Research Process

How big is that museum?

What is a small museum? Or for that matter a medium or large museum? In the museum sector, size is generally measured in relation to visitor numbers, and in cases where several criteria are used, such as income or staff numbers, they are still taken into account. The Mapping Museums research team has followed suit in this respect. We decided to group the museums within our dataset into size categories that are based on visitor numbers. Thus the question for us was: how should we establish the thresholds for these categories? How many visitors equate to small, medium, and large? And should we just use those three categories? What about very tiny or really massive museums?

Arts organisations define size in slightly different ways, and in some cases, single organisations may use a variety of measures. For example, the Association of Independent Museums (AIM) uses the following categories in their ‘toolkit’:

  • Small = visitor numbers of up to 10,000
  • Medium = visitor numbers of 10,001 to 50,000
  • Large = visitor numbers of 50,001+

However, AIM uses slightly different bands when museums are applying for membership. In this case, the smallest category is defined as being up to 20,000, not 10,000, and there is an additional category of ‘largest museums’, which attract over 100,000 visitors. Arts Council England (ACE) data uses the same measures as the AIM toolkit, but only in relation to independent museums. When they assess the size of local authority museums, they use a different yardstick:

  • Band One: 30,000 visitors
  • Band Two 30-100,000 visitors
  • Band Three 100, 000+ visitors

These differences are sensitive to the realities of museum practice. By categorising museums that have less than 20,000 visitors as being small for the purposes of membership, AIM enables more organisations to pay the lower rates of subscription than if it had put the bar at 10,000 visitors. Similarly, ACE recognise that local authority and independent museums operate under different conditions. For the Mapping Museums team, however, the use of different size bands was problematic because it would be difficult to know how to categorise museums that have hybrid forms of governance, for instance, when local authorities retain ownership of museum buildings and collections but outsource its management. Using different size categories according to governance, also meant we would have to change size designations when museums changed status, and it also prevented any direct comparisons across categories of governance.

In the absence of an established rubric for museum size, we needed to decide what size bands to use in the Mapping Museums research. In order to first make that decision we looked to the data and at the overall spread of museums according to visitor numbers.

Figure 1- Distribution of Museum Yearly Visits
Figure 1

At this point, we had no visitor numbers for 45% of the museums in our dataset. However, when we plotted the information that was available to us (Figure 1), we could see that there was a clear peak in the data between 10,000 and 32,000 visits per year (with median about 13,000), but that there were no obvious points where the distribution of museums divided into bands. Thus the data did not suggest any clear categories for allocating size.

We then divided the distribution into quartiles, which showed that 50% of museums had between about 4,000 and 40,000 visitors per year (Figure 2).

Figure 2 - Distribution of Museum Yearly Visits with Quartiles
Figure 2

One option was to create a band that covered the broad group of museums that gain between 4,000 and 40,000 visitors. The problem was that approach would elide the significant differences in scale. A museum that gains 4,000 visitors per year is likely to be run solely by volunteers or by a private individual, with limited opening, and to operate in a relatively ad-hoc fashion. A museum that attracts 40,000 visitors is reasonably well established and likely to have a professional orientation. Thus, grouping these museums did not make sense in an analytic context.

First categorisation

Our next step was to consider how various size categories would support our research. One of the problems of using three bands for sizing is that it lacks nuance. A museum that has 100,000 visitors is clearly very popular and well established but it is not in the same league as one that has visitors in the millions, yet both would normally be classified as ‘large’. Thus we initially decided to introduce more categories (see Table 1).



Size category

Yearly visitor number range Number of museums (%)
Tiny 0-1,000 4.1
Very small 1,001-5,000 10.8
Small 5,001- 20,000 16.6
Medium 20,001 – 50,000 11.2
Large 50,001 – 100,000 5.6
Very Large 100,001 – 1million 6.1
Huge 1 million + 0.3
Unknown 45.3

Table 1

This approach initially seemed to work. However, when we began detailed analysis of the data the researchers found that they were constantly aggregating the three smallest categories. We did not need that degree of nuance for our work. However, we regularly used the category of ‘huge’ as a way of filtering out the very largest of institutions. Thus we decided to revert to a single category for small museums but to keep ‘huge’.

Second (and final) categorisation

Our second set of categories, which we are now using, reads: small, medium, large, and huge. Yet the question remained of where the thresholds would be set for each category. Again we turned to the data, and looked at how the distribution of museums would change if we used 10,000 or 20,000 visitors as the top limit for the category of small, and what difference it would make if 50,000 or 100,000 were used as the upper limit for ‘medium’.

  Museum counts
Category thresholds small medium large huge
0; 20k; 100k; 1m


1318 677 250 13
0; 10k; 100k; 1m 930 1065 250 13
0; 10k; 50k; 1m 930 839 476 13

Table 2

As Table 2 shows, splitting small and medium at 20,000 means that the former category is significantly larger than the later. Splitting the categories at 10,000 produces a more balanced distribution between the two. In both these figures the category of large has relatively few museums because it only includes organisations with over 100,000 visitors. When that threshold is dropped to 50,000, then the size of that category almost doubles.

Importantly, the different size categorisations give a very different impression of the UK museum sector. If small museums predominate then we might assume that the sector is dominated by museums that attract few visitors, are volunteer-run museums or have few paid staff, and that possibly are struggling to survive. In contrast, if there are larger numbers of museums of a medium size, then the sector seems to be more comfortably established, and, if there are high numbers of large museums, then onlookers may conclude that it is flourishing. Thus size categorisations can have a strong impact on perceptions of the sector, even if the actual visitor numbers and lived realities of museum practice remain the same throughout.

After considerable discussion the Mapping Museums team decided to set the size categories as follows (Table 3):

Size category Yearly visitor number range Number of museums (%)
Small 0–10,000 22.5
Medium 10,001–50,000 20.3
Large 50,001–1 million 11.5
Huge 1 million+ 0.3
Unknown 45.3

Table 3

For us, these categories chimed reasonably closely with norms of thinking about museum size, and are similar to those used by the AIM toolkit, which has the advantage of making them familiar within the sector. They lack nuance in the category of ‘large’, but this is not a particular issue for our research, as the focus of Mapping Museums is on smaller museums. Setting the bar at 10,000 also means that small museums do not merge into medium-sized, more established organisations, and we can examine them as a distinct group. For us, this is important because the smallest museums are often sidelined both in research and in professional discussions.

Copyright: Fiona Candlin and Andrea Ballatore, 2018

Image via S. Faric on Flickr

Research Process

Missing, massaged, and just wrong: Problems with visitor numbers

Visitor numbers provide some sense of the scale of a museum’s operations. If a museum has a large collection of priceless artefacts, occupies an impressive building, has professional curators and conservators, a nice café, and offers activities to its audiences, then it is unlikely to attract a mere 2,000 visitors per year. Conversely, if a museum is housed in a defunct railway station, with one retired locomotive on exhibition, and is staffed entirely by volunteers, then it would be surprising to discover that it gained millions of visitors. There is a link between a museum’s provision, and its visitor numbers. Thus by listing visitor numbers for the museums in our dataset, the Mapping Museums team intended to provide researchers with some guide as to the organisations’ size and character. However, this process was not as straightforward as it initially seemed.

One problem is that visitor numbers are not always available. Figures for larger institutions are reported in the national monitor Visit Britain. Information on attendance at accredited museums is published by Arts Council England, and the Museums Association usually includes visitor numbers on the Find-A-Museum service listings. Obviously, museums that are not accredited or are not members of the Museums Association do not appear on those sites. Unaccredited, unaffiliated museums may sometimes note their visitor numbers on their own website or annual report, but more often, that information cannot easily be found. Moreover, visitor numbers may not exist as such. Collecting that information requires staff capacity and resources that are beyond the reach of some organisations, and while the lack of documentation or the complete absence of data may indicate low visitor numbers, that correlation cannot be guaranteed.

Problems with visitor numbers are not confined to a lack of information. Even when visitor numbers do exist, they cannot be relied upon. One issue is that there is no accepted methodology for how visitor numbers are collected, and institutions each decide how to accomplish this task. In some instances, museums log everyone who comes through the door. However, if the museum or gallery has conveniently placed toilets, as was the case at Middlesbrough Museum of Art, then people coming to use the facilities raise the footfall. Cafes can similarly boost the total visitor count. Other museums only record the number of visitors who enter into a gallery or look at artwork, although those criteria can be met by putting artwork or displays into the foyer of a museum. It is also unclear whether people who participate in outreach or other activities are included in total numbers. We found one very small museum that reported 42,000 visitors because they organise an annual rally and included all the attendees. Who is doing the counting and how they count has a significant impact on the recorded visitor numbers.

Methodology aside, visitor numbers are sometimes actively massaged. Adrian Babbidge commented in a recent article for Cultural Trends, there are strategic reasons for inflating them and the Mapping Museums team found instances where disparate numbers had been reported. For instance, one museum stated that it had less than 20,000 visitors a year on its AIM membership forms yet claimed 30,000 visitors per year on a fund raising website. If its actual numbers were closer to 30,000 then by tweaking figures, the institution saved a little on membership fees, and if the lower number was more accurate, the upwardly adjusted figure might have improved their chances of raising money. The Mapping Museums team has also encountered cases where visitor numbers were purposefully deflated. At least one small museum had under-reported ticket sales to avoid paying tax on that income. This had the consequence of them appearing to have lower visitor numbers than is in fact the case.

Another set of difficulties obtain when dealing with historic visitor numbers. As we’ve noted before, the Mapping Museums team is documenting UK museums from 1960 until the present day. Where available we have recorded visitor numbers that pertain to that period, and most notably, we have included figures from the massive DOMUS survey that was run between 1994 and 1998. This has the advantage of providing size indicators for museums that have now closed but we have discovered that some of the DOMUS records are anomalous. For example, The Royal Electrical and Mechanical Engineers museum is listed as having the following audiences in successive years.

4,500 in 1994

20,000 in 1995

35,000 in 1996

5,000 in 1997

According to these figures, the number of visits increased eightfold in a two year period, and then reverted to its original numbers. This seemed unlikely so we contacted the museum. The director, Major Rick Henderson, told us that the museum had never attracted such high visitor numbers. Even now, with a dedicated staff and a new building, attendances are in the region of 20,000. It is therefore likely that the inflated figures are due to errors made when the data was entered into the DOMUS system. The problem is that we cannot check all the anomalies, partly because of time but mainly because many of the museums have since closed and the institutional memory lost.

Thus, there are several challenges to using visitor numbers to give a sense of the scale of a museum: it is difficult to find figures for unaccredited museums or they may never have been collected; there is no established methodology for collecting visitor numbers; museums massage audience numbers for strategic purposes; and historic records may be incorrect.

The Mapping Museums team decided to deal with these various issues by using categories for size rather than visitor numbers. Providing precise numbers may give the false impression that the figures all adhere to the same measure and can be compared, whereas categories provider a looser guide to a museums operations. Unfortunately, using size categories also has its complications, which I will outline in the next blog.

Copyright Fiona Candlin 2018