Offshoring Analytics: my personal experience of is it worth it?
Planning to be offshoring analytics to meet demand? Then I suggest you read this first.
There has been much on LinkedIn and Twitter in recent months about the shortfall in analytical resource, for the USA & UK markets especially.
Several years ago I had the learning experience of attempting offshoring analytics to India, Bangalore to be precise. Here are some lessons that I learnt as a result.
The excitement of offshoring analytics
It was all very exciting at first, flying out there and working with the team whilst visiting the UK. International experience for what had been a more UK focussed leadership role.
Plus, on paper, it looked a good idea to address the peaks and troughs of demand for analysis and modelling.
The pitch successfully communicated the ease of accessing highly trained Indian graduates at a fraction of UK wages. However, as with all software demos, the experience after purchase was a little different.
I always expected the model to take a while to bed down and you expect to give any new analysts time to get up to speed with our ways of working. However, after a few months the cracks began to show.
Coping with real data & the real world
Analysts in India were failing to understand what was required unless individual pieces of work were managed like mini-projects and requirements specified in great detail. There also appeared to be little ability to improvise or deal with “dirty data”, so significant data preparation was still required by my UK team (who were beginning to question the benefit).
Once the building of propensity modelling was attempted, a few months later, it became even more apparent that lack of domain knowledge and rote learning from textbooks caused problems in the real world.
Potential remedies to fix offshoring analytics
Several remedies were tried. Further visits to the UK, upgrading the members of the team to even more qualified analysts (as we started with graduates we were now working solely with those who held Masters degrees and in some cases PhDs).
Even after this, none of my Bangalore team were as able to “think on their feet” as well as any of my less qualified analysts in the UK. There were still no signs of domain knowledge (about insurance, our business, our customer types, previous insight learned etc) being retained.
Reasons to stop offshoring analytics
After 18 months and to some extent with a heavy heart (as all those I had worked with in Bangalore sincerely wanted to do the best job they could for you), I ended this pilot and instead recruited a few more analysts in the UK. Several factors drove the final decision but they included:
- Erosion on labour arbitrage (the most highly skilled cost more);
- Inefficiency (i.e. need for prep & guidance) impacting to UK team;
- Cost and effort to comply with data security requirements.
Since that time I have had a few customer insight leaders suggest that it is worth trying again (nowadays with China, Eastern Europe or South Africa), but I am not convinced.
On reflection my biggest concerns are around the importance of analysts understanding their domain (business/customers) and doing their own data preparation (as so much is learned from exploratory data analysis phase). The “project-ization” of analysis requests does not suit this craft.
Offshoring analytics – what is your perspective?
So, for me, the answer is no. For all the reasons outlined above & despite the positive attitude of offshore analysts.
Do you have any experience of trying this? Does it work for you or your business? If so, please let me know, I would love to share a different perspective, especially if you can offer advice as to how you got this working.
There are some consistent themes emerging Paul. Both this and the Org Design post rightly conclude that there are significant advantages in having analysts who work closely with the business. I’d take a competent analyst who “gets” the priorities of the business they’re supporting and can spot opportunities for value- add rather over a back office technical/statistical genius who is far removed from the coalface (and in some cases so far removed they’re in a different continent!).
Very true, James, it reminds me of the old recruitment adage (to recruit on attitude as skills can be learnt). Analysts who want to really engage with a business and understand customer problems are those with the attitudes needed in the right structure. Thanks for sharing.
Thanks for sharing, Paul.
Has anyone made off-shoring of insight/analytics work? I’m with you both James and Paul that the required relationship with the business managers and attitude beat statisical grunt. My stats is a bit rusty but my commercial rationale is a killer and vuts through days of modelling. I for 1 have been buolding and maintaining my own analytical team with “grow your own” style of bringing in thr right grads and training them up. Hopefully we get to keep the good ones and let the others go.
I agree with your “grow your own” strategy, Graeme. Many of my most effective hires were those who I recruited as graduates and grew with the team. It also helps avoid a lot of bad habits formed elsewhere or just different culture and priorities.
Very insightful from everyone so far – especially pleased to see from Graeme that this isn’t just something that James, Paul and I share from working together for many years but has been seen in other businesses too.
Even within my current organisation we have an issue with a central insight team not being close enough to the business and so us havign to spend a large amount of time re-cutting and editing work produced to make it fit the business need – not the most motivating exercise for fully trained analysts. If this is the case with a “out-sourced” team within the business then a team out-sourced to India (or anywhere else for that matter) doesn’t really have a chance!
I was heavily involved in setting up the Bangalore team Paul mentions in the orginal blog post and everyone of them had very impressive experience, CVs and motivation to deliver. The issue was our ability to translate the business question into an analysis question, get it answered and then translate the output back into business speak. It was quicker to do it yourself.
I find this conversation very interesting, mainly because I have been seeing a very strong trend in the US to move away from offshoring and instead bringing your analytics practices back within your own organizations. What my clients have been telling me is that they are better able to “control” and “secure” the data that is being used, plus they have more confidence in hiring the “right” resources in order to accomplish their goals.
When my clients had used Off-shoring they found varied results and weren’t getting the value they expected or had been promised when they first leveraged the resources.
S o I agree with this commentary that I don’t believe it is worth it to outsource your analytics at this time.
As Paul knows I was part of the onshore team in the UK and so I’ve been on the inside of this conversation previously. I also helped interview the team in Bangalore and select the team members.
Paul’s right that it didn’t work. But I’ve seen businesses make it work since. One of our suppliers at Woolworths does it and there are a couple of reasons why
1) bigger on shore teams. If the offshore team is a relatively small part of the overall and there are on shore teams who can translate then it can work
2) the arrival of the agile model. With more frequent daily stand ups and collaboration tools such as JIRA then cross team working is easier.
3) more senior oversight. A lot of us in Paul’s team at the time were relatively junior and overseeing junior teams In bbangalore. More senior resource to act as the account manager with the business helps
I’m still unconvinced in the long term on the benefits. However, given the skills gap in these areas (Australia even more so than US/UK) it’s inevitable other companies are going to try this
Thanks for sharing your experience too, Andy. Very useful.
Great post Paul!
And if I may, I believe it may also tie in closely to our ongoing conversation around the need to break out “Insights & Analytics”, in terms of using the right data to solve the key business problems, from “Data Infrastructure”, where it’s more about the IT work: building the right data silos, reporting and flows to be able to answer those questions.
From the experiences I’ve had and see in others, you may be able to outsource the IT side IF you have carefully considered what data you have, what you need, and how you need it all run, but you can’t outsource those early decisions, nor the end users who apply the data to either algorithms or reporting.
Great point, Charlie & I think it does tie in to that part of our conversation. As a reminder, to readers, you can hear my interview with Charlie Ballard (from Meta) here: https://open.spotify.com/episode/7udkMz3N4CsdwFFsQPuIjT