Home The Humans Upgrade Specialist AI In Practice Go Deeper Let's Talk →

Is Your Offshore Research Provider Underperforming?

Darren SharmaCEO & Founder

Signs your offshore research provider is underperforming

Underperformance in outsourced research shows up in three places. The desk sees it in the work. The provider’s set-up predicts it, usually before the desk sees anything. And the COO’s office sees it in the numbers: an engagement that has stopped growing.

  • What the desk sees: rework; the same instruction given twice; analysts who never bring an idea or push back; drafts that are hard to understand; slow turnaround on events; errors at review; drafts a general AI tool could have written.
  • What predicts it: tenure on the account; the bait and switch; where the analysts studied; training; how oversight is run; client engagement, and whether the people running it have done the desk’s job; whether the analyst can explain a conclusion to the third line of defence.
  • What the COO’s office sees: growth and penetration that have stalled.

The scorecard

What the desk sees

#SignalWhat to count
1ReworkShare of deliverables the desk materially rewrites before use, by month
2The same instruction twiceTimes the desk repeats an instruction, a correction, or the background on a name it has already explained
3ProactivityIdeas and challenges the analysts bring unasked: an issue the desk did not raise, a view that disagrees with the desk’s, a suggestion that would improve the desk’s own idea. An analyst who always agrees is a signal in itself
4Hard to understandTimes the desk has to ask what a draft means, or rewrites it for clarity rather than content
5Event turnaroundTime from a results release or rating action to a usable draft, against the agreed level
6Accuracy at reviewErrors found per deliverable at the desk’s check
7The AI testWhether a prompt to a general AI tool would produce the same draft; if it would, the analyst’s view is missing. Unattributed AI text is the plainest case

What predicts it (supply side)

Some of these are counts; some are matters of judgement.

#SignalTypeWhat to ask for
8Analyst tenure on the accountQuantitativeMedian months the current analysts have spent on the desk’s work, not the firm-wide figure (note 1)
9Bait and switchQuantitativeHow many of the analysts met at pilot or sample stage are doing the work now
10Where the analysts studiedQuantitativeEach analyst’s institution and its place in India’s NIRF management ranking (note 2)
11Training before live workQuantitativeHow long, and who ran it
12OversightQualitativeWho supervises, and how review steps down as an analyst proves themselves: close checking while an analyst is new, then sampling once they are cleared to work on their own
13Client engagementQualitativeWhether the desk talks to the analyst directly, and who runs the engagement: professional client-engagement people alongside people who have done the desk’s job themselves, such as experienced bankers from the markets the desk works in, or only a delivery manager who has never done the work
14The audit questionQualitativeWhether the analyst can explain how a conclusion was reached when the third line of defence asks

What the COO’s office sees

Growth and penetration. An engagement that works spreads: more desks use it, more kinds of work move across, and the team grows. When that has stopped, and nobody onshore is asking for more, the business has reached its own verdict before any scorecard does.

Note 1. LinkedIn shows median employee tenure of 2.4 to 5.2 years at the main mass-market providers (November 2026). Where that median sits under three years, an analyst on the account for two years is already one of the longer-serving people in the firm.

Note 2. The NIRF (National Institutional Ranking Framework) is published each year by India’s Ministry of Education; the current management ranking is at nirfindia.org. A provider that will not say where its analysts studied is telling the desk something.

Why the output is the late signal

By the time the output is visibly worse, the cause has usually been in place for months. An analyst left in the spring. The replacement was briefed by a delivery manager rather than by the desk. The house style held while the context drained away, so the drafts still looked right and read thinner. That is why the supply-side half of the scorecard matters: it moves first.

Qualitative signs, made countable

Most of what goes wrong in outsourced research is qualitative: the analyst does not anticipate, does not understand the context, does not own the work. Each of those can be turned into something the desk counts.

  • Does not anticipate becomes proactivity: ideas and challenges raised unasked, per month (signal 3).
  • Does not understand the context becomes the same instruction twice (signal 2) and drafts that are hard to understand (signal 4).
  • Does not own the work becomes rework (signal 1).

A simple log kept by whoever reviews the drafts, for one quarter, is enough. The point is not precision. It is to turn a vague sense that the work has got worse into a conversation with numbers in it.

After 21 years doing this, what we find is that bankers and asset managers have a strong intuition and, even without the counting, know when some of these indicators are going in the wrong direction.

The bait and switch

The most damaging pattern does not show in the first months at all. A pilot is staffed with the provider’s strongest analysts. The pilot goes well, the contract is signed, and the production work moves to other people. The desk notices a change in the drafts before anyone tells it about a change in the team.

Three checks catch it:

  • Name the analysts in the agreement, not only the number of analysts.
  • Compare the names on deliverables with the names met at pilot stage.
  • Ask the provider to confirm in writing who will do the work in month six.

What underperformance can mean

A poor scorecard means one of three things, and the fix depends on which.

  1. The provider is bad at what it does. The model suits the desk; the provider runs it badly, with high turnover, a thin bench, or a manager standing between the desk and the analyst.
  2. The provider runs one equilibrium well, and it is the wrong one for the desk. Offshore research settles into one of two equilibria. In the lower value-added one, the provider delivers output to a specification, such as a pitchbook or a set of spreads. In the higher value-added one, the analyst holds an opinion, such as a buy or sell recommendation on a bond. Both work, for different desks. A desk that wants opinions from a product team, or a specified product from a team built for opinions, will see the scorecard go red with nobody doing their job badly. A desk that bought product work should judge it on rework, turnaround and accuracy; proactivity is not part of that job. More on the two types of research outsourcing.
  3. Outsourcing does not suit the desk. Some work has to sit next to the people who use it, and some organisations are not open to outside analysts’ views on it. That is a fair conclusion, not a failure.

Check the desk’s own side as well. A vague brief, no feedback, or a desk that has stopped reading the drafts will produce the same signs.

Putting it into the agreement

The scorecard works best written into the service agreement as the service levels, reviewed at the same cadence as the desk’s other third-party arrangements. For outsourcing of critical or important functions, the EBA’s guidelines already expect agreed service levels to include precise quantitative and qualitative performance targets, so that corrective action can be taken without undue delay (EBA/GL/2019/02, paragraph 75(i)).

Corrective measures

What to correct depends on which of the three applies.

  • The provider runs the model badly: put the scorecard to them, with names. Ask for the analysts met at pilot back on the work, direct contact with the analysts, and a quarter to show the numbers moving. If they do not move, change provider.
  • The wrong equilibrium: do not try to push a provider into the other one. It is staffed differently, run differently and priced differently. Move the work to a provider that runs the equilibrium the desk needs, or split the work between the two.
  • Outsourcing does not suit the desk: bring the work back onshore, or keep offshore only the parts that are specified products.

Not sure which applies? Ask us. Few providers look at underperformance this way, because most work in only one equilibrium, usually the lower value-added one, so the full range of options is not always apparent to buyers. We are happy to talk it through, with no commitment: contact us.

Median employee tenure at Frontline is 9.6 years, against 2.4 to 5.2 years at the main mass-market providers (LinkedIn, November 2026). The full comparison is at Upgrade to Frontline.

Frequently asked questions

What are the signs an offshore research provider is underperforming?
Rising rework, the same instruction given twice, analysts who bring no ideas and never disagree, drafts that are hard to understand, and slow turnaround on events. On the supply side: falling tenure on the account and pilot analysts no longer on the work. At business level: an engagement that has stopped growing.

Which KPIs belong in an outsourced research service agreement?
Accuracy at review, rework rate, on-time delivery, event turnaround and continuity: tenure on the account and named analysts. Qualitative targets can be made countable, for example ideas and challenges raised unasked, per month.

What is bait-and-switch staffing?
A pilot staffed with a provider’s strongest analysts, followed by production work done by others. Check it by naming analysts in the agreement and comparing deliverables against the pilot team.

How can a desk judge where offshore analysts studied?
Ask for each analyst’s institution and check its place in India’s NIRF management ranking, published each year by the Ministry of Education.

What should I do if my offshore provider is underperforming?
Work out which of three things is happening: the provider runs its model badly, the provider runs the wrong equilibrium for the desk, or outsourcing does not suit the work. The first is fixed with the provider, the second by changing model, the third by bringing the work back.

Part of: Offshore Analyst Teams – what works and what breaks

Compare: Upgrade to Frontline

Go deeper: How to evaluate an offshore analyst team · When outsourced credit research fails · Why quality degrades after year one · The lament of the KPO middle manager · How to switch offshore research provider