Used car buying guide: Volvo 850

Used Volvo 850

They were the most powerful cars Volvo had produced at that point in its history, a claim given extra emphasis by the company fielding a pair of 850 estates and saloons in the British Touring Car Championship from 1994 to 1996. Admittedly, they differed from their road car brethren in being powered by a 2.0-litre, non-turbocharged engine producing 286bhp and driving the front wheels through a six-speed sequential gearbox, in place of the production cars’ choice of five-speed manual or four-speed automatic ’boxes.

No matter. They showed that more than just building safe and secure cars, Volvo could build exciting ones as well. To ram the point home, the T-5R road car was available in yellow. Today, in estate form and in this colour with a manual gearbox, it’s the most sought after of the three hot 850s. Naturally, importers have spotted an opportunity and, as this is written, there are a number of 1995 N-reg T-5Rs and the later Rs imported from Japan for sale at prices starting from around £7000.

But all this talk of T5s, T-5Rs and Rs is confusing so, to begin, the T5 was the first hot 850. Available as a saloon and estate, it was launched in 1993. Its turbocharged 2.3-litre five-pot motor produced 222bhp, good for 0-62mph in 7.3sec. A torque limiter in first gear was designed to contain wheelspin (the 850 is front-wheel drive) but even with a steady driving style, you’ll be lucky to get 10,000 miles from the front tyres.

The T5 was quick but, apart from the presence of a discreet boot spoiler, didn’t look it. Not so the T-5R that replaced it in 1994. With its side skirts, sportier bumpers, 17in titanium grey alloy wheels and the choice of yellow paint (it also came in green and black), this is the version that changed people’s impression of Volvo, more so when they discovered its overboosted 237bhp engine could propel estate and saloon versions from zero to 62mph in just 6.9sec.

Porsche helped with tweaks to its suspension and the design of the interior, which featured dark grey leather, synthetic suede and wood inlays. Just 400 examples came to the UK but they were enough to prepare the way for its successor, the even more powerful R of 1995.

This time, the 2.3 engine wielded 247bhp thanks to a larger Garrett turbo with improved engine management to reduce turbo lag and a more efficient intercooler. A limited-slip diff on the five-speed manual version helped the car, again offered in saloon and estate forms, to achieve 0-62mph in 6.7sec.

Few of the less charismatic T5s remain, but the T-5Rs and Rs that have survived seem to have stood up well. Rust-free imports from Japan are tempting but check the spec and service history. All good? Then give those XC90s a fright.

How to get one in your garage

An expert’s view 

Richard Ellison, 850 R owner: “My dad was in the traffic police and brought home a T5 estate one night. I fell in love with it and then all over again when I saw them in the BTCC. I bought my T-5R seven years ago, since when I’ve been restoring it. It had good provenance and a load of workshop invoices but it soon became apparent it had been maintained by a chimp. He’d even used wood screws! It’s now concours. I converted it from an auto to a manual and upgraded the Garrett turbo from the G15 to the G16 to suit. I love the T-5R for the fact that it represents Volvo’s first break with its staid image, while being a bit of a sleeper. It’s a proper 1990s premium car, too: reliable and solidly built.”

Buyer beware…

■ Engine

A blocked PCV system can cause the rear main oil seal to blow. To check, with the engine running, hold a rubber glove over the oil filler. If it inflates, the PCV is blocked. Also, remove the dipstick (don’t be alarmed if there’s milky gunk on the end) and watch for oil smoke rising from the tube. Check for leaks from the oil cooler flow and return pipes caused by flexing. Oil (10W 40 semi-synthetic) and filter changes every 10,000 miles are essential. Cambelt and water pump are best changed every 60k miles.

■ Cooling system 

Heater matrix failure is a possibility, so check for coolant loss and an antifreeze smell in the car, a steamed-up windscreen and a wet carpet behind the centre console.

■ Gearboxes 

Ensure fluids have been changed at regular intervals.

■ Suspension 

Check estates for sagging of the rear self-levelling suspension, where fitted. If rear trailing arm bushes are worn, the rear end can ‘clunk’ and feel light.

■ Body

Inspect the front offside wing for rust caused by the washer bottle leaking. Rear bumper mounts rust, too.

One we found 

Volvo 850 2.3 R estate, 1996/N, 140K MILES, £3000 The owner has converted this from an automatic to a manual. He has changed the Garrett turbo to the right 16G type to suit. It has new Bilstein suspension all round, a powerflow exhaust, 18in alloy wheels and service history. Seller says the car runs a treat.


Used car buying guide: Suzuki Jimny

Suzuki Jimny

Since it was launched 21 years ago, in 1998, its design has barely changed so that today even an old £1500 example looks like a more expensive, later model. It’s so reliable that one dealer has stopped selling anything else and devoted himself to it, saving, he says, a fortune in workshop bills.

It loses little money as the high prices of late-plate versions prove (at least until the new model establishes itself and used ones become available) and it costs peanuts to run. Finally, if off-roading is important to you, thanks to its light weight, short entry and departure angles, separate chassis and high and low four-wheel-drive ratios, it can pretty much keep up with a Land Rover Defender.

Prices for leggy Jimnys start at around £1200. Stretch to £3500 and you’ll bag a 2006 car in good nick with 80,000 miles. Prices go all the way to £16,000 for fully loaded, last-of-the-line 2018 cars.

These prices and comments refer to hard-roof versions in five-speed manual and four-speed automatic form, by the way. They were built in Japan and are as solid as a rock. Not so the soft-top version. It was built in Spain for the rental market and to a much lower standard.

It sounds like one of those rumours pedalled by someone with a field of hard-top Jimnys to shift but at least one specialist bitten once too often by a soft-top Jimny won’t buy them for stock, selling only those he gets in part-exchange. The problem is that because they’re a soft-top, they’re often priced at least as high as standard Jimnys despite being more expensive to tax and insure, and probably in need of a new roof (about £500 plus fitting).

Throughout its life, the Jimny was powered by a 1.3-litre petrol engine. From launch, it had just 78bhp, but in 2005, it was replaced by one producing 83bhp and with variable valve timing (VVT) for a broader spread of torque. You feel its benefit in more relaxed motorway cruising and better low-speed pick-up off road but the gains are marginal. Early or later engine, all versions of the Jimny sit on an old-school, ladder-frame chassis and put their power down via a proper selectable four-wheel drive system offering a choice of high and low ratios, as well as rear drive only.

You may prefer to change gears yourself but don’t dismiss an automatic Jimny. It’s a tough gearbox and its smoother changes make for calmer, more relaxed progress. Anything that counters the little car’s choppy ride has to be good.

Until 2009, the dominant trim level was JLX (roof rails, electric windows and mirrors). From 2009, badging was changed to today’s more familiar SZ3 and SZ4, the latter with alloy wheels, air-con and part-leather trim. Options included a sat-nav and a reversing camera. Nice to have but better still is the car they’re fitted to.

An expert’s view 

Will Chappell, director, Cotswold Jimnys: “We started the business 32 years ago selling all sorts of 4x4s but changed exclusively to Jimnys seven years ago when we realised they were the only model that didn’t cost us anything in the workshop. We’ve since sold upwards of 700. Unless they come in as a part-exchange, we don’t touch the soft-top because the quality is terrible. Otherwise, at all years, a Jimny is very reliable and cheap to run. Just make sure you rustproof the chassis.”

Buyer beware… 


The early 1.3 uses a belt that should be changed at 70,000 miles. The engine works hard so listen for camshaft noise and check around the filler for water and oil emulsion, a clear sign of head gasket failure. On older cars, check the condition of coolant hoses and the radiator for leaks. Faulty crank, intake and exhaust sensors will cause the engine to misfire and lose power.


The four-speed auto is generally trouble-free. The manuals are tough but hard work can cause the occasional synchro problem. A worn clutch release bearing may squeal on engagement. On cars little used to running in four-wheel drive, the system can need coaxing back to life. On older cars, the vacuum hoses to the front hubs can perish.

Suspension and brakes 

Check for broken springs, leaky dampers, perished bushes and scored brake discs.

The ‘death wobble’

It can suffer so-called ‘death’ or wheel wobble at around 50mph caused by any number of problems, including worn kingpins, wheel bearings and CV joints, or simply poor wheel balancing.

Chassis and body 

Rust is generally surface and rarely terminal but check joints and mountings for perforation. Check behind plastic body mouldings and under the bonnet for rust caused by trapped mud and water.


Early cloth seats are likely to have burst but otherwise the interior is surprisingly tough and control and switch issues are rare.

Also worth knowing 

Want to boost your Jimny’s off-roading capability but spend most of your time on-road? Invest in a good set of off-road rubber. BF Goodrich is the best known but cheaper tyres from £70 are just as good.

How much to spend 

£1250-£2450: Hard-top Jimnys with high mileages.

£2500-£3450: Up to 2007 but mileages as low as 60k and many with full history.

£3500-£4995: More low-mileage 2006-08 cars.

£5000-£6450: Good choice of 2008-on cars with very low mileages

£6500-£8450: Up to 2013-reg SZ4s in good condition.

£8500-£10,450: More 2012-14 low-mileage cars.

£10,500 and higher: Good choice of late ’n’ low cars.

One we found 

Suzuki Jimny 1.3 JLX, 2003/03, 50K miles, £2600 

Private-sale Jimny with two previous owners and full service history. Rear springs replaced recently. Comes with two spare wheels. Also has a towbar, so check the gearbox and head gasket for signs of strain. Always been garaged. Seller says it’s “extremely reliable.”


Used car buying guide: Mk2 Mazda MX-5

Used Mk2 Mazda MX-5

If Volkswagen ever thought launching a new Golf was tough, launching a new MX-5 must have been harder still for Mazda’s dutiful executives, still reeling from the success of the original. Of course, it could never be a straight copy but, even so, the absence of the Mk1’s pop-up lights – a move forced on Mazda by US safety officials – was a blow. A bit like when BL was forced to replace the MG B’s traditional chrome bumpers with ugly rubber affairs.

Still, what else were you going to buy back in 1998, especially since word was that the car possessed the same sweet, rear-drive handling, delicate steering and crisp gearchange (plus some extra kit including a heated glass rear window) that had made the Mk1 such a winner?

Twenty-one years later, these very qualities continue to attract second-hand buyers, especially now that prices for Mk2s begin from as little as £350 for runners with an MOT. In fact, they don’t really go any higher than £3500 for the best cars. In part, they’re held back by the Mk3, which kicks off at around £2250 but also by demand for clean Mk1s. You’ll easily pay £6000 for a good one of those.

Condition and not age or specification or size of engine is key to valuing a Mk2. You’re just as likely to find a tatty but late-plate 2004 1.8i Sport for a few pounds as you are an early but bright 1998 1.6 for a few thousand. Those two engines are your lot. The 140bhp 1.8 Sport had a Torsen limited-slip differential and 15in alloy wheels.

Cars dating from the facelift of 2001 are known as Mk2.5s. It was a mild affair, the biggest change being the headlights, which now had three bulb chambers. The front bumper also gained a couple of foglights, or mesh grilles if the trim level didn’t permit them. An automatic gearbox slipped onto the price list but missed the point and is rare today. The 1.8i S gained larger, 16in alloys.

Being an MX-5, there are a few special editions to disentangle. Two that caught our eye were a 2001 Y-reg 1.8 Jasper Conran 1.8 with 75,000 miles, a new belt and a good service record for £1795, and a 2003 53-reg 1.8 Angels, inspired by the film Charlie’s Angels, with 79,000 miles for £2190.

However, standard, Sport or special edition, it doesn’t matter. Condition is king and until you’ve crawled around the car testing the sills with a magnet and lifting the carpets checking for damp, there’s no point getting excited about the specification. But by all means get excited about how it drives.

How to get one in your garage

An expert’s view 

Ashley Martin, The MX-5 Restorer: “I know Mk2 MX-5s inside out, literally so, due to the cavity rust issues afflicting them and Mk1s. We do a really thorough repair job on the cars. I own a Mk1 and a Mk2, both 1.8s, and love them. I’ve driven a few Mk3s but there’s something about the earlier cars. They just want you to have fun. There’s a huge and growing following for both, with the Mk1 still the most popular. The trouble with the Mk2 is that it’s not a question of if it will rust but when. They are cheap, though.” Buyer beware…

■ Engine: Cambelt should be changed at 60k miles. With luck, the water pump will have been changed at the same time. Check for oil leaks from the cam cover (not serious) and listen for misfires caused by the coil packs. (There are two.) HT leads routinely break down.

■ Gearbox: On the overrun, a sound like marbles rattling in a tin is gear lash caused by the gearbox being misaligned when being refitted after, for example, clutch removal. Many garages don’t realise it has to be aligned. Later cars got a rubber damper. The later, six-speed gearchange is not as quick as the five-speed. Listen for failing propshaft joints (they’ll squeak), especially in reverse.

■ Suspension, steering and brakes: Check for uneven tyre wear, suggesting incorrect four-wheel geometry. All four should be set up independently of each other. Suspension bushes will look terrible but they rarely fail. Check for seized brake caliper pistons and springs broken by speed humps.

■ Body: Regarding rust, later Mk2s fare worse than early ones. Rotting rear wheel arches are easy to spot but the sills, which rust from the inside out, will require much closer examination. Front chassis rails suffer badly. Unlike the Mk1’s, they’re a double-skinned box section for additional impact strength but this traps moisture and they corrode. Surface rust blights the underside, rear subframes and wishbones. Check for bonnet corrosion caused by stone-chipping.

■ Interior: Expect worn seats and collapsed armrests. Also worth knowing The first thing to do when you get your Mk2 home is brush and then rustproof the underside with something like Dinitrol. It’ll stop things like subframes becoming terminal. Then go online and find out where all the entry holes are and flood the cavities to delay rot from the inside out.

How much to spend 

£350-£1249: All ages and mileages and in running order but verging on the tatty, such as a 2003 1.6 with rotten sills for £1150.

£1250-£1999: Huge choice of better-looking cars, many with full or near-full service histories but check for filler in places such as sills and wheel arches.

£2000-£3500: More of the same, including a promising 2002/52 1.6 with 60k miles and an excellent Mazda service record for £2299. Also, a nice, one-owner, 1998 S 1.8 Sport with 22k miles and full Mazda service history for £3000.

One we found 

Mazda MX-5 1.8 Sport, 1999/V, 64K miles, £1999: ‘Full service history!’ exclaims the ad before admitting that it has had 11 services… No worries. Your biggest concern will be body rot but this is an early Mk2 and these were better protected, so fingers crossed.


Used car buying guide: Audi RS5

Used Audi RS5

The one we’re thinking of has done a not inconsiderable 152,000 miles but they’ve been mainly clocked on the motorway apparently and, miraculously, it’s a one-owner car with full service history and new Pirelli P Zeros. Crucially, the engine has had a carbon-clean, too.

It’s also loaded with kit to the extent that it cost around £65,000 new. But its owner has certainly had their money’s worth and appears to be entirely satisfied with it.

It would be our ‘one we found’ but for its mileage which, while not a problem to a savvy buyer like you guided more by condition and provenance, would be a turn-off to those looking to cash in on the growing appeal of early RS5s rather than later, pricier facelift cars.

The RS5 V8 coupé was launched in 2010. A facelift followed two years later bringing a larger, single-frame grille, a restyled bumper (making the headlights look more aggressive) and new, 10-spoke, 19in alloys. Inside, the biggest change was Audi’s latest multimedia interface system.

Mechanically, though, it was business as usual: in addition to the naturally aspirated V8 – derived from the R8’s 5.2 V10 – and quattro drivetrain, it retained the pre-facelift model’s seven-speed S tronic dual clutch gearbox (there was never a manual option) and mechanical centre differential. Bringing up the rear was the same electronic differential shuffling torque on demand. Dynamic Ride Control dampers remained as standard.

The thing to remember with any used performance car is that while it might now be cheaper, the cost of maintaining it isn’t. It’s this crucial fact that escapes starry-eyed owners, who then cut corners on servicing.

That’s a bad move since something like a fast-fit stamp is the last thing you want to see in an RS5’s service book. No disrespect to the able technicians who work in such places, but knowing some of the RS5’s idiosyncrasies is crucial to its trouble-free running. They include rattly chain tensioners at around the 70,000-mile mark and, as already hinted at by the value of a carbon-clean, the tendency for the inlet valves to coke up.

Theories as to the cause of this include the engine breathing oily fumes from the crankcase ventilation system and the inlet valves not being hosed down with fuel, as they would be were the injection system port-based rather than direct.

It’s all horribly technical but all you need to know is that a full set of specialist or main dealer service stamps and evidence of supplementary work including regular transmission fluid changes is the secret to RS5 happiness. That and the sound of its 444bhp V8.

How to get one in your garage

An expert’s view 

Simon Howarth, founder, AMD Technik: “We’ve been servicing performance Audis for 30 years and we know RS models inside out. Personally, I prefer the earlier B7 RS4 to the RS5. It’s the purist’s choice: more practical and exclusive. There are loads of RS5s for sale, but have you ever tried opening the door in a car park? They are so long, you have to squeeze out to avoid bashing them. It feels like a car for the American market. But I can understand why people love them for the wonderful V8 and those coupé looks. The facelift didn’t bring much so early, low-mileage ones will be in demand as prices fall and people see they’re a performance bargain.”

Buyer beware… 

■ Engine: Start the engine from cold and listen for timing chain rattles caused by a sloppy tensioner. Favour a car that has had regular servicing since frequent oil changes are key. Post-70,000-mile cars can suffer coked inlet valves indicated by reluctant starting, uneven idling or a shortfall in power. If a sustained blast up the road doesn’t cure it, it’s a heads-off job.

■ Gearbox: A fluid and filter change every 40,000 miles. Don’t go by the service record, ask to see the invoice. Check changes are smooth in manual and auto modes.

■ Suspension: Worn front suspension arms (there are four each side) are common and expensive (£100 per arm). Tramlining is a sign of wear. Check the adaptive dampers for leaks. Abnormal wear on the inner edges of the tyres could be worn bushes forcing the wheels to toe outwards. Check the brake discs for wear – replacements are £400 each.

■ Electrics: Blocked windscreen scuttle drains are a problem. The pooling water flows into the pollen filter and down to the carpets, flooding the comfort ECU that works the central locking and windows.

■ Body: Any corrosion will be damage related.

■ Interior: Things should be Audi-tight in here, although the firm suspension and unforgiving wheels do loosen things a touch. Just avoid anything too vocal. Also worth knowing The RS5 is not for the faint-hearted. Covering our ‘one we found’ example (below) with an official Audi warranty for 10,000 miles and 12 months, paying a £100 excess, would cost £2135, or £181 per month. Still, you’ll sleep at night.

How much to spend 

£14,000-£19,999: High-mileage (over 120,000) to mid-mileage (around 60k) 2010-11 cars, most with full service history.

£20,000-£21,999: 45k to 70k-mile 2010-11 cars with solid histories, few previous owners.

£22,000-£23,999: Mainly 25k to 50k-mile 2011-reg prefacelift, including cheapest approved used: 2011 with 42k miles for £22,450.

£24,000-£25,999: Facelift 2012-13 cars with less than 50k miles plus some higher-mileage 2014s.

£26,000-£29,999: More 2013-14-on cars with solid histories and fewer than 45k miles.

£30,000-£34,999: Mainly 2015-on cars. Lots of cabriolets.

One we found 

Audi RS5 4.2 V8 FSI, 2011/11, 78K miles, £18,795: Full service history, just two previous owners and described as being in ‘pristine condition’, this RS5 is also sold with a service pack and 12-month warranty. Listen for timing chain rattle.


James Ruppert: the fastest-selling used cars on the market

Ford Ka

I was fascinated to read what the fastest-selling used cars in 2018 were. These figures come courtesy of Indicata, the used vehicle management portal from Autorola UK, which checked out the fastest-selling used cars from the 1.1 million aged between six and 36 months old sold by dealers during 2018.

These fascinating figures reveal just how long popular cars hang around on the forecourt. When you try to sell a car, you might start to panic after a couple of weeks if no one has actually paid up and taken it away. Dealers are playing the longer game and so should you.

Anyway, the least shocking result is that the Ford Ka was the fastest-selling used car in the UK in 2018, taking an average of 32 days for dealers to sell. A nearly new Ka, then, should be the easiest to sell. I hadn’t been keeping up with what a 2018 Ka looks like (I was a bit taken aback), but there seem to be a few four-door Zetecs around with hardly any miles for an asking price of £8995. It will have the City pack as well, so you will be able to park it because things will be going beep to make it easier.

Then again, if you’re trying to sell a less popular model, be prepared for quite a long wait. Subaru’s Outback (103 days to sell on average), Forester (105 days) and XV (107 days) were the slowest-selling used cars. I like Scoobies because they can do things and are rather good value. Like the latest Ka, I had not logged what an XV actually looked like. I don’t mind it. Don’t think I have ever seen one on the Queen’s Highway, though. A 2017 2.0d SE Premium with just shy of 20,000 miles will cost just under £17,000. But the bottom line is that it could take three months to shift a Subaru. Blimey, that’s worth remembering when you buy one.

Basically, Ford makes up the bulk of super-fast used car sales, with six models in the top 10. The Galaxy, Kuga, Tourneo Connect, Fiesta and Ecosport are the others.

Maybe the most interesting aspect of all this is that the pure-electric Renault Zoe, which was ranked in joint-fourth place, on average takes just 38 days to sell as a used car. The point here is that the monthly charge for the battery can be more than some are planning to spend on petrol. So if you’re tempted to turn your back on petrol for a Zoe, do your sums first. But if you do buy one and it doesn’t work out, at least you shouldn’t have to hang around long before selling it on to its next owner.

What we almost bought this week

Isuzu Trooper

Readers of a certain age will get our fascination with the Trooper, one from an era of unbreakable, Japanese, ladder-framed 4x4s we called off-roaders, not SUVs. For farmers, there was a diesel; for the rest of us, this, the 3.2 V6 petrol. This one’s a 1999 S-reg with 165,000 miles, for £1850. Just 15mpg and 358g/km CO2 – those were the days.

Tales from Ruppert’s garage

BMW 320, mileage – 82,709: The weather has taken a turn for the worst and that means I am running around in an almost 40-year-old car. Obviously, it should be stored away and not used on our salty roads. Well, never mind: I believe in using cars, because who knows how long I have to enjoy interesting motors?

Well, I decided that the handling of a 1970s E21-gen 3 Series on ice wasn’t good, so I’ve added some ballast in the form of compost. The bag cost £2.99 and effectively this is analogue 1970s’ ASC. It drives great, but there remains a list of to dos.

Reader’s ride

Smart Roadster

Christopher Walker Brown purchased this Smart Roadster and was chuffed with his £550 bargain sports car. “Then oil sprayed over the hot exhaust, and caught alight!” he tells us. “I jumped out and emptied a powder extinguisher over the exhaust and rear of the tiny engine, but the fire had burnt the wire harness, coil packs, fuel lines and intake manifold. Things looked terminal. But after just £98 of purchases I had enough new parts to repair the fire damage. Now the car is running great and has put a smile back on my face.”

Readers’ questions

Question: I’m considering buying a Hyundai Nexo, the brand’s new hydrogen car. How big is the refuelling network? George Bodian, Redhill

Answer: The Nexo will be available to lease from March. However, there are just 13 publicly accessible hydrogen filling locations in the UK. Most are located around London but ITM Power, the UK’s biggest hydrogen fuel operator, has recently opened sites in Swindon and Rotherham and this year will open one in Gatwick, another in Birmingham and two in Derby. UKH2Mobility, a government-funded organisation promoting the fuel, says that 65 filling stations will be up and running across the country by 2020. John Evans

Question: I’m 23 with three years’ driving under my belt and a clean record. I fancy a hot hatch for £5500. Any suggestions? Ryan Jones, via email

Answer: You sound like a responsible chap, so our first thought is something respectable and grown-up such as a Volkswagen Golf GTI. Your budget will easily net a tidy 2005-07 Mk5 with around 70,000 miles, few previous owners and full service history. All very sane and sensible, but if you fancy something a little edgier, how about a Ford Focus ST Mk2? We saw a mint 2007 car with 68,000 miles and full Ford history for £5500. Head or heart, Ryan? John Evans


Why We Need More Visibility Into The Social Media Analytics Algorithms We Use

Social media has increasingly become the lens through which we observe the modern human world. In turn, the vast industry of social media analytics platforms have become the instruments through which we use that lens to make sense of society. Yet, despite real world business and governmental decisions being made on their results, we know surprisingly little about the algorithms powering most of these platforms, especially their edge cases and interpretative nuances. How can we make meaningful decisions about social media based on the results of algorithms we know nothing about?

Nearly every social media analytics platform today offers some form of sentiment analysis, typically a simple positive-neutral-negative score, but also occasionally a few additional more nuanced categories like joy or fear.

These sentiment scores have become a go-to resource for understanding social trends, classifying reaction to topics and helping guide business and marketing decisions.

Calculating social media sentiment is an extremely difficult task, filled with complexities.

Unfortunately, few social media companies offer any substantive technical detail on how their sentiment algorithms work, treating them as proprietary business secrets.

Some disclose that their systems are traditional simple “bag of words” word counters that simply have two lists of words, one for “positive” words and one for “negative” words and just count up how many of a tweet’s words are in each list. Some add a score to each word recording just how “positive” or “negative” it is, to differentiate between “love” and “like” or “loathe” and “dislike.” Some use more sophisticated statistical or even neural algorithms. However, almost none share their actual word lists or algorithms.

Rudimentary visibility into sentiment algorithms can be found by filtering to just negative or positive tweets and then using the word cloud histogram feature offered by most platforms to see which words seem to dominate those tweets. This can help identify particularly glaring vocabulary mismatches.

For example, one tool consistently ranked basketball tweets as substantially more negative than football tweets. The culprit turned out to be the fact that many basketball tweets in the sample contained the word “court” to refer to the basketball court, while the tool’s sentiment dictionary labeled “court” as a highly negative word, assuming it always referred to a legal court.

Similarly, another analysis showed tweets about Republicans being far more positive than those about Democrats because during the sample period Democrats were often referred to as simply “democrats” whereas Republicans were consistently referred to as “the republican party.” The word “party” was being incorrectly labeled by the algorithm as a very positive word.

Reverse engineering sentiment algorithms in this fashion can help identify misalignments between the algorithm’s dictionaries and the specific domain being examined. Some social analytics platforms permit their users to manually adjust the sentiment dictionaries applied to a given analysis, supporting domain adaptation, though not all do. Such algorithmic analysis can also yield clues as to the provenance of its dictionary, with some platforms using lightly modified well known open source sentiment dictionaries.

Some provide clues buried in their internal user documentation, such as noting that their sentiment system was initially trained on a few hundred thousand or a few million tweets when the company was first founded and has not been updated since. In fact, few update their dictionaries in realtime hour-by-hour to capture the latest linguistic nuances of Twitter.

Using a dictionary based on a few million tweets sampled from a decade ago raises grave concerns about just what those results are actually measuring.

Most social media analytics platform users are not data scientists, meaning they likely aren’t thinking critically about these sorts of questions or performing systematic evaluations of the results they receive.

Language detection is another opaque but critically important algorithm when searching for words that have different meanings in different languages or which represent a brand’s name in one language but a common unrelated word in another.

At first glance it might seem to be relatively trivial to determine the language of a given tweet. However, the small amount of text and prevalence of acronyms and slang terminology makes social media content especially difficult for traditional language detection algorithms. Tools like the Google Chrome Language Detection (CLD2) library can be readily applied to tweets with quite usable results, but many analytics platforms deploy their own custom algorithms that have been optimized for social use, especially Twitter.

Widely used libraries like CLD2 have well understood performance characteristics and extensive documentation on their edge cases. Some, like CLD2, are completely open source, allowing advanced users to fully understand precisely how the algorithm arrives at its determinations and to proactively identify environments where it may struggle.

In contrast, few social media analytics firms provide much in the way of documentation of their proprietary language detection systems. Many decline to answer specific technical questions, including the algorithm type and size and origin of its training data, treating such information as proprietary business information.

Similar to sentiment, it is sometimes possible to reverse engineer that a given company’s “proprietary” algorithm is in reality just a standard library like CLD2 with a few basic commonsense preprocessing steps like removing hyperlinks and @username references.

Most of the time, however, it is simply impossible to know how a company’s language detection algorithm functions.

Relying on a third-party algorithm without any understanding of its nuances and edge cases is extremely dangerous when it comes to interpreting the results it yields. If an analytics platform reports that Dutch tweets about a topic have decreased by five-fold over a 24-month period down to almost zero, does that really mean that that the Dutch have simply stopped talking about that topic or could it instead simply be that Dutch Twitter practices, from slang usage to abbreviations, have evolved in such a way that the company’s language detection algorithm is becoming less and less accurate at detecting the language?

Without additional information there is no way to know whether the linguistic trends being observed are real or merely algorithmic artifacts.

Comparing results across multiple social analytics companies can lend confidence to observed trends, but the lack of technical detail on the underlying algorithms used by each platform makes it impossible to know whether they are all in reality using the same shared algorithms under the hood.

Many platforms offer vague measures of “importance” or “impressions” or “influence” of those tweeting about a given topic. Some provide at least basic definitions of those terms, such as summing the total number of followers of all users who tweeted about a given topic. Yet, none of these approaches is truly satisfying or meaningful across all queries.

If Donald Trump tweets his support of a new book, his endorsement is likely to lead one half of the US population to embrace the book and the other half to demonize it. Similarly, if Barack Obama tweets his support of a book, the reaction will likely be precisely inverse from that of Trump’s endorsement. In short, both individuals have a specific demographic and ideological base they are highly influential to.

A marketer wishing to pitch a liberal-leaning new book can’t just rank all Twitter users by one magical score of “influence” and pick Donald Trump from the top of the list to ask to endorse it nor could they pick Barack Obama to pitch a conservative leaning book. They have to look at the “demographics of influence” of each user.

Few platforms offer such demographic-level influencer scores as part of their routine summary displays.

In fact, few platforms divulge how they compute the demographic information they offer for Twitter users, from age to geography to income and education levels. Estimating the location of non-geotagged tweets is an extremely difficult task and the majority of the most obvious approaches don’t actually work.

Filtering tweets by country of origin is thus an incredibly error prone process at best, with uncertain results.

More broadly, there is often a sharp divide between the marketing materials many social media analytics platforms tout and the limited realities of how those platforms actually work. For example, platforms may aggressively market themselves as deep learning companies that harness the full power of neural networks to make sense of Twitter. In reality, some caveat those claims upon further scrutiny, acknowledging that they limit their use of deep learning to just a few minor specialty tools and that those tools are in turn limited to small random samples of data, with the overwhelming majority of their results being based on non-neural approaches.

In short, don’t trust a company’s marketing brochures – ask the hard questions about whether each algorithm you will be using is neural based, Naïve Bayesian or simply word counting.

To reduce the computational complexity of tools like word clouds, attribute histograms, maps, customized sentiment analysis, clustering and other higher order analyses, some companies limit their analytics tools to small samples of the total data. While the original query might match more than a quarter billion tweets, the resulting word cloud might be based on just the most recent 1,000 tweets or a random sample of 10,000 tweets, vastly reducing its coverage. Some platforms place prominent warnings in their user interfaces about this sampling, while others bury these caveats deep in their technical documentation.

Putting this all together, the social media analytics revolution reflects a broader trend of the big data world: as we assemble ever greater archives of human behavior, we explore that data through ever more opaque algorithms. We shovel petabytes of data through these black boxes and report what came out the other end without even the faintest understanding of whether those reported results are in any way accurate or meaningful or how their accuracy might vary from query to query. In turn, businesses and governments make very real economic and policy decisions based on numbers that could for all intents and purposes have simply been produced by a random number generator.

In the end, how can we make meaningful decisions about social media based on the results of algorithms we know nothing about?



11 Things Climate Change ‘Dismissive’ People Say On Social Media

It is clear that climate is changing, and there is a human component on top of the naturally varying system. Most climate scientists understand this, and most logical people do too. The 4th National Climate Assessment report is a good place to find affirmation for these statements. Each year, the Yale Climate Communication group and George Mason University scholars query the American public about their views on climate change. Within this study, there is always a certain percentage that fall into a category called “Dismissive.” According to the study authors,

the Dismissive are very sure it (climate change) is not happening and are actively involved as opponents of a national effort to reduce greenhouse gas emissions.

Currently, the percentage is roughly nine percent. While their numbers are small, they are often very loud, persistent, aggressive and vitriolic in social media. Over time, I have noticed 11 “Dismissive” tactics on social media.


Ice Ages. Ice ages always seem to come up and some statement about natural cycles. It is honestly stunning this happens since most climate scientists are very aware of the various ways that climate changes naturally. The discussion about climate change is not an “either/”or” discussion. It is an “and” discussion. Grass grows naturally, “and” it grows differently with fertilized soil. Trees fall naturally in the forest, “and” they can be cut down by a chain saw.

That magazine article from the 1970s. Apparently there was an article in Newsweek in 1975 that ran a story about a “cooling world.” It is amusing to see how often this is cited in social media. As a wrote previously in Forbes,

No, a magazine article, a few people, and some literature said this not the majority of scientists or scientific studies. The writer of that magazine article has even debunked this himself.

Citing one random study. I call this 1-study mania. Over the years, I have seen people criticize the peer review literature. They talk about how it is unreliable or biased. While there are certainly issues with the literature, it is still an important gatekeeper against bad science in the same way the FDA is for bad food or drugs. Here’s the kicker though. As soon as a study appears that supports a “confirmation bias viewpoint”, they are quick to cite the study to support their point.

“Grand Poobah” effect. I observe this often in social media. A person doesn’t necessarily have a strong background in climate science but relies on some  scientist or personality for talking points or to validate their positions. They will often even mention or tag that person in their social media post. I call it the “Grand Poobah” effect and have written about it previously.

Doubt and its merchants. There is typically some sample of comments about scientists and grant money. That statement illustrates a lack of understanding of the science grants process. Here is a good “101” at this link. There are rigorous processes for attaining grants based on science inquiry and review. There are also “other” funding models and grey literature publications designed to advocate certain positions or misinformation. Merchants of Doubts by Naomi Oreskes and Erik Conway is a good book to dig deeper into the latter.

Credentialing. This is the era in which a “Tweet” is presumed to carry as much weight as a degree or years of scientific inquiry. The “Dunning – Kruger Effect” is in full-effect on social media. An oft-used strategy is “I have a degree in (fill in favorite discipline that is not climate science or climatology)” or “I study this in my spare time while in my basement eating cookies.”

Deflection. Another tactic that I notice is lobbing questions of deflection. This is usually some random, “seemingly” intellectual, provocative or irrelevant question that has the intent of “public gotcha” to the climate scientist.

It’s cold. This winter I am certain that you have seen this one: “It’s cold or snowing so global warming must not exist.” Nope, it means the day or week is a manifestation of weather. It is not “where you live” warming. It is not “my little part of the planet on this particular day” change.

They changed the name. Speaking of global warming, there are always a handful of folks that finds devious intent in the use of climate change or global warming. I discussed the reasons why that is another “smoke and mirrors” tactic in a previous Forbes piece.

No profile and few followers. Many of the dismissive comments come from accounts with few followers (less than 10) or no profile picture. I am guessing these are “bots.”  I suppose this is technically not “saying” anything as the article title indicates, but you get the point.

Storms always happened. This is a very common one. Aforementioned comments about trees falling in the forest or the grass-fertilizer relationship address such statements.

I am sure you can name others that I missed, and I certainly hope I caught all of the “typos” because they hyper-focus on those like a laser too.

Enough of this, it is time to watch some college hoops.

The six Americas on climate changeYALE CLIMATE COMMUNICATION

Dr. Marshall Shepherd, Dir., Atmospheric Sciences Program/GA Athletic Assoc. Distinguished Professor (Univ of Georgia), Host, Weather Channel’s Popular Podcast, Weather Geeks, 2013 AMS President


Three Social Media Ad Techniques To Boost Your Brand’s Visibility

In 2013, Facebook advertising had already been around for a few years, but many entrepreneurs were still slow to adopt it for their business. For some, the advertising channel was just a source of confusion. Heck, maybe it still is today.

If you were advertising online in 2013, you’ll know exactly why that year was special. That was the first year Facebook launched video advertisements — and wow, if you ask me, the results were tremendous. Back then, you could run a video advertisement and get an insanely low cost per click, which usually meant high return on investment (ROI), as long as you had a good product to sell.

The reason I mention this is because, for most marketers who have been around the block, you’ll hear the stories of “the good old days,” when advertising was super cheap and extraordinarily profitable (at least in comparison to today). You may even have some of those stories yourself.

The good news is that right now in 2019, there are still worthwhile social advertising strategies and techniques available. That’s right — the “good old days” are still here. I find that most marketers are unaware of how powerful some of the new features out there are, so I’m going to explain three of the top social media advertising techniques that you can use to create visibility and profits for your company.

Ready? Let’s dive in.

1. Instagram Stories Ads 

Those quirky 15-second videos you see on Instagram are actually my preferred, No. 1 choice for present-day social media advertising. The reason is simple. Instagram (and concurrently, Facebook) Stories ads are relatively new, and as with most new advertising techniques, costs are currently low compared to other more time-tested strategies.

If you’re new to the scene, let me give you a quick background. A few years ago, Snapchat created the “stories” product, and Instagram later launched its own version. A “story” is a 15-second video that people film, often about their day or something funny, which lasts for 24 hours on their Snapchat or Instagram profile. Most people use stories just to have fun or, in some cases, give quick value.

Now, for the good stuff. Here’s why you want to pay attention to Stories:

We, as users, tend to respond to stories differently than a traditional news feed ad. When someone sits down to watch through their Instagram stories, they are often in a much more relaxed state of mind. They are used to seeing funny, quick 15-second videos. So, if you can make your ads engaging and match this type of content, they have the potential to do well.

Another benefit of Instagram Stories ads is that they only need to be 15-second videos (and they can even be still pictures). So, if you have a hard time getting creative, you only need to produce 15 seconds of content. Easy.

Start testing Instagram Stories ads today — in my opinion, this is probably the most important opportunity to immediately utilize.

2. Instagram Feed Video Ads

Now that you understand Instagram Stories ads, let me tell you about the second most profitable ad source, in my opinion: Instagram feed video ads.

It’s no surprise that videos can be powerful, especially with so many people using services like Netflix or YouTube. And today, many people are inclined to watch quick videos versus long-form videos. That makes your job super easy as an advertising creator.

Here’s what you need to know about Instagram video ads: First of all, they only need to be 60 seconds. They actually can’t be longer based on Instagram’s parameters. Another good thing about Instagram video ads is you can post them onto your Instagram profile and then boost them. So by default, you can get a lot of traffic directly to your profile page.

If your company has a strong presence on Instagram, you can potentially gain more sales due to the social proof you’ve created on your page. For example, if your brand page has 50,000 followers and someone sees your ad, they may be more inclined to like and trust your brand because 50,000 other people also seem to like and trust it (because they follow it).

And if your brand is new and doesn’t have that many followers, don’t worry. Instagram has advertising options to help you find more followers, which leads me to my next point.

3. Instagram ‘View Profile’ Ads

The theme today has been Instagram ads, and your Insta toolkit wouldn’t be complete without using some “view profile” ads. A view profile ad is actually more of a call to action than a type of advertising. Let me explain.

In traditional advertising, you would have an ad (say, your 60-second video) and then a call to action (this is a link to sell your product or get the viewer of your ad to take action). On Instagram, you can make the call to action “view profile.” What this means is that instead of sending the person clicking on your ad to a sales page or to another website, you can send them right to your Instagram profile.

Now, you might say, “Why would I do that when I could send them to a link to buy something?”

Here’s why: You can use this type of advertising to get followers, build up social proof and set the scene to maximize and amplify your other advertisements.

When people view your profile, they may follow, like or comment. So, when you run other ads in the future, other people will see the follows, likes and comments, and they may be more likely to do business with you based on your social proof.

In my opinion, view profile ads are super powerful, and they are a perfect complement to the other tools in your tool belt.



Serbia Protests: How Social Media Facilitates Political Protest

Screen with social media icons applications Facebook, Twitter, Instagram, WhatsApp, WeChat, Telegram, Skype, Youtube, Snapchat etc.GETTY

Anti-government demonstrations in Serbia have entered their third month – against the ruling Progressive Party (SNS); a conservative, populist, and pro-EU party. Fresh rallies are rolling out across the country and in parts of Kosovo, and social media is playing an increasingly important and necessary role.

Led by Aleksandar Vučić- the President of Serbia,and his Progressive Party is currently facing its biggest crisis since coming to power in 2012. The initial protests were triggered when opposition leader Borko Stefanovic of the Serbian Left party was severely beaten by masked men. The iron rods used by Stefanovic’s attackers were widely described as having “painted” his shirt red. As a result of the assault, social media was flooded with #StopKrvavimKosuljama (#StopTheBloodyShirts). Further outrage followed when several weeks later journalist Milan Jovanovic’s home was torched in an attack widely believed to have been carried out by elements of the Progressive Party – on January 25th Dragoljub Simonovic, a Progressive Party member and head of the Grocka municipality, was arrested on suspicion of ordering the attack on Jovanovic. Three more people were also arrested on suspicion of having executed the crime.

However, the protests go beyond political violence. There is increasing dissatisfaction with the rule of the President and declining political and civil liberties. Many media professionals, particularly from independent outlets, have reported increasing cases of verbal threats, surveillance, harassment, attacks against property and intimidation- to name a few.

Citizens are unable to express their dissatisfaction through established institutional channels – the Progressive Party holds a tight grip on television and print media, the latter of which continues to dominate Serbian media consumption. Thus protesters have taken to social media – where unfettered by the government, use of the internet by Serbians as their primary news source has tripled in just six years

Initially only concentrated in Belgrade, the protests have slowly begun to organise and expand nationwide.

Serbian actor Branislav Trifunovic, a leading figure of the protest movement has been actively engaging and mobilising people via twitter to take part in the protests. He says: “Social networks have remained the only possible way for the people to be told, truthfully, about what is happening. Broadcasters are, in one way or another, owned by the ruling party. Therefore, social networks are seen as dangerous to the regime.”

Evidence shows internet access and social media has been vital to information exchange during protest movements in the last decade. For instance, during the 2013-14 protests in Turkey and Ukraine, Facebook and Twitter were used to facilitate organization, transportation, advise protesters, encourage support and catch the attention of international news coverage- and now we are seeing similar uses in Serbia.

During the demonstrations, protesters tweet and live stream under the hashtag #1of5million- a hashtag coined late last year after Vucic said he would not bow to a single demand even “if there were 5 million of you.” Other popular protest slogans include “Wake Up”, “Rise”, “It Has Started” and “I’m Not Stupid”.

“We try to inform people (via social media) about the abuses of Vučić and his associates, announce further protests, and often in desperate ways, try to keep the good spirit of the people,” says Trifunovic.

“That is why we are often targeted by tabloids owned by the Serbian Progressive Party. Currently, social media is the only means of (distributing) information. Therefore, it is absolutely necessary.”

The core demands of the opposition movement include; free and fair elections, media freedom, reporting of the protests by the Serbian public broadcaster (RTS), a space for opposition voices on TV with national coverage and an end to political violence. Protesters are leveraging social media in order to try and break state control over what remains the most widely consumed media source: TV.

The response from Vučić has been a careful one and he rejects claims he has become autocratic. Vučić has challenged the opposition by calling for early elections in the spring. However, opposition groups have expressed a reluctance to take part in any snap election under what they consider to be unfair conditions.

Serbia’s opposition parties remain badly divided, though many of their leaders and members have joined the protests regardless. Students, academics, and prominent writers, actors and singers have also joined.


Rules For Legal, Wine-Focused Social Media Posts Are Changing

The legal regulations for social media posts in the wine industry have long been quite rigid. Many of these rulings date back to the repeal of Prohibition, in 1933, and the establishment of a three-tier liquor sales system that was created to theoretically not give undo control to any one specific sales tier.

The upshot of this has included incredibly controversial legal perspectives about what type of social media promotions might favor one sales outlet, or retailer, over another venue. So directly promoting wine dinners at one restaurant or wine tastings at another retailer has long been officially illegal, even if most wineries didn’t always comply.

Photocredit: GettyGETTY

Fortunately, these laws are changing and becoming more permissive, in a way that should benefit both retailers and wineries. California drinks attorney Rebecca Stamey-White, a partner in the firm of the San Francisco-based firm of Hinman & Carmichael, shared some of the inside story about how—and why—these laws are changing in one of her firm’s recent newsletters.

How the Law is Evolving

Recently the state of California has expanded an alcohol supplier’s ability to advertise events at retail locations starting this year. According to an assembly bill issued last fall suppliers of wine and spirits may post photos, however strangely, no video, of the retail premises, where they host events, on social media in connection with a range of events, according to Stamey-White. According to John Hinman, another partner at the same the firm, these laws went into effect this year.

Stamey-White says laws in California are changing. Photo credit: Hinman & Carmichael.HINMAN AND CARMICHAEL

These new laws allow wineries to post pictures of winemaker dinners, tastings at retailer venues with specific licenses and tastings at on-premise venues. She added, in the newsletter, that, “Previously, the law only allowed suppliers to mention the location of the event and basic details, but you couldn’t explicitly include pictures of the retail premises, so it made it difficult [or folks had to take the risk] to actually promote events on Instagram, Facebook and on [other] websites. Now it’s explicitly permissible to advertise these events with photos, which is helpful, since every social media platform these days focuses on photos and video,” she adds.

“The rule of thumb that many brands have been following…and telling their agencies to follow…. is to include references to at least three accounts in social media posts. With social media dominating as a source of info on ‘what’s happening, where?’ it requires some creativity to make that content relevant and not sound foolish when you’re having an event at just one store or bar, for example,” shares Steve Raye, the director of Bevology, a Simsbury, Connecticut-based drinks consulting firm.

He adds that, “The good news though is that now we know the rules. Where before everyone was guessing where the line was. Nobody wants to cross it, but it’s been a bit of a grey area than an actual line. Looking forward, social media is a juggernaut that beverage-alcohol legislation and judicial rulings are falling way behind.”

What Changed this Year

The laws, relative to social media, have changed as of this year. According to Stamey-White, “California expanded an alcohol supplier’s ability to advertise events at retail locations starting this year.”

“Basically California had made it very hard for wineries to promote events over social media that had any reference to a retailer. There were even some citations for an event that had been run annually in Sacramento, I guess it was close enough to the regulators for them to see it,” shared Jon Moramarco, a partner in number-crunching firm of Gomberg & Fredrikson.

“Posting pictures of the location where your wines are being poured is a far cry from anything anti-competitive. This is a step in the right direction,” shares Rabobank’s New York City-based global strategist for beverages Stephen Rannekleiv.

Moramarco adds that, “This prohibition made it very hard for wineries to use social media the way most people use it without thinking. This will make it easier to have a natural social media program around events but as John [Hinman] notes wineries will still need to have good practices to avoid the remaining pitfalls.”

In terms of social media, wine-focused concerns in other states, Moramarco says that they are minor. “The short answer is that suppliers posting photos of tasting events have not been a big priority in other states. No other state has been so explicit about social media advertising, Tied House [laws] and experiential events as California,” Stamey-White adds.

“We also recommend implementing a well-designed, social media policy, with guidance about how to engage with retailers and events on retail premises specifically,” she concludes.