another dominant interpretation, but we had a dedicated category for children. This was not the only complication involving children’s projects – for almost any problem, we found projects specialising in children suffering from that problem. This was particularly apparent when we received applications, when a large number of projects involved holidays for sick children and their families. We considered this to be a market distortion driven by biases in selection processes, and it was a factor in our decision to look for alternative discovery methods.
The elderly category carried a complication in that it mainly comprises a collection of ailments suffered disproportionately but not uniquely nor universally by the elderly, including reduced health and physical capability, increased isolation, and poverty. Looking into the work done by existing grant-makers for the elderly showed that our characterisation of the category was in line with them, and we believe that donors would not be surprised by the projects involved.
We consider it unambiguous that some current and major environmental issues are polarising, which means that they are not considered in our scope, and a donor should continue to use alternative methods to support them. Our decision has been to prioritise and preserve a simple interpretation of charity, and make it widely acceptable. We were still able to accept several environmental projects into our portfolio, but found that candidate projects were more likely than those in other categories to state strong political positions alongside their work, which often meant that we needed to discount them.
In the final version of our charity categorisation, each category is focused on UK-based projects, with the exception of Development (exclusively for projects abroad) and Crisis Relief (about 10% UK-based). In total, 70% of funds were allocated to projects in the UK.
Some donors reported finding it difficult or unintuitive initially to order the project categories, but all returned a form that was complete and comprehensible. Some donors omitted some categories, some used the same number across multiple categories, and some reduced the set of numbers to produce a coarser set of relative preferences (e.g. using only 1, 4 and 9, or only 1 and 2).
While very broad and open to different interpretations, this gave us a basic set of quantitative preference data. We looked at the effect of different mappings from these figures to grant-budgets, agreed reasonable criteria and found a mapping to satisfy them. This ensured a significant difference between different rankings while leaving no category barely funded – the least popular categories received 2.9% of the total budget, while the most popular received 16.9%.
Though we considered alternatives, we ultimately chose to interpret the preference figures linearly. So, a category scoring a 9 from you would be allocated 9 times the budget as a category scoring 1. This served as a reasonable interpretation of the data for our small scale.
Following common sense, a budget was calculated separately for each donor, and then aggregated, so that each donor’s preference figures only affected their own portion of the total budget.
For donors who had not specified preferences, the budget allocation was at our discretion. We decided to use an aggregated set of preferences from all donors to allocate these funds in a similar way, except that this aggregated set was not weighted by the size of each donation, i.e. we gave each donor an equal say in how these unrestricted funds were used. In this we also included preference data from non-donating members, who wanted to have their views taken into account as appropriate while either getting to know the service or unable to donate for any reason.
Our initial approach of inviting applications from charities produced a selection of eligible projects from charities that we were previously unfamiliar with. As mentioned above, it also produced an undesirable clustering of projects so that most areas were either over- or under-represented, which did not get us far towards the goal of achieving good coverage of the sector. We needed a method that could both find projects based on quite specific criteria, and provide assurance that we weren’t forgetting about significant problem areas.
The database contains around 150,000 active charities. Taking a random sample of these, it was clear that a large proportion would not suit our needs. There are several common reasons for this, including that we cannot find detailed information about what the charity does, or there is a lack of evidence of recent financial or project activity, or that the charity only serves niche or particularly local interests and so is outside of our scope (for instance, a high proportion of registered charities are schools and churches). We were able to filter effectively to mitigate these phenomena, so that a high proportion of a filtered list were charities that merited active examination and discussion by the board.
Database filtering and sampling produced good shortlists in each area except Education. We put this down to a lack of definition both on our part and in general, for the same reasons as discussed under donor presentation. We took the approach of including a range of projects working with low-income, under-privileged or at-risk groups, rehabilitating offenders, and extra-curricular projects overlapping with other categories. For our overseas categories there was noticeably less variation in type of projects, but quite enough for the scale of our portfolio.
By sampling the whole database, we were also able to check whether our filters were leaving areas unrepresented. We had been concerned initially that we might find project areas that didn’t fit well into our general-or-local categorisation – perhaps because of a wide variation by geographical region – but we did not find strong evidence of this. We also ran searches across search engines and charity news and aggregation websites to find accidental omissions. Following these, we are confident that our database method achieves a high level of coverage.
It was clear early on that positioning charitable activities intentionally to fit categories is common. We saw applications covering multiple categories (e.g. helping only disabled children from ethnic minorities), in volumes that were intuitively skewed compared to incidence of the complex complaints in the population. With the database approach, we saw charities that had registered under additional categories that appeared to be tenuously connected to their work, and some charities that had ticked every box possible. In both cases it was clear that these practices reduced our shortlist quality by adding less relevant projects to our search results; we handled this by applying filters for the more extreme cases, and by factoring it into the board’s considerations.
Some projects were difficult to categorise – should a project educating school children about eating healthily be placed under Health, Children, or Education, or should the three categories take shared responsibility for it? We handled such cases individually through board-level discussion; at scale, we might resolve this problem by defining the categories, their responsibilities, and any mechanism for shared responsibility more clearly.
For the earlier discovery model where we solicited grant applications, we maintained lists of applications received in each category and would work from these when a sufficient budget became available. Significant trustee discretion was required in order to choose grant recipients, and imbalances in the numbers of applications received for each category made it difficult to construct or foresee a portfolio with sufficient coverage, which was the primary driver for us looking for more active discovery methods.
For the later model, some of the filtered lists contained thousands of charities, so we used random sampling to reduce the number under consideration without introducing bias. Different filtering was used for different categories, and for each charity making it past these a manual due diligence step was applied – this involved sanity checks of information on the charity’s website, and following up any causes for concern; a charity was removed from the list at this stage if there was either a lack of necessary information or the information provided raised serious concerns not accounted for elsewhere on the website. In a small number of cases where we felt the project was otherwise compelling, we contacted the charity with questions.
Once we moved to this method, the quality and appropriateness of projects reviewed by the board was consistently high, allowing time for deeper investigation and supporting a higher level discussion around how we could optimise the portfolio. By the time we sent out a grant offer, our level of confidence in the recipient was high.
Being contacted out of the blue with a grant offer was a new experience for many of our recipients, and was frequently seen by them as a risk. Several grant offers went unanswered, and several of those responding employed safety measures, including asking for assurance that the offer was genuine, and requesting payment by cheque to avoid giving out bank account numbers.
A minority of respondents were unable to agree to the terms we set out. Our standard terms were that the full amount of the grant be allocated to the specified project, with daily administration costs included, but fundraising and marketing activities excluded. The two reasons for charities being unable to comply with this were that their accounting procedures or policies would not allow the restriction (we only encountered this with large, household names), or that they operated as grant-providers themselves and could not guarantee that their own recipients would meet these criteria.
Problems aside, the speed of reply, tone, and helpfulness in the responses received all met a high standard. This remained true when our offering was a very small fraction of the recipient’s overall income.
With the small grant sizes we offered, we wanted to minimise the processing cost for our recipients. This was one driver for moving away from an application-based model. While it would have been difficult to do otherwise at our small scale, we ensured that no grant was a significant proportion of the recipient’s total income; at a scale where this no longer held, some consideration would need to be given to the longer-term outlook of the recipient, and a model centred around one-off grants would not be an appropriate approach. In anticipation of this, we decided to renew a small number of grant offers annually where a project continued to operate, subject to a summary report. Again to reduce cost, we limited this report to a basic confirmation that the grant had been spent as planned, and that the project had not changed substantially.
References [ + ]
|1.||↑||Projects were selected so that every location in Britain has at least 1-2 funded projects based within 50 miles.|
|2.||↑||Donation per month per donor per grant, with £25,971 over 108 months, 5 donors, and 78 grants (or 8.7 annually-recurring grants).|